United States
                Environmental Protection
                Agency
                  Office of Research and
                  Development
                  Washington DC 20460
EPA/600/P-99/002bB
March 2001
Second External Review Draft
?/EPA
JS EPA 0«ce c' 3esf»-eh and Ocv: oi
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-99/002bB
                                                 March 2001
                                     Second External Review Draft
  Air Quality Criteria for
       Participate 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.
        U.S. Environmental Protection Agency
        Regions, Library (PL-12J)
        77 West Jackson Boulevard, 12th Floor
        Chicago, 1L  60604-3590
     National Center for Environmental Assessment
         Office of Research and Development
        U.S. Environmental Protection Agency
          Research Triangle Park, NC 27711
                                       Printed on Recycled Paper

-------
                                    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.
March 2001                              Il-ii        DRAFT-DO NOT QUOTE OR CITE

-------
                                        Preface

      National Ambient Air Quality Standards (NAAQS) are promulgated by the United States
 Environmental Protection Agency (EPA) to meet requirements set forth in Sections 108 and 109
 of the U.S. Clean Air Act (CAA). Sections 108 and 109 require the EPA Administrator (1) to list
 widespread air pollutants that reasonably may be expected to endanger public health or welfare;
 (2) to issue air quality criteria for them that assess the latest available scientific information on
 nature and effects of ambient exposure to them; (3) to set "primary" NAAQS to protect human
 health with adequate margin of safety 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 years) review and revise, as appropriate, the criteria and NAAQS for
 a given listed pollutant or class of pollutants.
      The original NAAQS for particulate matter (PM), issued in 1971 as "total suspended
 particulate" (TSP) standards, were revised in  1987 to focus on protecting against human health
 effects associated with exposure to ambient PM less than 10 microns (< 10 /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 Aig/m3, 24-h; 50 //g/m3, annual average) were
 retained in modified form and new standards (65 //g/m3, 24-h; 15 (J-g/m3,  annual average) for
 particles <2.5 fj.m (PM2 5) were promulgated in July 1997.
      This Second External Review Draft of revised Air Quality Criteria for Particulate Matter
 assesses new scientific  information that has become available mainly between early 1996 through
 December 2000. The present draft is being released for public comment and review by the Clean
 Air Scientific Advisory Committee (CASAC) to obtain comments on the organization and
 structure of the document, the issues addressed, the approaches employed in assessing and
 interpreting the newly available information on PM exposures and effects, and the key findings
 and conclusions arrived at as a consequence of this assessment. Extensive additional pertinent
 information is expected to be published during the next 6 to 9 mo (including results from a  vastly
expanded EPA PM Research program and from other federal and state agencies, as well as  other
March 2001                                II-iii        DRAFT-DO NOT QUOTE OR CITE

-------
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. Public comments and
CASAC review recommendations will be taken into account, along with any pertinent newly
available information published or accepted for peer-reviewed publication by May/June 2001, in
making any appropriate further revisions to this document for incorporation into a Third External
Review Draft.  That draft is expected to be released in September/October, 2001 for further
public comment and CASAC review (December 2001) in time for a final version to be
completed by early 2002. Evaluations contained in the present document will be drawn on to
provide inputs to associated PM Staff Paper analyses prepared by EPA's Office of Air Quality
Planning and Standards (OAQPS) to pose options for consideration by the EPA Administrator
with regard to proposal and, ultimately, promulgation of decisions on potential retention or
revision of the current PM NAAQS.
     Preparation of this document was coordinated by staff of EPA's National Center for
Environmental Assessment in Research Triangle Park (NCEA-RTP).  NCEA-RTP scientific
staff, together with experts from other EPA/ORD laboratories and academia, contributed to
writing of document chapters, and earlier drafts of this document were 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. The
document describes the nature, sources, distribution, measurement, and concentrations of PM in
outdoor (ambient) and indoor environments. It also evaluates the latest data on human exposures
to ambient PM and consequent health effects in exposed human populations (to support decision
making regarding primary, health-related PM NAAQS).  The document also evaluates ambient
PM environmental effects on vegetation and ecosystems, visibility, and man-made materials, as
well as atmospheric PM effects on climate change processes associated with alterations in
atmospheric transmission of solar radiation or its reflectance from the Earth's surface or
atmosphere (to support decision making on secondary PM NAAQS).
     The NCEA of EPA acknowledges the contributions provided by authors, contributors, and
reviewers and the diligence of its staff and contractors in the preparation of this document.
March 2001                              II-iv        DRAFT-DO NOT QUOTE OR CITE

-------
                Air Quality Criteria for Particulate Matter
                                 VOLUME I
    1.  INTRODUCTION 	1-1

    2.  PHYSICS, CHEMISTRY, AND MEASUREMENT OF
       PARTICULATE MATTER	2-1

    3.  CONCENTRATIONS, SOURCES, AND EMISSIONS OF
       ATMOSPHERIC PARTICULATE MATTER 	3-1
       Appendix 3 A:  Organic Composition of Particulate Matter	  3A-1
       Appendix 3B:  Composition of Particulate Matter Source Emissions	3B-1

    4.  ENVIRONMENTAL EFFECTS OF PARTICULATE MATTER 	4-1
       Appendix 4A:  Excerpted Key Points from the Executive Summary of
                    the World Meteorological Organization 1998 Assessment
                    of Stratospheric Ozone Depletion	  4A-1
       Appendix 4B:  Excerpted Key Points from the Executive Summary of
                    the United Nations Environment Programme 1998
                    Assessment of Environmental Effects of Ozone
                    Depletion	4B-1
       Appendix 4C:  Excerpted Key Points from the Executive Summary of
                    the Special Report of the Intergovernmental Panel on
                    Climate Change Working Group II on the Regional
                    Impacts of Climate Change:  An Assessment of
                    Vulnerability	4C-1
       Appendix 4D:  Excerpted Materials from the U.S. Global Change
                    Research Program Assessment Overview Report on
                    Climate Change Impacts on the United States and
                    Subsidiary Regional Assessment Reports	  4D-1
       Appendix 4E:  Recent Model Projections of Excess Mortality Expected
                    in U.S. Cities During Summer and Winter Seasons
                    Because of Future Climate Change, Based on Kalkstein
                    and Greene (1997)	4E-1

    5.  HUMAN EXPOSURE TO PARTICULATE MATTER AND ITS
       CONSTITUENTS  	5-1
March 2001                           II-v         DRAFT-DO NOT QUOTE OR CITE

-------
               Air Quality Criteria for Particulate Matter
                                (cont'd)
                              VOLUME II


   6.  EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS FROM
      AMBIENT PARTICULATE MATTER  	6-1
      Appendix 6A: Demographic and Pollution Data for 90-City Analysis
                  of NMMAPS Project	  6A-1
      Appendix 6B: Heart Rate Variability as a Predictor of Serious Cardiac
                  Outcomes	6B-1

   7.  DOSIMETRY OF PARTICULATE MATTER	7-1

   8.  TOXICOLOGY OF PARTICULATE MATTER  	8-1

   9.  INTEGRATIVE SYNTHESIS: PARTICULATE MATTER
      ATMOSPHERIC SCIENCE, AIR QUALITY, HUMAN
      EXPOSURE, DOSIMETRY, AND HEALTH RISKS 	9-1
      Appendix 9A: Key Quantitative Estimates of Relative Risk for
                  Particulate Matter-Related Health Effects Based on
                  Epidemiologic Studies of North American Cities
                  Assessed in the 1996 Particulate Matter Air Quality
                  Criteria Document	  9A-1

   EXECUTIVE SUMMARY	E-1
March 2001                         Il-vi        DRAFT-DO NOT QUOTE OR CITE

-------
                                 Table of Contents
 List of Tables	II-xiv
 List of Figures  	II-xviii
 Authors, Contributors, and Reviewers	H-xxi
 U.S. Environmental Protection Agency Project Team for Development of Air
     Quality Criteria for Particulate Matter	I-xxix

 6.  EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS FROM AMBIENT
    PARTICULATE MATTER	6-1
    6.1   INTRODUCTION	6-1
         6.1.1  Types of Epidemiology Studies Reviewed	6-2
         6.1.2  Confounding and Effect Modification 	6-3
         6.1.3  Selection of Studies for Review	6-5
    6.2   MORTALITY EFFECTS OF PARTICULATE MATTER EXPOSURE  	6-6
         6.2.1  Introduction	6-6
         6.2.2  Mortality Effects of Short-Term Particulate Matter Exposure	6-7
               6.2.2.1  Summary of 1996 Particulate Matter Criteria Document
                       Findings and Key Issues	6-7
               6.2.2.2  Introduction to Newly Available Information	6-11
               6.2.2.3  New Multi-City Studies	6-39
               6.2.2 A  The Role of Particulate Matter Components	6-49
               6.2.2.5  New Assessments of Cause-Specific Mortality	6-71
               6.2.2.6  Salient Points Derived from Summarization of Studies of
                       Short-Term Particulate Matter Exposure Effects on Mortality  . . . 6-75
         6.2.3  Mortality Effects of Long-Term Exposure to Ambient Particulate
               Matter	6-79
               6.2.3.1  Studies Published Prior to the 1996 Particulate Matter
                       Criteria Document	6-79
               6.2.3.2  Prospective Cohort Analyses of Chronic Particulate Matter
                       Exposure Mortality Effects Published Since the 1996
                       Particulate Matter Air Quality Criteria Document	6-82
               6.2.3.3  Studies by Particulate Matter Size-Fraction and Composition  . . . 6-95
               6.2.3.4  Population-Based Mortality Studies in Children	6-101
               6.2.3.5  Shortening-of-Life Associated with Long-Term Ambient
                       Particulate Matter Exposure	6-105
               6.2.3.6  Salient Points Derived from Analyses of Chronic Particulate
                       Matter Exposure Mortality Effects  	6-106
    6.3  MORBIDITY EFFECTS OF PARTICULATE MATTER EXPOSURE	6-109
        6.3.1   Cardiovascular Effects Associated with Acute Ambient Particulate
               Matter Exposure 	,	6-109
               6.3.1.1   Introduction 	6-109
               6.3.1.2   Summary of Key Findings on Cardiovascular Morbidity from
                       the 1996 Particulate Matter Air Quality Criteria Document	6-110

March 2001                             II-vii       DRAFT-DO NOT QUOTE OR CITE

-------
                                Table of Contents
                                      (cont'd)
                                                                                Page
              6.3.1.3  New Particulate Matter-Cardiovascular Morbidity Studies	6-111
              6.3.1.4  Issues in the Interpretation of Acute Cardiovascular Effects
                      Studies	6-140
        6.3.2  Effects of Short-Term PM Exposure on the Incidence of Respiratory
              Hospital Admissions and Medical Visits  	6-141
              6.3.2.1  Introduction	6-141
              6.3.2.2  Summary of Key Respiratory Hospital Admissions Findings
                      from the 1996 Particulate Matter Air Quality Criteria
                      Document	6-142
              6.3.2.3  New Respiratory-Related Hospital Admissions Studies	6-143
              6.3.2.4  Key New Respiratory Medical Visits Studies  	6-178
              6.3.2.5  Summary of Key Findings on Acute Particulate Matter
                      Exposure and Respiratory-Related Hospital Admissions
                      and Medical Visits	6-181
        6.3.3  Effects of Particulate Matter Exposure on Lung Function and
              Respiratory Symptoms 	6-182
              6.3.3.1  Effects of Short-Term Particulate Matter Exposure on Lung
                      Function and Respiratory Symptoms	6-183
              6.3.3.2  Long-Term Particulate Matter Exposure Effects on Lung
                      Function and Respiratory Symptoms	6-204
   6.4  INTERPRETIVE ASSESSMENT OF EPIDEMIOLOGIC DATABASE
        ON HEALTH EFFECTS OF AMBIENT PARTICULATE MATTER	6-216
        6.4.1  Introduction	6-216
        6.4.2  New Assessments of Confounding	6-219
              6.4.2.1  Assessment  of Copollutant Confounding 	6-219
              6.4.2.2  Simulation Analysis of Confounding	6-225
              6.4.2.3  Alternative Approaches to Deal with Confounding	6-226
        6.4.3  Role of Particulate Matter Components  	6-228
              6.4.3.1  Fine- and Coarse-Particle Effects on Mortality	6-228
              6.4.3.2  Fine- and Coarse-Particulate Matter Effects on Morbidity	6-233
        6.4.4  The Question of Lags 	6-238
        6.4.5  New Assessments of Mortality Displacement  	6-243
        6.4.6  New Assessment of Threshold in Concentration-Response
              Relationships	6-245
        6.4.7  New Theoretical Assessments of Consequences of Measurement
              Error 	6-248
        6.4.8  New Assessment of Methodological Issues 	6-255
              6.4.8.1  Time Series Model Specification	6-255
              6.4.8.2  Case-Crossover Study Design	6-256
        6.4.9  Heterogeneity of Particulate Matter Effects Estimates	6-258
March 2001                              II-viii        DRAFT-DO NOT QUOTE OR CITE

-------
                                Table of Contents
                                      (cont'd)
               6.4.9.1   Evaluation of Heterogeneity of Particulate Matter Mortality
                       Effect Estimates	6-260
               6.4.9.2   Comparison of Spatial Relationships in the NMMAPS and
                       Cohort Reanalyses Studies	6-264
    6.5  KEY FINDINGS AND CONCLUSIONS DERIVED FROM PARTICULATE
        MATTER EPIDEMIOLOGY STUDIES  	6-266
    REFERENCES  	6-271

    Appendix 6A:  Demographic and Pollution Data for 90-City Analysis of
                 NMMAPS Project	  6A-1
    REFERENCES  	  6A-12

    Appendix 6B:  Heart Rate Variability as a Predictor of Serious Cardiac Outcomes	6B-1
    REFERENCES  	6B-6

7.  DOSIMETRY OF PARTICULATE MATTER	7-1
    7.1  INTRODUCTION	7-1
        7.1.1   Size Characterization of Inhaled Particles 	7-2
        7.1.2  Structure of the Respiratory Tract	7-4
    7.2  PARTICLE DEPOSITION	7-4
        7.2.1   Mechanisms of Deposition	7-5
        7.2.2  Deposition Patterns in the Human Respiratory Tract	7-6
               7.2.2.1   Total Respiratory Tract Deposition	7-7
               7.2.2.2   Deposition in the Extrathoracic Region	7-9
               7.2.2.3   Deposition in the Tracheobronchial and Alveolar Regions	7-12
               7.2.2.4   Local Distribution of Deposition	7-12
               7.2.2.5   Deposition of Specific Size Modes of Ambient Aerosol	7-14
        7.2.3   Biological Factors Modulating Deposition	7-16
               7.2.3.1   Gender	7-16
               7.2.3.2   Age	7-18
               7.2.3.3   Respiratory Tract Disease  	7-20
               7.2.3.4   Anatomical Variability	7-23
        7.2.4   Interspecies Patterns of Deposition	7-24
    7.3  PARTICLE CLEARANCE AND TRANSLOCATION	7-29
        7.3.1   Mechanisms and Pathways of Clearance  	7-29
               7.3.1.1  Extrathoracic Region	7-31
               7.3.1.2  Tracheobronchial Region	7-32
               7.3.1.3  Alveolar Region	7-32
        7.3.2   Clearance Kinetics  	7-34
               7.3.2.1  Extrathoracic Region	7-34
               7.3.2.2  Tracheobronchial Region	7-34

March 2001                              Il-ix        DRAFT-DO NOT QUOTE OR CITE

-------
                             Table of Contents
                                  (cont'd)
             7.3.2.3  Alveolar Region	7-36
       7.3.3  Interspecies Patterns of Clearance  	7-38
       7.3.4  Biological Factors Modulating Clearance	7-39
             7.3.4.1  Age	7-40
             7.3.4.2  Gender	7-40
             7.3.4.3  Physical Activity  	7-40
             7.3.4.4  Respiratory Tract Disease  	7-40
   7.4  PARTICLE OVERLOAD	7-42
   7.5  COMPARISON OF DEPOSITION AND CLEARANCE PATTERNS
       OF PARTICLES ADMINISTERED BY INHALATION AND
       INTRATRACHEAL INSTILLATION	7-43
   7.6  MODELING THE DISPOSITION OF PARTICLES IN THE RESPIRATORY
       TRACT	7-45
       7.6.1  Modeling Deposition and Clearance	7-45
       7.6.2  Models To Estimate Retained Dose	7-51
   REFERENCES 	7-53

8.  TOXICOLOGY OF PARTICULATE MATTER 	8-1
   8.1  INTRODUCTION	8-1
   8.2  RESPIRATORY EFFECTS OF PARTICULATE MATTER IN
       HEALTHY HUMANS AND LABORATORY ANIMALS:
       IN VIVO EXPOSURES 	8-2
       8.2.1  Acid Aerosols	8-3
       8.2.2  Metal, Particles, Fumes,  and Smoke	8-4
       8.2.3  Ambient Combustion-Related and Surrogate Particle Matter  	8-9
       8.2.4  Ambient Bioaerosols	8-23
   8.3  SYSTEMIC EFFECTS OF PARTICULATE MATTER IN HEALTHY
       HUMANS AND LABORATORY ANIMALS:  IN VIVO EXPOSURES  	8-25
   8.4  SUSCEPTIBILITY TO THE EFFECTS OF PARTICULATE MATTER
       EXPOSURE  	8-34
       8.4.1  Effects of Particulate Matter on Cardiopulmonary Compromised
             Hosts	8-35
       8.4.2  Genetic Susceptibility to Inhaled Particles	8-39
       8.4.3  Effect of Particulate Matter on Allergic Hosts	8-41
       8.4.4  Resistance to Infectious Disease	8-45
   8.5  MECHANISMS OF PARTICULATE MATTER TOXICITY AND
       PATHOPHYSIOLOGY: IN VITRO EXPOSURES 	8-47
       8.5.1  Introduction	8-47
       8.5.2  Experimental Exposure Data  	8-47
             8.5.2.1  Ambient Particles	8-48
             8.5.2.2  Residual Oil Fly Ash 	8-56

March 2001                           II-x       DRAFT-DO NOT QUOTE OR CITE

-------
                                Table of Contents
                                      (cont'd)
                                                                               age
         8.5.3   Potential Cellular and Molecular Mechanisms	8-58
               8.5.3.1  Reactive Oxygen Species	8-58
               8.5.3.2  Intracellular Signaling Mechanisms	8-63
               8.5.3.3  Other Potential Cellular and Molecular Mechanisms  	8-67
         8.5.4   Specific Particle Size and Surface Area Effects	8-68
         8.5.5   Pathophysiological Mechanisms for the Effects of Low
               Concentrations of Particulate Air Pollution  	8-72
               8.5.5.1  Direct Pulmonary Effects	8-72
               8.5.5.2  Systemic Effects Secondary to Lung Injury	8-74
               8.5.5.3  Direct Effects on the Heart	8-76
    8.6   RESPONSES TO PARTICULATE MATTER AND GASEOUS
         POLLUTANT MIXTURES  	8-77
    8.7   SUMMARY	8-83
         8.7.1   Biological Plausibility	8-83
               8.7.1.1  Link Between Specific Particulate Matter Components and
                      Health Effects	8-83
               8.7.1.2  Susceptibility	8-87
         8.7.2   Mechanisms of Action  	8-88
    REFERENCES  	8-89

9.  INTEGRATIVE SYNTHESIS: PARTICULATE MATTER ATMOSPHERIC
    SCIENCE, AIR QUALITY, HUMAN EXPOSURE, DOSIMETRY, AND
    HEALTH RISKS	9-1
    9.1  INTRODUCTION	9-1
    9.2  ATMOSPHERIC SCIENCE CONSIDERATIONS	9-2
        9.2.1   Ambient Particulate Matter Size Distinctions	9-3
        9.2.2   Fine- and Coarse-Mode Particle Distinctions vis-a-vis Sources,
               Formation Mechanisms, and Atmospheric Behavior	9-7
        9.2.3   Particle Size-Related Distinctions vis-a-vis Number, Surface Area,
               and Mass	9-10
        9.2.4  Nuclei-Mode Particles (Ultrafme Particles)	9-11
    9.3  CHARACTERIZATION OF U.S. AMBIENT PARTICULATE
        MATTER CONCENTRATIONS AND CONTRIBUTING SOURCES
        AND EMISSIONS 	9-15
        9.3.1  Ambient Particulate Matter Measurement Methods	9-15
        9.3.2  Patterns and Trends in U.S. Particulate Matter Concentrations	9-18
              9.3.2.1  PM10 Trends and Concentrations	9-18
              9.3.2.2  PM25 Trends and Concentrations	9-19
              9.3.2.3  Spatial Variability in PM2 5 Concentrations	9-20
              9.3.2.4  Relationships Among Particulate Matter in Different Size
                      Fractions	9-20

March 2001                             Il-xi        DRAFT-DO NOT QUOTE OR CITE

-------
                                                                                 age
                                Table of Contents
                                      (cont'd)
              9.3.2.5  Short-Term Temporal Variability of Particulate Matter
                      Concentrations	9-21
        9.3.3  Sources of Particulate Matter	9-21
   9.4   HUMAN EXPOSURES TO AMBIENT PARTICULATE MATTER	9-24
   9.5   DOSIMETRY CONSIDERATIONS  	9-26
        9.5.1  Particle Deposition in the Respiratory Tract	9-28
        9.5.2  Particle Clearance and Translocation	9-32
        9.5.3  Deposition and Clearance Patterns of Particles Administered by
              Inhalation Versus Intratracheal Instillation	9-34
        9.5.4  Inhaled Particles as Potential Carriers of Toxic Agents	9-35
   9.6   HEALTH EFFECTS OF AMBIENT PARTICULATE MATTER 	9-36
        9.6.1  Introduction	9-36
        9.6.2  Community-Health Epidemiologic Evidence for Ambient Particulate
              Matter Effects	9-38
              9.6.2.1  Short-Term Particulate Matter Exposure Effects on Mortality . .  . 9-40
              9.6.2.2  Updated Epidemiologic Findings for Long-Term Particulate
                      Matter Exposure Effects on Mortality	9-61
              9.6.2.3  Relationships of Ambient Particulate Matter Concentrations
                      to Morbidity Outcomes  	9-67
              9.6.2.4  Methodological Issues	9-80
        9.6.3  Coherence of Reported Epidemiologic Findings  	9-87
        9.6.4  Toxicologic Insights on Biological Plausibility	9-89
              9.6.4.1  Mechanisms of Action	9-89
              9.6.4.2  Links Between Specific Particulate Matter Components and
                      Health Effects	9-95
   9.7   RISK FACTORS AND POTENTIALLY SUSCEPTIBLE POPULATION
        GROUPS	9-99
        9.7.1  Introduction	9-99
        9.7.2  Preexisting Disease as a Risk Factor for Particulate Matter Health
              Effects	9-99
              9.7.2.1  Ambient Particulate Matter Exacerbation of Cardiovascular
                      Disease Conditions	9-99
              9.7.2.2  Ambient Particulate Matter Exacerbation of Respiratory
                      Disease Conditions	9-102
        9.7.3  Aged-Related At-Risk Population Groups: The Elderly and Children ...9-103
   REFERENCES  	9-107
March 2001                              Il-xii       DRAFT-DO NOT QUOTE OR CITE

-------
                                Table of Contents
                                      (cont'd)
    Appendix 9A:  Key Quantitative Estimates of Relative Risk for Particulate
                 Matter-Related Health Effects Based on Epidemiologic Studies
                 of North American Cities Assessed in the 1996 Particulate Matter
                 Air Quality Criteria Document 	  9A-1
    REFERENCES 	  9A-6

 EXECUTIVE SUMMARY	E-1
March 2001                            II-xiii       DRAFT-DO NOT QUOTE OR CITE

-------
                                    List of Tables

Number                                                                          Page

6-1      Short-Term Particulate Matter Exposure Mortality Effects Studies	6-12

6-2      Synopsis of Short-Term Mortality Studies That Examined Relative Importance
         of PM25 Versus PM10_25  	6-51

6-3      Excess Total Mortality Risks Estimated To Be Associated with Various
         Ambient Particle Size-Related Indices	6-61

6-4      Summary of Particulate Matter Chemical Components Analyzed in Recent
         Studies	6-63

6-5      Summary of Source-Oriented Evaluations of Particulate Matter Components
         in Recent Studies	6-68

6-6      Comparison of Six Cities and American Cancer Society Study Findings from
         Original Investigators and Health Effects Institute Reanalysis	6-84

6-7      Relative Risk of Mortality from All Nonexternal Causes, by Sex and Air
         Pollutant, for an Alternative Covariate Model in the ASHMOG Study	6-88

6-8      Relative Risk of Mortality from Cardiopulmonary Causes, by Sex and  Air
         Pollutant, for an Alternative Covariate Model  	6-89

6-9      Relative Risk of Mortality from Lung Cancer by Air Pollutant and by Gender
         for an Alternative Covariate Model  	6-90

6-10     Comparison of Reported PM10, PM25, and PM10_25 IQR Relative Risks for
         Various Mortality Causes in a Male Subset of the AHSMOG Study for
         One-Pollutant Models	6-92

6-11     Comparison of AHSMOG, Six Cities, and American Cancer Society Study
         Findings	6-93

6-12     Comparison of Estimated Relative Risks for All-Cause Mortality in  Six U.S.
         Cities Associated with the Reported Inter-City Range of Concentrations of
         Various Particulate Matter Metrics	6-95

6-13     Comparison of Reported SO4= and PM2 5 Relative Risks for Various  Mortality
         Causes in the American Cancer Society Study	6-96

6-14     Comparison of Total Mortality Relative Risk Estimates and T-Statistics for
         Particulate Matter Components in Three Prospective Cohort Studies	6-97

March 2001                             II-xiv        DRAFT-DO NOT QUOTE OR CITE

-------
                                   List of Tables
                                       (cont'd)

 Number                                                                         Page

 6-15     Comparison of Cardiopulmonary Mortality Relative Risk Estimates and
         T-Statistics for Participate Matter Components in Three Prospective Cohort
         Studies	6-98

 6-16     Acute Particulate Matter Exposure and Cardiovascular Hospital Admissions .... 6-112

 6-17     Acute Particulate Matter Exposure and Respiratory Medical Visits and
         Hospital Admissions Studies  	6-144

 6-18     Percent Increase in Hospital Admissions per 10-yUg/m3 Increase in PM10 in
         14 U.S. Cities 	6-173

 6-19     Short-Term Particulate Matter Exposure Effects on Pulmonary Function
         Tests in Studies of Asthmatics 	6-185

 6-20     Short-Term Particulate Matter Exposure Effects on Symptoms in Studies
         of Asthmatics 	6-189

 6-21     Short-Term Particulate Matter Exposure Effects on Pulmonary Function
         Tests in Studies of Nonasthmatics 	6-195

 6-22     Short-Term Particulate Matter Exposure Effects on Symptoms in Studies
         of Non-Asthmatics 	6-201

 6-23     Long-Term Particulate Matter Exposure:  Respiratory Symptom, Lung
         Function, and Biomarker Effects  	6-206

 6-24     Comparison of PM10 Effect Sizes Estimated by NMMAPS Analyses for 0, 1,
         and 2 Day Lags for the 20 Largest U.S. Cities  	6-241

 6A-1     The 90 Cities and Their Included Counties by Population Size with Mean
         Daily Number of Deaths by Category (1987-1994) 	 6A-2

 6A-2     Mean Daily Pollution Levels by City (1987-1994)	 6A-6

 6A-3    Number of Days for Which Monitoring Was Available by Pollutant for Cities
         (1987-1994)	 6A-9

6B-1    Terms Used in Expressing Heart Rate Variability	6B-4
March 2001                              II-xv        DRAFT-DO NOT QUOTE OR CITE

-------
                                    List of Tables
                                        (cont'd)

Number                                                                          page

7-1      Overview of Respiratory Tract Particle Clearance and Translocation
         Mechanisms 	7-30

8-1      Respiratory Effects of Acid Aerosols in Humans and Laboratory Animals	  8-5

8-2      Respiratory Effects of Metal Particles, Fumes, and Smoke in Humans and
         Laboratory Animals  	8-7

8-3      Respiratory Effects of Ambient Particulate Matter	8-11

8-4      Respiratory Effects of Complex Combustion-Related
         Particulate Matter	8-13

8-5      Respiratory Effects of Surrogate Particulate Matter	8-18

8-6      Respiratory Effects of Ambient Aerosols	8-24

8-7      Cardiovascular Effects and Other Systemic Effects of Particulate Matter	8-26

8-8      hi Vitro Effects of Particulate Matter and  Particulate Constituents	8-49

8-9      Numbers and Surface Areas of Monodisperse Particles of Unit Density
         of Different Sizes at a Mass Concentration of 10 /ug/m3  	8-69

8-10     Respiratory and Cardiovascular Effects of Mixtures 	8-78

9-1      Comparison of Ambient Particles, Fine and Coarse Mode	9-8

9-2      Constituents of Atmospheric Particles and Their Major Sources	9-22

9-3      Effect Estimates per Variable Increments  in 24-Hour Concentrations of Fine
         Particle Indicators from U.S. and Canadian Studies	9-47

9-4      Effect Estimates per Variable Increments  in 24-Hour Concentrations of
         Coarse-Fraction Particles from U.S. and Canadian Studies	9-53

9-5      Summary of Source-Oriented Evaluations of Particulate Matter Components
         in Recent Studies   	9-59

9-6      Effect Estimates per Increments in Long-Term Mean Levels of Fine and
         Inhalable Particle Indicators from U.S. and Canadian Studies	9-62

March 2001                              II-xvi       DRAFT-DO NOT QUOTE OR CITE

-------
                                    List of Tables
                                        (cont'd)

 Number                                                                         Page

 9-7      Percent Increase in Hospital Admissions per 10-yUg/m3 Increase in 24-Hour
         PM10 in 14 U.S. Cities	9-70

 9-8      Percent Increase in Mortality per 10 Mg/m3 PMIO in Seven U.S. Regions 	9-83

 9-9      Incidence of Selected Cardiorespiratory Disorders by Age and by Geographic
         Region, 1996	9-100

 9-10     Number of Acute Respiratory Conditions per 100 Persons per Year, by Age:
         United States, 1996	9-103

 9A-1     Effect Estimates per 50-yUg/m3 Increase in 24-Hour PM10 Concentrations
         from U.S. and Canadian Studies	  9A-2

 9A-2     Effect Estimates Per Variable Increments in 24-Hour Concentrations of Fine
         Particle Indicators from U.S. and Canadian Studies	  9A-4

 9A-3     Effect Estimates per Increments in Annual Mean Levels of Fine Particle
         Indicators from U.S. and Canadian  Studies 	  9A-5
March 2001                             II-xvii       DRAFT-DO NOT QUOTE OR CITE

-------
                                    List of Figures

Number                                                                             Page

6-1      Estimated excess risks for particulate matter mortality (1 day lag) for the
         90 largest U.S. cities as shown in the original NMAPS report	6-41

6-2      Map of the United States showing the 90 cities and the seven regions
         considered in the NMMAPS geographic analyses  	6-42

6-3      Percent excess mortality risk (lagged 0, 1, or 2 days) estimated in the NMMAPS
         90-City Study to be associated with 10-yUg/m3 increases in PM10 concentrations
         in cities aggregated within U.S. regions shown in Figure 6-2  	6-43

6-4      Percent excess risks estimated per 25-^g/m3 increase in PM2 5 or PM10.2 5
         from new studies evaluating both PM2 5 and PM10_2 5 data for multiple years	6-53

6-5      Excess risks estimated for sulfate per 5-/^g/m3 increase from the studies in
         which both PM2 5 and PM,0_2 5 data were available	6-66

6-6      Acute cardiovascular hospitalizations and particulate matter exposure excess
         risk estimates derived from selected U.S. PM10 studies	6-134

6-7      Maximum excess risk of respiratory-related hospital admissions and visits
         per 50-//g/m3 PM|0 increment in selected studies of U.S. cities	6-182

6-8      Selected acute pulmonary function change studies of asthmatic children 	6-193

6-9      Odds ratios with 95% confidence interval for cough per 50-/ug/m3 increase
         in PM10 for selected asthmatic children studies at lag 0	6-194

6-10     Marginal posterior distributions for effect of PM,0 on total mortality at lag 1
         with and without control for other pollutants, for the 90 cities  	6-221

6-11     Marginal posterior distribution  for effects of PM10 on all cause mortality at
         lag 0,  1, and 2 for the 90 cities  	6-240

6-12     The U. S. Environmental Protection Agency-derived plot showing
         relationship of PMIO total mortality effects estimates and 95% confidence
         intervals for all cities in the Samet et al. (2000a,b) NMMAPS 90-cities
         analyses in relation to study size	6-261
March 2001                              II-xviii       DRAFT-DO NOT QUOTE OR CITE

-------
                                    List of Figures
                                         (cont'd)

 Number                                                                            Page

 6-13    The U. S. Environmental Protection Agency-derived plots showing
         relationships of PM10-mortality (total, nonaccidental) effects estimates
         and 95% confidence intervals to study size (defined as in Figure 6-10)
         for cities broken out by regions as per the NMMAPS regional analyses of
         Samet et al.  (2000a,b)	6-262

 7-1      Total deposition data (percentage deposition of amount inhaled) in humans
         as a function of particle size	7-8

 7-2      Major clearance pathways for particles deposited in the extrathoracic region
         and tracheobronchial tree 	7-30

 7-3      Diagram of known and suspected clearance pathways for poorly soluble
         particles depositing in the alveolar region  	7-31

 9-1      Volume size distribution, measured in traffic, showing fine- and coarse-mode
         particles and the nuclei and accumulation modes within the fine-particle mode .... 9-5

 9-2      An idealized distribution of ambient particulate matter showing fine- and
         coarse-mode particles and the fractions collected by size-selective samplers  	9-6

 9-3      Number of particles as a function of particle diameter:  number concentrations
         are shown on a logarithmic scale to display the wide range by site and size, and
         number concentrations for the average urban distribution are shown on a
         linear scale for which the area under any part of the curve is proportional to
         particle number in that size range	9-11

 9-4      Particle volume distribution as a function of particle diameter:  for the
         averaged rural and urban-influenced rural number distributions shown in
         Figure 9-3 and a distribution from south central New Mexico, and for the
         averaged urban and freeway-influenced urban number distributions shown
         in Figure 9-3 	9-12

 9-5      Distribution of coarse, accumulation, and nuclei- or ultrafine-mode particles
         by three characteristics: (1) number, (2) surface area, and (3) volume for the
         grand average continental size distribution	9-13

 9-6      Total human respiratory tract  deposition as a function of particle size  	9-29

9-7      Percent excess risks estimated per 25-/ug/m3 increase in PM2 5 or PM10_2 5 from
         new studies evaluating both PM2 5 and PM10_2 5 data for multiple years	9-56

March 2001                               II-xix        DRAFT-DO NOT QUOTE OR CITE

-------
                                    List of Figures
                                         (cont'd)

Number                                                                            page

9-8      Relative risks estimated per 5-^g/m3 increase in sulfate from U.S. and
         Canadian studies in which both PM2 5 and PM10_2 5 data were available	9-57

9-9      Acute cardiovascular hospitalizations and particulate matter exposure excess
         risk estimates derived from selected U.S. PM10 studies	9-69

9-10     Maximum excess risk in selected studies of U.S. cities relating PM10 estimate
         of exposure to respiratory-related hospital admissions and visits  	9-74

9-11     Selected acute pulmonary function change studies of asthmatic children 	9-77

9-12     Odds ratios for cough for a 50-/^g/m3 increase in PM10 for selected asthmatic
         children studies, with lag 0 with 95% confidence interval	9-78

9-13     Marginal posterior distributions for effect of PM10 on total mortality at lag 1,
         with and without control for other pollutants, for the 90 cities  	9-82

9-14     Inhalation rates on a per body-weight basis for males and females by age  	9-105
March 2001                               II-xx        DRAFT-DO NOT QUOTE OR CITE

-------
                      Authors, Contributors, and Reviewers
        CHAPTER 6. EPIDEMIOLOGY OF HUMAN HEAL TH EFFECTS FROM
                         AMBIENT PARTICULATE MATTER
 Principal Authors

 Dr. Lester D. Grant—National Center for Environmental Assessment (MD-52),
 U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

 Dr. Vic Hasselblad—29 Autumn Woods Drive, Durham, NC 27713

 Dr. Kazuhiko Ito—New York University Medical Center, Institute of Environmental Medicine,
 Long Meadow Road, Tuxedo, NY  10987

 Dr. Patrick Kinney, Columbia University, 60 Haven Avenue, B-1, 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
March 2001                            Il-xxi       DRAFT-DO NOT QUOTE OR CITE

-------
                     Authors, Contributors, and Reviewers
                                       (cont'd)
Contributors and Reviewers
(cont'd)

Dr. Ralph Delfmo—University of California at Irvine, Epidemiology Division, Department of
Medicine, University of California at Irvine, Irvine, CA 92717

Dr. Douglas Dockery—Harvard School of Public Health, 665 Huntingdon Avenue, 1-1414,
Boston, MA  02115

Dr. Peter Guttorp—University of Washington, Department of Statistics, Box 354322
Seattle, WA  98195

Dr. Scott R. Kegler—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Dr. Fred Lipfert—23 Carll Court, Northport, NY 11768

Dr. Lee-Jane Sally Liu—University of Washington, Department of Environmental Health,
Box 357234, Seattle, WA 98195

Dr. Suresh Moolgavakar—Fred Hutchinson Cancer Research Center, 1100 Fairview Avenue,
N-MP 665, Seattle, WA 98109

Dr. Robert D. Morris—Tufts University, 136 Harrison Avenue, Boston, MA 02111

Dr. Lucas Neas—National Health and Environmental Effects Research Laboratory (MD-58),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Dr. James Robins—Harvard School of Public Health, Department of Epidemiology,
Boston, MA  02115

Dr. Isabelle Romieu—Centers for Disease Control (CDC), 4770 Bufford Hwy, NE,
Atlanta, GA  30341

Dr. Mary Ross—Office of Air Quality Planning and Standards (MD-15),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Dr. Lianne Sheppard—University of Washington, Box 357232, Seattle, WA 98195-7232

Dr. Richard L. Smith—University of North Carolina,  Department of Statistics, Box 3260
Chapel Hill, NC 27599
March 2001                             II-xxii       DRAFT-DO NOT QUOTE OR CITE

-------
                      Authors, Contributors, and Reviewers
                                       (cont'd)
 Contributors and Reviewers
 (cont'd)

 Dr. Leonard Stefanski—North Carolina State University, Department of Statistics, Box 8203,
 Raleigh, NC  27695

 Dr. Duncan Thomas—University of Southern California, Preventative Medicine Department,
 1540 Alcazar Street, CH-220, Los Angeles, CA  90033-9987

 Dr. Clarice Weinberg—National Institute of Environmental Health Sciences, P.O. Box 12233,
 Research Triangle Park, NC 27709

 Dr. William Wilson—National Center for Environmental Assessment (MD-52),
 U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

 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. DOSIMETRYOFPARTICULATEMATTER
 Principal Authors

 Dr. Lawrence J. Folinsbee—National Center for Environmental Assessment (MD-52),
 U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

 Dr. Richard Schlesinger—New York University School of Medicine, Department of
 Environmental Medicine, 57 Old Forge Road, Tuxedo, NY 10987

 Dr. James McGrath—National Center for Environmental Assessment (MD-52),
 U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

 Contributors and Reviewers

 Dr. Dan Costa—National Health and Environmental Effects Research Laboratory (MD-82),
 U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

 Dr. Robert Devlin—National Health and Environmental Effects Research Laboratory
 (MD58),U.S. Environmental Protection Agency, Research Triangle  Park, NC 27711
March 2001                             II-xxiii       DRAFT-DO NOT QUOTE OR CITE

-------
                     Authors, Contributors, and Reviewers
                                      (cont'd)
Contributors and Reviewers
(cont'd)

Dr. Kevin Dreher—National Health and Environmental Effects Research Laboratory (MD-82),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Dr. Andrew Ohio—National Health and Environmental Effects Research Laboratory (MD-58D),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Dr. Judith Graham—National Exposure Research Laboratory (MD-75),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Dr. Chong Kim—National Health and Environmental Effects Research Laboratory (MD-58B),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Dr. Hillel Koren—National Health and Environmental Effects Research Laboratory (MD-58A),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

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^21 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

March 2001                             II-xxiv      DRAFT-DO NOT QUOTE OR CITE

-------
                      Authors, Contributors, and Reviewers
                                       (cont'd)
 Contributors and Reviewers
 (cont'd)

 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 8.  TOXICOLOGY OF PARTICULA TE MA TTER
 Principal Authors

 Dr. Lung Chi Chen—New York University School of Medicine, Nelson Institute of
 Environmental Medicine, 57 Old Forge Road, Tuxedo, NY  10987

 Dr. Lawrence J. Folinsbee—National Center for Environmental Assessment (MD-52),
 U.S. Environmental Protection Agency, Research Triangle Park, NC  27711

 Dr. Terry Gordon—New York University Medical Center, Department of Environmental
 Medicine, 57 Old Forge Road, Tuxedo, NY  10987

 Dr. Christine Nadziejko—Department of Environmental Medicine, New York University School
 of Medicine, Tuxedo, NY

 Mr. James Raub—National Center for Environmental Assessment (MD-52),
 U.S. Environmental Protection Agency, Research Triangle Park, NC  27711

 Contributors and Reviewers

 Dr. 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
March 2001                             II-xxv       DRAFT-DO NOT QUOTE OR CITE

-------
                     Authors, Contributors, and Reviewers
                                      (cont'd)
Contributors and Reviewers
(cont'd)

Dr. Judith Graham—National Exposure Research Laboratory (MD-75),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Dr. Chong Kim—National Health and Environmental Effects Research Laboratory (MD-58B),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Dr. Hillel Koren—National Health and Environmental Effects Research Laboratory (MD-58A),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

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. James McGrath—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Dr. Kent Pinkerton—University of California, ITEH, One Shields Avenue, Davis, CA 95616

Dr. Peter J.A. Rombout—National Institute of Public Health and Environmental Hygiene,
Department of Inhalation Toxicology, P.O. Box 1, NL-3720 BA Bilthoven, The Netherlands

Dr. Vanessa Vu—Office of Research and Development, U.S. Environmental Protection Agency
(8601), Waterside Mall, 401 M St. S.W., Washington, DC 20460

March 2001                             II-xxvi      DRAFT-DO NOT QUOTE OR CITE

-------
                      Authors, Contributors, and Reviewers
                                      (cont'd)
         CHAPTER 9. INTEGRA TIVE SYNTHESIS: PARTICULA TE MA TIER
           ATMOSPHERIC SCIENCE, AIR QUALITY, HUMAN EXPOSURE,
                         DOSIMETRY, AND HEALTH RISKS
 Principal Authors

 Dr. Lester D. Grant—National Center for Environmental Assessment (MD-52),
 U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

 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. Allan Marcus—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. Joseph P. Pinto—National Center for Environmental Assessment (MD-52),
 U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

 Mr. James A. Raub—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. Vanessa Vu—Office of Research and Development, U.S. Environmental Protection Agency
 (8601), Waterside Mall, 401 M St. S.W., Washington, DC 20460
March 2001                            II-xxvii      DRAFT-DO NOT QUOTE OR CITE

-------
                     Authors, Contributors, and Reviewers
                                       (cont'd)
                              EXECUTIVE SUMMARY
Principal Authors

Dr. Lester D. Grant—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Ms. Beverly Comfort—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. Lawrence J. Folinsbee—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Dr. J.H.B. Garner—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 Exposure Research Laboratory (MD-56),
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
March 2001                            II-xxviii      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

 Dr. William E. Wilson—Air Quality Coordinator, Physical Scientist, National Center for
 Environmental Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle
 Park,NC  27711

 Dr. Lawrence J. Folinsbee—Health Coordinator,  Chief, Environmental Media Assessment
 Group, National Center for Environmental Assessment (MD-52), U.S. Environmental Protection
 Agency, Research Triangle Park, NC 27711

 Dr. J.H.B. Garner—Ecological Scientist, National Center for Environmental Assessment
 (MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

 Dr. Dennis J. Kotchmar—Project Manager, Medical Officer, National Center for Environmental
 Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC
 27711

 Dr. Robert Chapman—Medical Officer, National Center for Environmental Assessment
 (MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

 Ms. Beverly Comfort—Health Scientist, National Center for Environmental Assessment
 (MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

 Mr. William Ewald—Health Scientist, National Center for Environmental Assessment (MD-52),
 U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

 Dr. David Mage—Physical Scientist, National Center for Environmental Assessment (MD-52),
 U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

 Dr. Allan Marcus—Statistician, National Center for Environmental Assessment (MD-52),
 U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

 Dr. James McGrath—Visiting Senior Health Scientist, National Center for Environmental
 Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC
 27711
March 2001                            II-xxix       DRAFT-DO NOT QUOTE OR CITE

-------
             U.S. ENVIRONMENTAL PROTECTION AGENCY
  PROJECT TEAM FOR DEVELOPMENT OF AIR QUALITY CRITERIA
                       FOR PARTICULATE MATTER
                                     (cont'd)
Scientific Staff
(cont'd)

Dr. Joseph P. Pinto—Physical Scientist, National Center for Environmental Assessment,
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Mr. James A. Raub—Health Scientist, National Center for Environmental Assessment (MD-52),
U. S. Environmental Protection Agency, Research Triangle Park, NC 27711

Technical Support Staff

Mr. Randy Brady—Deputy Directory, National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Mr. Douglas B. Fennell—Technical Information Specialist, National Center for Environmental
Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC
27711

Ms. Emily R. Lee—Management Analyst, National Center for Environmental Assessment
(MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC  27711

Ms. Diane H. Ray—Program Specialist, National Center for Environmental Assessment
(MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Ms. Eleanor Speh—Office Manager, Environmental Media Assessment Branch, National Center
for Environmental Assessment (MD-52), U.S. Environmental Protection Agency, Research
Triangle Park, NC 27711

Ms. Donna Wicker—Administrative Officer, National Center for Environmental Assessment
(MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711

Mr. Richard Wilson—Clerk, National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
March 2001                            II-xxx       DRAFT-DO NOT QUOTE OR CITE

-------
              U.S. ENVIRONMENTAL PROTECTION AGENCY
   PROJECT TEAM FOR DEVELOPMENT OF AIR QUALITY CRITERIA
                        FOR PARTICULATE MATTER
                                      (cont'd)


 Document Production Staff

 Mr. John R. Barton—Document Processing Coordinator, OAO Corporation, 2222 Chapel
 Hill-Nelson Highway, Beta Building, Suite 100, Durham, NC 27713

 Ms. Diane G. Caudill—Graphic Artist, OAO Corporation, 2222 Chapel Hill-Nelson Highway,
 Beta Building, Suite 100, Durham, NC 27713

 Ms. Yvonne A. Harrison—Word Processor, OAO Corporation, 2222 Chapel Hill-Nelson
 Highway, Beta Building, Suite 100, Durham, NC  27713

 Ms. Bettye B. Kirkland—Word Processor, OAO Corporation, 2222 Chapel Hill-Nelson
 Highway, Beta Building, Suite 100, Durham, NC  27713

 Mr. David E. Leonhard—Graphic Artist, OAO Corporation, 2222 Chapel Hill-Nelson Highway,
 Beta Building, Suite 100, Durham, NC 27713

 Ms. Phyllis  H. Noell—Technical Editor, OAO Corporation, 2222 Chapel Hill-Nelson Highway,
 Beta Building, Suite 100, Durham, NC 27713

 Ms. Carolyn T. Perry—Word Processor, OAO Corporation, 2222 Chapel Hill-Nelson Highway,
 Beta Building,  Suite 100, Durham, NC 27713

 Technical Reference Staff

 Mr. R. David Belton—Reference  Specialist, OAO Corporation, 2222 Chapel Hill-Nelson
 Highway, Beta Building, Suite  100, Durham, NC 27713

 Mr. John A. Bennett—Technical Information Specialist, OAO Corporation, 2222 Chapel
 Hill-Nelson Highway, Beta Building, Suite 100, Durham, NC 27713

 Ms. Sandra L. Hughey—Technical Information Specialist, OAO Corporation, 2222 Chapel
 Hill-Nelson Highway, Beta Building, Suite 100, Durham, NC 27713

 Mr. Jian Ping Yu—Reference Retrieval and Database Entry Clerk, OAO Corporation,
 2222 Chapel Hill-Nelson Highway, Beta Building, Suite 100, Durham, NC 27713

Ms. Kun Zhang—Records Management Technician, OAO Corporation, 2222 Chapel Hill-Nelson
Highway, Beta Building, Suite 100, Durham, NC 27713
March 2001                           II-xxxi       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 in human populations being associated with ambient levels of PM2 5,
 11      PM10, and other indicators of PM exposure. The numerous more recent epidemiologic studies
 12     reviewed in this chapter generally identify more cities where ambient PM-relationships with
 13      morbidity and mortality have been found and, thereby, both extend the earlier findings and
 14      provide an expanded evidence base that substantiates health effects being associated with
 15      exposures to PM at concentrations currently encountered in the United States.
 16           The epidemiology studies presented here should be considered in combination with the
 17      ambient concentration information presented in Chapter 3, the human exposure studies in
 18      Chapter 5, and the dosimetry and toxicology studies in Chapters 7 and 8. The contribution of the
 19      epidemiology studies is to evaluate associations between health effects and exposures of human
 20      populations to ambient PM and to help identify susceptible subgroups and associated risk factors.
 21      Chapter 9 provides a more detailed interpretive synthesis of information.
 22           This chapter opens with a brief overview of key general features of the several types of
 23      epidemiologic studies assessed in the chapter and a discussion of important general
 24      methodological issues that must be considered in their critical assessment. After this brief
 25      introduction, Section 6.2 assesses studies  of PM effects on mortality. Section 6.3 evaluates
 26      studies of morbidity as a health endpoint.  Section 6.4 then provides  an interpretive assessment of
 27      the overall PM epidemiologic data base in relation to a variety of key issues and potential
28      inferences associated with studies reviewed in Sections 6.2  and 6.3.  The overall key findings and
29      conclusions for this chapter are then summarized in Section 6.5.
30

       March 2001                               6-1         DRAFT-DO NOT QUOTE OR CITE

-------
 1      6.1.1  Types of Epidemiology Studies Reviewed
 2           Definitions of various types of epidemiology studies used here were provided in the 1996
 3      PM AQCD (U.S. Environmental Protection Agency, 1996) and are briefly summarized here.
 4      Briefly, the epidemiology studies are divided into mortality studies and morbidity studies.
 5      Mortality studies evaluating PM effects on total (non-accidental) mortality and cause-specific
 6      mortality have provided the most unambiguous evidence of a clearly adverse endpoint. The
 7      morbidity studies further substantiate PM effects on a wide range of health endpoints, such as:
 8      cardiovascular and respiratory-related hospital admissions, medical visits, reports of respiratory
 9      symptoms, self-medication in asthmatics, changes in pulmonary function tests (PFT), low
10      birthweight infants, etc.
11           The epidemiology strategies most commonly used in PM health studies are of four types, in
12      order of increasing inferential strength as delineated by Rothman and Greenland (1998):
13      (1) ecologic studies; (2) time-series semi-ecologic studies; (3) longitudinal panel and prospective
14      cohort studies; and (4) case-control studies.  All of these are observational studies rather than
15      experimental studies, since participants are not assigned at random to air pollution exposures.
16      In general, the exposure of the participant is not directly observed, and the concentration of
17      airborne particles and other air pollutants at one or more stationary air monitors is used as a
18      proxy for individual  exposure to ambient air pollution.
19           In ecologic studies, the responses are at a community level  (for example, annual mortality
20      rates), as are the exposure indices (for example, annual average particulate matter concentrations)
21      and covariates (for example, the percentage of the population greater than 65 years of age).
22      No individual data is used in the analysis, therefore the relation between health effect and
23      exposure calculated across different communities may not reflect individual-level associations
24      between health outcome and exposure.  The use of proxy measures for individual exposure and
25      covariates or effects  modifiers may also bias the results, and within-city or within-unit
26      confounding may be overlooked.
27           Time series studies are more informative because they allow study of associations between
28      changes in outcomes and changes in exposure indicators preceding or simultaneous with the
29      outcome. The temporal relationship supports a conclusion of a causal relation, even when both
30      the outcome (for example, the number of non-accidental deaths in a city during a day) and the
31      exposure (for example, daily air pollution concentration) are community indices.
        March 2001                                6-2         DRAFT-DO NOT QUOTE OR CITE

-------
  1           Prospective cohort (or panel) studies use data from individuals, including health status,
  2     individual exposure, and individual covariates or risk factors, observed over time.  The
  3     participants in a prospective cohort study are ideally recruited independently of prior health status
  4     or exposure, using a simple or stratified random sample so as to represent a target population, so
  5     that exposure of the participants is known before the health endpoint occurs.  The use of
  6     individual-level data is believed to give prospective cohort studies greater inferential strength
  7     than other epidemiology strategies, but the use of community-level or estimated exposure data
  8     may weaken this advantage, as in time-series studies.
  9           Case-control studies are retrospective studies in that exposure  is determined after the health
 10     endpoint occurs (this is common in occupational health studies).  As Rothman and Greenland
 11     (1998) describe it, "Case-control studies are best understood by defining a source population,
 12     which represents a hypothetical study population in which a cohort study might have been
 13     conducted ... In a case-control study, the cases are identified and their exposure status is
 14     determined just as in a cohort study . . . [and] a control  group of study subjects is sampled from
 15     the entire source population that gives rise to the cases  . . . the cardinal requirement of control
 16     selection is that the controls must be sampled independently of their  exposure status."
 17
 18     6.1.2  Confounding and Effect Modification
 19          A pervasive problem in the analysis of epidemiology data, no matter what design or
 20     strategy, is the unique attribution of the health outcome to the nominal causal agent-airborne
 21      particles in this  document. The health outcomes attributed to particles are not very specific (for
 22     example, mortality in a broad range of ICD-9 categories) and may also be attributable to high or
 23      low temperatures, influenza and other diseases, and/or exposure to gaseous criteria air pollutants.
 24      Many of the other factors can be measured, directly or by proxies. Some of these co-variables
25      are confounders, others are effects modifiers.  The distinctions are important.
26           Confounding is "... a confusion of effects. Specifically, the apparent effect of the
27      exposure  of interest is distorted because the effect of an extraneous factor is mistaken for or
28      mixed with the actual exposure effect (which may be null)."  (Rothman and Greenland, 1998,
29      p. 120). These authors list three criteria for a confounding factor:
30
31
        March 2001                                6-3         DRAFT-DO NOT QUOTE OR CITE

-------
 1           (1)  A confounding factor must be a risk factor for the disease [health effect].
 2           (2)  A confounding factor must be associated with the exposure under study in the source
 3               population (the population at risk from which the cases are derived).
 4           (3)  A confounding factor must not be affected by the exposure or the disease, i.e., it cannot
 5               be an intermediate step in the causal path between the exposure and the disease.
 6     Gaseous criteria pollutants (CO, NO2, SO2, O3) are candidates for confounders since:  (1) all of
 7     these have adverse health effects, with CO more often identified with cardiovascular effects and
 8     the others with respiratory effects (including symptoms and hospital admissions), as part of the
 9     wide spectrum of cardiopulmonary disease also associated with particles; (2) the gaseous criteria
10     pollutants may be associated with particles for several reasons, including (a) common sources,
11     (b) correlated changes in response to wind and weather, and (c) SO2 and NO2 may be precursors
12     to sulfate and nitrate components of ambient particle mixes.
13           A common source, such as combustion of gasoline in motor vehicles emitting CO, NO2,
14     and primary particles, may play an important role in confounding among these pollutants, as does
15     weather and seasonal effects. Even though O3 is a secondary pollutant also associated with
16     emission of NO2, it is often less highly associated with particles.  Levels of SO2 in the western
17     U.S. are often quite low, so that secondary formation of particle sulfates plays a much smaller
18     role and there is usually relatively little confounding of SO2 with PM mass concentration in the
19     west. On the other hand, in the industrial midwest and northeastern states, SO2 and sulfate levels
20     during many of the epidemiology studies were relatively high, and highly correlated with fine
21     particle mass concentrations, so that criterion 3 (no causal path leading from confounder to
22     exposure, or exposure to confounder to health effect) may not be strictly true for SO2 vs sulfate
23     or overall fine particle mass. If there is a causal pathway, then it is not clear whether the
24     observed relation of exposure to health effect is a direct effect of the exposure, an indirect effect
25     mediated by the confounder, or a mixture of these.
26           Most extraneous variables fall into the category of effects modifiers.  "Effect-measure
27     modification differs from confounding in several ways.  The main difference is that, whereas
28     confounding is a bias that the investigator hopes to prevent or remove  from the effect estimate,
29     effect-measure modification is a property of the effect under study ... In epidemiologic analysis
30     one tries to eliminate confounding but one tries to detect and estimate effect-measure
31     modification." (Rothman and Greenland, 1998, p. 254).  Examples of effects modifiers in some

       March 2001                                 6-4         DRAFT-DO NOT QUOTE OR CITE

-------
  1      of the studies evaluated in this chapter include environmental variables (such as temperature or
  2      humidity in time-series studies), individual risk factors (such as education, cigarette smoking
  3      status, age in a prospective cohort study), and community factors (such as percent of population
  4      > 65 years old).  It is often possible to stratify the relationship between health outcome and
  5      exposure by one or more of these risk factor variables.
  6
  7      6.1.3  Selection of Studies for Review
  8           Numerous PM epidemiology papers have been published since the 1996 PM AQCD.
  9      An ongoing medline search has been and is continuing to be conducted in conjunction with other
 10      strategies to identify PM literature pertinent to developing criteria for PM NAAQS. Those
 11       epidemiologic studies that relate measures of ambient PM to human health outcomes are
 12      assessed in this chapter, but occupational exposures studies are not.  Some of the criteria used for
 13       selecting relevant literature for consideration here include whether a given study presents:
 14       (1) pertinent ambient PM indices:  e.g., PM10, fine or coarse fractions of PM10, etc.; (2) analyses
 15       of health effects of specific PM chemical constituents (e.g.,  metals, sulfates, nitrates or ultrafines,
 16       etc.);  (3) health endpoints not previously extensively researched; (4) multiple pollutant analyses;
 17       and/or (5) for long-term effects, mortality displacement information. The publication of
 18      pertinent new studies has been  and is proceeding at a prodigious rate; and the review and
 19      evaluation of pertinent literature in this PM AQCD development process is an ongoing process
 20      which continues to obtain and assess new evidence.  Efforts have been made to assess  here
 21      pertinent new studies published mainly through December, 2000.
 22           In the sections that follow on PM mortality and morbidity effects, key points derived from
 23      the 1996 PM AQCD assessment of then-available information are first concisely highlighted.
 24      The numerous individual new studies that have become available since that prior PM AQCD are
 25      then summarized in tabular form, in which important methodological features and results are
 26      presented. The tables have a uniform general organization with four divisions: (1) information
 27      about  study location and ambient PM levels, (2) study description of methods employed,
28      (3) results and comments and (4) quantitative outcomes for PM measures. For consistency with
29      the prior PM AQCD (U. S. Environmental Protection Agency, 1996), the pollutant increments
30      utilized here to report Relative Risks (RR's) or Odds Ratio for various health effects are:

        March 2001                                6-5         DRAFT-DO NOT QUOTE OR CITE

-------
 1      for PMto, 50 Mg/m3; for PM2 5, 25 //g/m3; for SO4=, 155 nmoles/m3 (=15 //g/m3); and, for H+,
 2      75 nmoles/m3 (=3.6 //g/m3, if as H2SO4).
 3           Greater emphasis is placed in text discussions on integrating and interpreting findings from
 4      the body of evidence provided by the newer studies (as well as relating them to those reviewed in
 5      the 1999 PM AQCD), rather than detailed evaluation of each of the numerous newly available
 6      studies. Particular emphasis is focused in the text on those studies and analyses thought to
 7      provide the most pertinent information for U.S. standard setting purposes. For example, North
 8      American  studies conducted in the U.S. or Canada are generally accorded more text discussion
 9      than those from other geographic regions; and analyses using gravimetric (mass) measurements
10      are generally accorded more text attention than those using non-gravimetric ambient PM
11      measures, e.g., black smoke (BS) or coefficient of haze (COH). Also, more emphasis is placed
12      on text discussion of new multi-city studies that employ standardized methodological analyses
13      for evaluating PM effects across several or numerous cities and often provide overall effects
14      estimates based on combined analyses of information pooled across multiple cities.
15
16
17      6.2 MORTALITY EFFECTS OF PARTICIPATE MATTER EXPOSURE
18      6.2.1 Introduction
19           The relationship of PM and other air pollutants to excess mortality has been intensively
20      studied and has played an important role in previous PM health assessments (U.S. Environmental
21      Protection Agency, 1986, 1996).  Mortality is the most severe adverse health endpoint and, in
22      some ways, the easiest to study. Excellent death records are maintained at every level of
23      government in most all nations and are typically made available to researchers. Also, from a
24      narrowly technical point of view, individual deaths are more amenable to statistical analyses,
25      since individual deaths from natural causes (typically respiratory and cardiovascular diagnoses)
26      are statistically independent except in rare extremely infectious instances.  Individual deaths are
27      also non-recurring events, unlike hospital admissions or respiratory symptoms.
28           Recent findings are evaluated here for the two most important epidemiology designs by
29      which mortality is studied: time-series mortality studies (Section 6.2.2); and prospective cohort
30      studies (Section 6.2.3). The time-series studies mostly assess  acute responses to short-term PM

        March 2001                               6-6        DRAFT-DO NOT QUOTE OR CITE

-------
  1     exposure, although some recent work suggests that time-series data sets are also useful to
  2     examine responses to exposures over a longer time scale. Time-series studies use community-
  3     level air pollution measurements to index exposure and community-level response (i.e., the total
  4     number of deaths each day by age and/or by cause of death). Prospective cohort studies usefully
  5     complement time-series studies; they use individual health records, with survival lifetimes or
  6     hazard rates adjusted for individual risk factors, and typically evaluate human health impacts of
  7     long-term PM exposures indexed by  community-level measurements.
  8
  9     6.2.2  Mortality Effects of Short-Term Particulate Matter Exposure
 10     6.2.2.1  Summary of 1996 Particulate Matter Criteria Document Findings and Key Issues
 11           The time-series mortality  studies reviewed in the 1996 and other past PM AQCD's
 12     provided much evidence that ambient PM air pollution is associated with increases in daily
 13      mortality. The 1996 PM AQCD assessed about 35 PM-mortality time-series studies published
 14     between 1988 and 1996. Information derived from those studies was  consistent with the
 15      hypothesis that PM is a causal agent in the mortality impacts of air pollution.
 16           The PM10 relative risk estimates derived from short-term PM10 exposure studies reviewed
 17      in the 1996 PM AQCD suggested that an increase of 50 /wg/m3 in the 24-h average of PM10 is
 18      most clearly associated with an  increased risk of premature total nonaccidental mortality (total
 19      deaths minus those from accident/injury) on the order of relative risk (RR) = 1.025 to 1.05 in the
 20      general population or, in other words, 2.5 to 5.0% excess deaths per 50 /ug/m3 PM10 increase,
 21      with all statistically significant RR estimates ranging more broadly from 1.015 to 1.085 (i.e.,
 22      1.5 to 8.5% excess risk). Higher relative risks were indicated for the elderly and for those with
 23      pre-existing cardiopulmonary conditions.  Also, based on the then recently published Schwartz
 24      et al. (1996a) analysis of Harvard Six City data, the 1996 PM AQCD found the RR for excess
 25      total mortality in relation to 24-h fine particle concentrations to be in the range of RR = 1.026 to
 26      1.055 per 25 //g/m3 PM2 5 (i.e., 2.6 to  5.5% excess risk per 25 //g/m3 PM2 5).
 27           While numerous studies reported PM-mortality associations, important issues needed to be
 28      addressed in interpreting their findings.  The 1996 PM AQCD extensively discussed most critical
29      issues, including:  (1) seasonal confounding and effect modification; (2) confounding by weather;
30      (3) confounding by co-pollutants; (4)  measurement error; (5) functional form and threshold;
31      (6) harvesting and life shortening; and (7) the role of PM components.  As important issues
        March 2001                               6-7         DRAFT-DO NOT QUOTE OR CITE

-------
 1     related to model specification became further clarified, more studies began to address the most
 2     critical issues, with some having been at least partially resolved, whereas others required still
 3     further investigation.  The next several paragraphs summarize the status of these issues at the
 4     1996 PM AQCD publication time.
 5           One of the most important components in time-series model specification is adjustment for
 6     seasonal cycles and other longer-term temporal trends. Residual over-dispersion and
 7     autocorrelation result from inadequate control for these temporal trends, and not adequately
 8     adjusting for them could result in biased RRs.  Modern smoothing methods allow efficient fits of
 9     temporal trends and minimize such statistical problems. Thus, most recent studies controlled for
10     seasonal and other temporal trends, and it was unlikely that inadequate control for such trends
11     seriously biased estimated PM coefficients.  Effect modification by season was examined in
12     several studies.  Season-specific analyses are often not feasible in small-sized studies (due to
13     marginally significant PM effect size), but some studies (e.g., Samet et al., 1996; Moolgavkar
14     and Luebeck, 1996) suggested that estimated PM coefficients varied from season to season.
15     It was not fully resolved, however, if these results represent real seasonal effect modifications or
16     may be due to varying extent of correlation between PM and co-pollutants or weather variables
17     by season.
18           While most available studies included control for weather variables, some reported
19     sensitivity of PM coefficients to weather model specification, leading some investigators to
20     speculate that inadequate weather model specifications may still have erroneously ascribed
21     residual weather effects to PM. Two PM studies (Samet et al., 1996, 1998; Pope and Kalkstein,
22      1996) involved collaboration with a meteorologist and utilized more elaborate weather modeling,
23     e.g., use of synoptic weather categories.  These studies found that estimated PM effects were
24     essentially unaffected by the synoptic weather variables and also indicated that the synoptic
25     weather model did not provide better model fits in predicting mortality when compared to other
26     weather model specifications used in previous PM-mortality studies.  Thus, these results
27     suggested that the reported PM effects were not explained by weather effects.
28           Many earlier PM studies considered at least one co-pollutant in the mortality regression,
29     and some also examined several co-pollutants. In most cases, when PM indices were significant
30     in single pollutant models, addition of a co-pollutant diminished the PM effect size somewhat,
31     but  did not eliminate the PM associations. When multiple pollutant models were performed by
       March 2001                               6-8         DRAFT-DO NOT QUOTE OR CITE

-------
  1      season, the PM coefficients became less stable, again, possibly due to PM's varying correlation
  2      with co-pollutants among season and/or smaller sample sizes. However, in many studies, PM
  3      indices showed the highest significance (versus gaseous co-pollutants) in single and multiple
  4      pollutant models.  Thus, it was concluded that PM-mortality associations were not seriously
  5      distorted by co-pollutants, but interpretation of the relative significance of each pollutant in
  6      mortality regression as relative causal strength was difficult because of limited quantitative
  7      information on relative exposure measurement/characterization errors among air pollutants.
  8           Measurement error can influence  the size and significance of air pollution coefficients in
  9      time-series regression analyses and is also important in assessing confounding among multiple
 10      pollutants, as varying the extent of such error among the pollutants could also influence the
 11      corresponding relative significance.  The 1996 PM AQCD discussed several types of such
 12      exposure measurement or characterization errors, including site-to-site variability and site-to-
 13      person variability—errors thought to bias the estimated PM coefficients downward in most cases.
 14      However, there was not sufficient quantitative information available to estimate such bias.
 15           The  1996 PM AQCD also reviewed evidence for threshold and various other functional
 16      forms of short-term PM mortality associations.  Several studies indicated that associations were
 17      seen monotonically below the existing PM standards. It was considered difficult, however, to
 18      statistically identify a threshold from available data because of low data density at  lower ambient
 19      PM concentrations, potential influence of measurement error, and adjustments for  other
 20      covariates. Thus, the use of relative risk (rate ratio) derived from the log-linear Poisson models
 21      was considered adequate and appropriate.
 22          The extent of prematurity of death (i.e., mortality displacement, or harvesting) in observed
 23      PM-mortality associations has important public health policy implications.  At the  time of the
 24       1996 PM AQCD review, only a few studies had investigated this issue. While one of the studies
 25      suggested that the extent of such prematurity might be only a few days, this  may not be
 26      generalized because this estimate was obtained for identifiable PM episodes. There was not
 27      sufficient evidence to suggest the extent of prematurity for non-episodic periods, from which
28      most of the recent PM relative risks were derived. The 1996 PM AQCD concluded:
29           "In summary, most available epidemiologic evidence suggests that increased mortality
30           results from both short-term and long-term ambient PM exposure.  Limitations of available
31           evidence prevent quantification of years of life lost to such mortality in the population. Life

        March 2001                                6-9        DRAFT-DO NOT QUOTE OR CITE

-------
 1           shortening, lag time, and latent period of PM-mediated mortality are almost certainly
 2           distributed over long time periods, although these temporal distributions have not been
 3           characterized." (p. 13-45)
 4           Only a limited number of PM-mortality studies analyzed fine particles and chemically
 5      specific components of PM. The Harvard Six Cities Study (Schwartz et al., 1996a) analyzed
 6      size-fractionated PM (PM2 5, PM10/15, and PM10/15.2 5) and PM chemical components (sulfates and
 7      H+).  The results suggested that PM2 5 was most significantly associated with mortality among the
 8      components of PM. While FT was not significantly associated with mortality in this and an
 9      earlier analysis (Dockery et al., 1992), the smaller sample size for H+ than for other PM
10      components made a direct comparison difficult.  The 1996 PM AQCD also noted that mortality
11      associations with BS or COH reported in earlier studies in Europe and the U.S. during the 1950s
12      to 1970s most likely reflected contributions from fine particles, as those PM indices had low 50%
13      cut-off diameters (~ 4.5/um).  Furthermore, certain respiratory morbidity studies showed
14      associations between hospital admissions/visits with components of PM in the fine particle
15      range.  Thus, the U.S. EPA 1996 PM AQCD concluded that there was adequate evidence to
16      suggest that fine particles play especially important roles in observed PM mortality effects.
17           Overall, then, the status of key issues raised in the 1996  PM AQCD can be summarized as
18      follows: (1) the observed PM effects are unlikely to be seriously biased by inadequate statistical
19      modeling (e.g., control for seasonality); (2) the observed PM effects are unlikely to be
20      significantly confounded by weather; (3) the observed PM effects may be to some extent
21      confounded or modified by co-pollutants, and such extent may vary from season to season;
22      (4) determining the extent of confounding and effect modification by co-pollutants requires
23      knowledge of relative exposure measurement characterization error among pollutants (there was
24      not sufficient information on this);  (5) no clear evidence for any threshold for PM-mortality
25      associations was reported (statistically identifying a threshold  from existing data was also
26      considered difficult, if not impossible); (6) some limited evidence  for harvesting, a few days of
27      life-shortening, was reported for episodic periods (no study was  conducted to investigate
28      harvesting in non-episodic U.S. data); (7) only a relatively limited number of studies suggested a
29      causal role of fine particles in PM-mortality associations, but in  the light of historical data,
30      biological plausibility, and the results from morbidity studies,  a  greater role for fine particles than
31      coarse particles was suggested in the 1996 PM AQCD as being likely.  The AQCD concluded:

        March 2001                               6-10        DRAFT-DO NOT QUOTE OR CITE

-------
  1           "The evidence for PM-related effects from epidemiologic studies is fairly strong, with most
  2           studies showing increases in mortality, hospital admissions, respiratory symptoms, and
  3           pulmonary function decrements associated with several PM indices. These epidemiologic
  4           findings cannot be wholly attributed to inappropriate or incorrect statistical methods,
  5           misspecification of concentration-effect models, biases in study design or implementation,
  6           measurement of errors in health endpoint, pollution exposure, weather, or other variables,
  7           nor confounding of PM effects with effects of other factors. While the results of the
  8           epidemiology studies should be interpreted cautiously, they nonetheless provide ample
  9           reason to be concerned that there are detectable human health effects attributable to PM at
 10           levels below the current NAAQS." (p. 13-92)
 11
 12      6.2.2.2  Introduction to Newly Available Information
 13           Since the 1996  PM AQCD, numerous new studies have examined short-term associations
 14      between PM indices and mortality. The newly available studies on relationships between short-
 15      term PM exposure and daily mortality are summarized in Table 6-1. The table describes the
 16      location,  study period, levels of PM, outcomes, methods, results, and reported risk estimates and
 17      lags. This table does not include review papers and simulation-only studies that did not include
 18      analyses of real data.  In addition to the table, discussion in the text below highlights findings
 19      from several multi-city studies.  Discussion of implications of new study results for types of
 20      issues identified in foregoing text is mainly deferred to Section 6.4.
 21            The summarization of studies in Table 6-1 (and in subsequent tables) is not meant to imply
 22       that all listed studies should be accorded equal weight in the overall interpretive assessment of
 23       evidence  regarding PM-associated health  effects. In general, increasing scientific  weight should
 24       be accorded to those studies (i.e., those not clearly flawed and which have adequate control for
 25       confounding) in proportion to the precision of their estimate of a health effect. Small studies and
 26       studies with an inadequate exposure gradient generally produce less precise estimates than  large
 27       studies with an adequate  exposure gradient.  Therefore, the range of exposures (e.g., as indicated
 28      by the IQR), the size of the study as indexed by the total number of observations (e.g., days) and
29      total number of events (i.e., total deaths), and the inverse variance for the principal effect
30      estimate are all important indices useful in determining the likely precision of health effects
31      estimates  and in according relative scientific weight to the findings of a given study.

        March 2001                                6-11         DRAFT-DO NOT QUOTE OR CITE

-------
e?
o
K)
O
o
                      TABLE 6-1.  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
Reference,
Location, Years,
PM Index, Mean or
Median, IQR in pig/
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, and covanates.
Results and Comments.  Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
CTs

K)
H
O
O
2
O
H
O
C
O
tn
O
!*
O
HH
H
m
          United States

          Samet et al. (2000a,b).
          90 largest U.S. cities.
          1987-1994.
          PM1(1 mean ranged from
          15.3 (Honolulu) to 52 0
          (Riverside).
                           Non-accidental total deaths and cause-specific
                           (cardiac, respiratory, and the other remaining) deaths,
                           stratified in three age groups (<65, 65-75, 75+), were
                           examined for their associations with PMUI, O3, SO2,
                           NO2, and CO (single, two, and three pollutant
                           models) at lags 0, 1, and 2 days. In the first stage of
                           the hierarchical model, RRs for the pollutants for
                           each city were  obtained using GAM Poisson
                           regression models, adjusting for temperature and
                           dewpomt (0-day  and average of 1 -3 days for both
                           variables), day-of-week, seasonal cycles, intercept
                           and seasonal cycles for three age groups. In the
                           second stage, between-city variation in RRs were
                           modeled within region.  The third stage modeled
                           between-region variation (7 regions).  Two alternative
                           assumptions were made regarding the prior
                           distribution: one with possibly substantial
                           heterogeneity and the other with less or no
                           heterogeneity within region. The weighted second-
                           stage regression included five types of county-
                           specific variables: (1) mean weather and pollution
                           variables; (2) mortality rate; (3) socio-demographic
                           variables; (4) urbanization; (5) variables related to
                           measurement error
                                                  The estimated city-specific coefficients were mostly
                                                  positive at lag 0, 1, and 2 days (estimated overall effect
                                                  size was largest at lag 1, with the estimated percent
                                                  excess death rate per 10 ,ug/m3 PMi() being about
                                                  0.5%).  The posterior probabilities that the overall
                                                  effects are greater than 0 at these lags were 0.99, 1.00,
                                                  and 0.98, respectively. None of the county-specific
                                                  variables (effect modifiers) in the second-stage
                                                  regression significantly explained the heterogeneity of
                                                  PMH) effects across cities. In the 3-stage regression
                                                  model with the index for 7 geographical regions, the
                                                  effect of PMH1 varied somewhat across the 7  regions,
                                                  with the effect in the Northeast being the greatest.
                                                  Adding O3 and other gaseous pollutants did not
                                                  markedly change the posterior distributions of PMU,
                                                  effects. O3 effects, as examined by season, were
                                                  associated with mortality in summer (=0.5 percent per
                                                  10 ppb increase), but not in all season data (negative in
                                                  winter).
                                                   Posterior mean estimates and 95%
                                                   credible intervals for total mortality
                                                   excess deaths per 50 Aig/m3 increase
                                                   in PM,0 at lag 1 day 2.3% (0.1,
                                                   4 5) for "more heterogeneity"
                                                   across-city assumption; 2 2% (0.5,
                                                   4.0) for "less or no heterogeneity"
                                                   across cities assumption. The
                                                   largest PM10 effect estimated for
                                                   7 U.S. regions was for the
                                                   Northeast:  4.6% (2.7, 6.5) excess
                                                   deaths per 50 AJg/m3 PMUI
                                                   increment.

-------
s
S3
 O
 o
                 TABLE 6-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
          Reference,
          Location, Years,
          PM Index, Mean or
          Median, IQR in //g/m3.
Study Description. Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, and covariates.
                                                                                        Results and Comments. Design Issues, Uncertainties,
                                                                                        Quantitative Outcomes.
                                                    PM Index, lag, Excess
                                                    Risk% (95% LCL, UCL),
                                                    Co-pollutants.
 Tl
 H
 6
 o
 o
o
 c
 o
 H
 w
o
I—I
          United States (cont'd)

          Dominici et al. (2000).
          20 largest U.S. cities.
          1987-1994. PM10 mean
          ranged from 23.8 ,ug/m3
          (San Antonio) to
          52.0 Aig/m3 (Riverside).
          Braga et al. (2000). Five
          U.S. cities: Pittsburgh, PA;
          Detroit, MI; Chicago, IL;
          Minneapolis-St. Paul, MN;
          Seattle, WA. 1986-1993
          PMH) means were 35, 37,
          37, 28, and 33 ^g/m3,
          respectively in these cities.
                                     Non-accidental total deaths (stratified in three age
                                     groups:  <65, 65-75, 75+) were examined for their
                                     associations with PM,,, and O3 (single, 2, and 3
                                     pollutant models) at lags 0, 1, and 2 days. In the first
                                     stage of the hierarchical model, RRs for PM10 and O3
                                     for each city were obtained using GAM Poisson
                                     regression models, adjusting for temperature and
                                     dewpoint (0-day and average of 1 -3 days for both
                                     variables), day-of-week, seasonal cycles, intercept
                                     and seasonal cycles for three age groups.  In the
                                     second stage, between-city variation in RRs were
                                     modeled as a function of city-specific covariates
                                     including mean PM10 and O3 levels, percent poverty,
                                     and percent of population with age 65 and over. The
                                     prior distribution assumed heterogeneity across cities.
                                     To approximate the posterior distribution, a Markov
                                     Chain Monte Carlo (MCMC) algonthm with a block
                                     Gibbs sampler was implemented. The second stage
                                     also considered spatial model, in which RRs in closer
                                     cities were assumed to be more correlated.
Potential confounding caused by respiratory
epidemics on PM-total mortality associations was
investigated in a subset of the 10 cities evaluated by
Schwartz (2000a,b), as summarized below.  GAM
Poisson models were used to estimate city-specific
PM,0 effects, adjusting for temperature, dewpoint,
barometric pressure, time-trend and day-of-week.
A cubic polynomial was used to for each epidemic
period, and a dummy variable was used to control for
isolated epidemic days  Average of 0 and 1 day lags
were used.
The lag 1 day PMU) concentration was positively
associated with total mortality in most locations (only
2 out of 20 coefficients were negative), though the
estimates ranged from 2.1% to -0.4% per 10 Atg/m3
increase in PMU). PM,,,-mortality associations changed
little with the addition of O3 to the model, or with the
addition of a third pollutant in the model. The pattern
of PM]0 effects with respiratory  and cardiovascular was
similar to that of total mortality.  The PM,0 effect was
smaller (and weaker) with other causes of deaths. The
pooled analysis of 20 cities data confirmed the overall
effect on  total and cardio-respiratory mortality, with lag
I day showing the largest effect  estimates.  The
posterior distributions for PMH)  were generally not
influenced by addition of other pollutants.  In the data
for which the distributed lags could be examined (i.e.,
nearly daily data), the sum of the 7-day distributed lag
coefficients was greater than each of single day
coefficients.  The city-specific covariates did not
predict the heterogeneity across  cities. Regional model
results  suggested that PM,,, effects in the West U.S.
was larger than in the East and South U.S.

When respiratory epidemics were adjusted  for, small
decreases in  the PM]() effect were observed (9% in
Chicago, 11 % in Detroit, 3% in  Minneapolis, 5% in
Pittsburgh, and 15% in Seattle).
                                                                                                      Total mortality excess deaths per 50
                                                                                                      A(g/m3 increase in PM1(): 1.8 (-0.5,
                                                                                                      4.1) for lag 0; 1.9 (-0.4, 4.3) for lag
                                                                                                      1; 1.2 (-1.0, 3.4) for lag 2.

                                                                                                      Cardiovascular disease excess
                                                                                                      deaths per 50 ,ug/m3 PMI():
                                                                                                      3.4(1.0,5.9).
The overall estimated percent
excess deaths per 50 /^g/m3 increase
in PMIO was 4.3% (3.0, 5.6) before
controlling for epidemics and 4.0%
(2.6, 5.3) after. Average of 0 and
1 day lags.

-------
I
tr
O
O
       TABLE 6-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES

Reference,
Location, Years,
PM Index, Mean or
Median, IQR in ,ug/m3.
Study Description:  Outcomes, Mean outcome rate,
and ages  Concentration measures or estimates.
Modeling methods: lags, smoothing, and covariates.
Results and Comments. Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
 O
 O
 z;
 o
 H
O
 O
 H
 tfl
 O
 i*
 O
 h-H
 tfl
          United States (cont'd)

          Schwartz (2000a). Ten
          U.S. cities: New Haven,
          CT; Pittsburgh, PA;
          Detroit, MI; Birmingham,
          AL; Canton, OH; Chicago,
          IL; Mmneapohs-St. Paul,
          MN, Colorado Springs,
          CO; Spokane, WA; and
          Seattle, WA. 1986-1993.
          PMU) means were 29, 35,
          36,37,29,37,28,27,41,
          and 33, respectively in
          these cities
Schwartz (2000b). Ten
U.S. cities: New Haven,
CT; Pittsburgh, PA;
Birmingham, AL; Detroit,
MI; Canton, OH; Chicago,
IL; Minneapolis-St. Paul,
MN; Colorado Springs,
CO; Spokane, WA; and
Seattle, WA. 1986-1993.
PMH| means were 29, 35,
36,37,29,37,28,27,41,
and 33, respectively in
these cities.
Daily total (non-accidental) deaths (20, 19, 63, 60,
10, 133, 32, 6, 9, and 29, respectively in these cities
in the order shown left). Deaths stratified by location
of death (in or outside hospital) were also examined.
For each city, a GAM Poisson model adjusting for
temperature, dewpomt, barometric pressure, day-of-
week, season, and time was fitted.  The data were also
analyzed by season (November through April as
heating season). In the second stage, the PMH)
coefficients were modeled as a function of city-
dependent covanates including copollutant to PM,()
regression coefficient (to test confounding),
unemployment rate, education, poverty level, and
percent non-white  Threshold effects were also
examined. The inverse variance weighted averages of
the ten cities' estimates were used to combine results.

The issue of distributed lag effects was the focus of
this study. Daily total (non-accidental) deaths of
persons 65 years of age and older were analyzed.
For each city, a GAM Poisson model adjusting for
temperature, dewpomt, barometric pressure, day-of-
week, season, and time was fitted.  Effects of
dlstnbuted lag were examined using four models:
(1) 1-day mean at lag 0 day; (2) 2-day mean at lag 0
and 1 day; (3) second-degree dlstnbuted lag model
using lags 0 through 5 days; (4) unconstrained
dlstnbuted lag model using lags 0 through 5 days.
The inverse vanance weighted averages of the ten
cities' estimates were used to combine results.
                                                                             PMH1 was significantly associated with total deaths, and
                                                                             the effect size estimates were the same in summer and
                                                                             winter.  Adjusting for other pollutants did not
                                                                             substantially change PM10 effect size estimates. Also,
                                                                             socioeconomic vanables did not modify the estimates.
                                                                             The effect size estimate for the deaths that occurred
                                                                             outside hospitals was substantially greater than that for
                                                                             inside hospitals.  The effect size estimate was larger for
                                                                             subset with PMUI less than 50 ,ug/m3.
                                                                                       The effect size estimates for the quadratic dlstnbuted
                                                                                       model and unconstrained distributed lag model were
                                                                                       similar. Both distributed lag models resulted in
                                                                                       substantially larger effect size estimates than the single
                                                                                       day lag, and moderately larger effect size estimates
                                                                                       than the two-day average models.
                                                   The total mortality RR estimates
                                                   combined across cities per
                                                   50 ^g/ni3 increase of mean of lag 0-
                                                   and 1-days PMUI: overall 3.4 (2.7,
                                                   4  1); summer 3.4 (2.4, 4.4); winter
                                                   3.3 (2.3, 4.4); in-hospital 2.5 (1.5,
                                                   3.4); out-of-hospital 4.5 (3.4, 5.6);
                                                   days < 50 //g/m3 4.4 (3.1, 5.7); with
                                                   SO22.9(1.2, 4.6); with CO 4 6
                                                   (3.2,6.0); withO33 5 (1 6, 5 3).
                                                   Total mortality percent increase
                                                   estimates combined across cities
per 50
                                                                increase in PMU)- 3.3
                                                   (2.5, 4.1) for 1-day mean at lag 0;
                                                   5.4 (4.4, 6.3) 2-day mean of lag 0
                                                   and 1; 7.3 (5.9, 8.6) for quadratic
                                                   distributed lag; and 6.6 (5.3, 8 0)
                                                   for unconstrained distributed lag

-------
g.
 O
 o
       TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES

Reference,
Location, Years,
PM Index, Mean or
Median, IQR i
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods:  lags, smoothing, and covanates.
Results and Comments.  Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
 o
 Z!
 O
 H
O
 O
 H
 m
 O
 J0
 O
 HH
 H
          United States (cont'd)

          Schwartz and Zanobetti
          (2000). Ten U.S. cities.
          Same as above.
Zanobetti and Schwartz
(2000).  Four U.S. cities:
Chicago, IL; Detroit, MI;
Minneapolis-St. Paul, MN;
Pittsburgh, PA. 1986-1993.
PM10 median = 33, 33,25,
and 31 respectively for
these cities.
The issue of a threshold in PM-mortahty exposure-
response curve was the focus of this study. First, a
simulation was conducted to show that the "meta-
smoothmg" could produce unbiased exposure-
response curves  Three hypothetical curves (linear,
piecewise linear, and logarithmic curves) were used
to generate mortality series in 10 cities, and GAM
Poisson models were used to estimate exposure
response curve. Effects of measurement errors were
also simulated.  In the analysis of actual 10 cities
data, GAM Poisson models were fitted, adjusting for
temperature, dewpoint, and barometric pressure, and
day-of-week. Smooth function of PM,,, with the same
span (0.7) in each of the cities. The predicted values
of the log relative risks were computed for 2 //g/m3
increments between 5.5 jUg/m3 and 69.5 jUg/m3 of
PM10 levels.  Then, the predicted values were
combined across cities using  inverse-variance
weighting.

Separate daily counts of total non-accidental deaths,
stratified by sex, race (black and white), and
education (education > 12yrs or not), were examined
to test hypothesis that people in each of these groups
had higher nsk of PM,,,. GAM Poisson models
adjusting for temperature, dewpoint, barometric
pressure, day-of-week, season, and time were used.
The mean of 0- and 1 -day lag PM,() was used.  The
inverse variance weighted averages of the four cities'
estimates were used to combine results
                                                                             The simulation results indicated that the "meta-
                                                                             smoothing" approach did not bias the underlying
                                                                             relationships for the linear and threshold models, but
                                                                             did result in a slight downward bias for the logarithmic
                                                                             model.  Measurement error (additive or multiplicative)
                                                                             in the simulations did not cause upward bias in the
                                                                             relationship below threshold. The threshold detection
                                                                             in the simulation was not very sensitive to the choice of
                                                                             span in smoothing.  In the analysis of real data from
                                                                             10 cities, the combined curve did not show evidence of
                                                                             a threshold in the PM^-mortality associations.
The differences in the effect size estimates among the
various strata were modest. The results suggest effect
modification with the slope in female deaths one third
larger than in male deaths. Potential interaction of
these strata (e.g., black and female) were not
investigated.
                                                   The combined exposure-response
                                                   curve indicates that an increase of
                                                   50 ptg/m3 is associated with about a
                                                   4% increase in daily deaths. Avg.
                                                   of 0 and 1 day lags.
The total mortality RR estimates
combined across cities per 50
^ig/rn3 increase of mean of lag 0-
and 1-days PM,,,:  white 5.0 (4.0,
6.0); black 3.9 (2.3, 5 4); male 3.S
(2.7, 4.9); female 5.5 (4.3, 6.7);
education <12y 4.7 (3.3, 6.0);
education > 12y 3.6 (1.0, 6.3).

-------
                TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
O
O
Reference,
Location, Years,
PM Index, Mean or
Median, IQR in /ug/m3.
Study Description- Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, and covanates
Results and Comments.  Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
Ox
TI
o
O
z;
o
o
c
o
H
m
O
&
0
H
m
          United States (cont'd)

          Moolgavkar (2000a)
          Cook County, Illinois
          Los Angeles County, CA
          Mancopa County, AZ
          1987-1995
          PMIO, CO, O3, NO,, SO, in
          all three locations.  PM, 5 in
          Los Angeles County.
          Cook Co:  PM,,,
          Median = 47 /ug/m3.
          Mancopa Co:  PMI()
          Median = 41.
          Los Angeles Co PMU,
          Median = 44; PM25
          Median = 22
                           Associations between air pollution and time-series of
                           daily deaths evaluated for three U.S. metropolitan
                           areas with different pollutant mixes and climatic
                           conditions.  Daily total non-accidental deaths and
                           deaths from cardiovascular disease (CVD),
                           cerebrovascular (CrD), and chronic obstructive lung
                           disease and associated conditions (COPD) were
                           analyzed by generalized additive Poisson models in
                           relation  to 24-h readings for each of the air pollutants
                           averaged over all monitors in each county.  All
                           models included an intercept term for day-of-week
                           and a  spline smoother for temporal trends Effects of
                           weather were first evaluated by regressing daily
                           deaths (for each mortality endpomt) against temp and
                           rel  humidity with lag times of 0 to 5 days Then lags
                           that minimized deviance for temp and rel. humidity
                           were kept fixed for subsequent pollutant effect
                           analyses. Each pollutant entered linearly into the
                           regression and lags of between 0 to 5 days examined
                           Effects of two or more pollutants were then evaluated
                           in multipollutant models.  Sensitivity analyses were
                           used to evaluate effect of degree of smoothing on
                           results.
                                                 In general, the gases, especially CO (but not O3) were
                                                 much more strongly associated with mortality than PM.
                                                  Specified pattern of results found for each county
                                                 were as follows.  For Cook Co., in single pollutant
                                                 analyses PM10, CO, and O3 were all associated (PMH)
                                                 most strongly on lag 0-2 days) with total mortality, as
                                                 were SO2 and NO, (strongest association on lag 1  day
                                                 for the latter two) In joint analyses with one of gases,
                                                 the coefficients for both PMU) and the gas were
                                                 somewhat attenuated, but remained stat. sig. for some
                                                 lags. With 3-pollutant models, PMIO coefficient
                                                 became small and non-sig. (except at lag 0), whereas
                                                 the gases dominated. For Los Angeles,  PMU), PM2S,
                                                 CO,  NO2, and SO2, (but not O3), were all associated
                                                 with total mortality.  In joint analyses with CO or  SO2
                                                 and  either PM,,, or PM, s,  PM metrics were markedly
                                                 reduced and non-sig., whereas estimates for gases
                                                 remained robust. In Mancopa Co. single-pollutant
                                                 analyses, PM,,, and each of the gases,  (except Oi), were
                                                 associated with total morality; in 2-pollutant models,
                                                 coefficients for CO,  NO2, SO2, were more robust than
                                                 for PM,,,.  Analogous patterns of more robust gaseous
                                                 pollutant effects were generally  found for cause-
                                                 specific (CVD, CrD, COPD) mortality analyses
                                                 Author concluded that while direct effect of individual
                                                 components of air pollution cannot be ruled out,
                                                 individual components best thought of as indices of
                                                 overall pollutant mix
                                                   In single pollutant models,
                                                   estimated daily total mortality %
                                                   excess deaths per 50 /xg/m3 PM,,,
                                                   was mainly in range of: 0.5-1 .0%
                                                   lags 0-2 Cook Co.; 0.25-1.0% lags
                                                   0-2 LA; 2.0% lag 2 Mancopa.
                                                   Percent per 25 A£g/m3 PM25 0.5%
                                                   lags 0, 1 for Los Angeles.

                                                   Maximum estimated COPD %
                                                   excess deaths (95% CI) per
                                                   50 /^g/m3 PMI(1:
                                                   Cook Co. 5.4 (0.3,10 7),  lag 2; with
                                                   O3, 3.0 (-1.8, 8.1) lag 2,  LA 5 9
                                                   (-1.6, 140) lag 1, Mancopa 8.2
                                                   (-4.2, 22.3) lag 1 ; per 25 ,ug PM, ,
                                                   in LA 2.7 (-3.4,9.1).

                                                   CVD%per50/ug/m3PM,,r
                                                   Cook 2.2 (0 4, 4. 1 ) lag 3; with O,,
                                                   SO2 1.99 (-006, 4 1) lag 3; LA 4  5
                                                   (1.7, 7.4) lag 2; with CO   056
                                                   (-3.8, 2.8) lag 2; Mancopa 8.9
                                                   (2.7, 15.4) lag 1, with NO,  7.4
                                                   (-0.95,  16.3) lag 1.  Percent per
                                                   25 //g/m1 PM2 5, LA 2.6 (0.4, 4 9)
                                                   lagl; with CO 0.60 (-2.1, 3.4).

                                                   CrD % per 50 ^g/m3 PMI(1.
                                                   Cook 3.3 (-0.12, 6.8) lag 2, LA 2  9
                                                   (-2.3, 8 4) lag 3; Mancopa 11  1
                                                   (0 54, 22.8) lag 5. Percent  per
                                                                                                     25 /
                                                                                                     lag 3.
                                                                                                                                        PM25, LA 3.6 (-0 6, 7.9)

-------
tu
3
tr
O
o
                 TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES

          Reference,
          Location, Years,
          PM Index, Mean or
          Median, IQR in ,ug/m3.
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods:  lags, smoothing, and covariates.
                                                                                     Results and Comments.
                                                                                     Quantitative Outcomes.
Design Issues, Uncertainties,
                                                                                                    PM Index, lag, Excess
                                                                                                    Risk% (95% LCL, UCL),
                                                                                                    Co-pollutants.
 \*>
 £
 Tl
 H
 I
 O
 O

 O
 H
O
 c
 o
 H
 tfl
 O
 O
 HH
 H
 W
         United States (cont'd)

         Ostroetal. (1999a).
         Coachella Valley, CA.
         1989-1992. PM10
         (beta-attenuation)
         Mean = 56.8 Aig/m3.
          Ostro et al. (2000).
          Coachella Valley, CA.
          1989-1998  PM25=16.8;
          PM,,,.25 = 25.8 in Indio;
          PM25 = 12.7;PM,,,25 =
          17.9 in Palm Springs.
Study evaluated total, respiratory, cardiovascular,
non-cardiorespiratory and age >50 yr deaths (mean =
5.4,0.6, 1.8, 3.0, and 4.8 per day, respectively). The
valley is a desert area where 50-60% of PM,,,
estimated to be coarse particles. Correlation between
gravimetric and beta-attenuation, separated by
25 miles, was high (r = 0.93). Beta-attenuation data
were used for analysis. GAM Poisson models
adjusting for temperature, humidity, day-of-week,
season, and time were used. Seasonally stratified
analyses were also conducted. Lags 0-3 days
(separately) of PMI() along with moving averages of
3 and 5 days examined, as were O3, NO2, and  CO.

A follow-up study of the Coachella Valley data, with
PM2 5 and PM10_2 5 data in the last 2.5 years. Both
PM2 5 and PM10_2 5 were estimated for the remaining
years to increase power of analyses.
                                                                                      Associations were found between 2- or 3-day lagged     Total mortality percent excess
                                                                                      PM1() and all mortality categories examined, except
                                                                                      non-cardiorespiratory series.  The effect size estimates
                                                                                      for total and cardiovascular deaths were larger for
                                                                                      warm season (May through October) than for all year
                                                                                      period. NO2 and CO were significant predictor of
                                                                                      mortality in single pollutant models, but in
                                                                                      multi-pollutant models, none of the gaseous pollutants
                                                                                      were significant (coefficients reduced), whereas PMHI
                                                                                      coefficients remained the same and significant.
                                                                                     Several pollutants were associated with all-cause
                                                                                     mortality, including PM2 5, CO, and NO2. More
                                                                                     consistent results were found for cardiovascular
                                                                                     mortality, for which significant associations were found
                                                                                                    deaths per 50 fj.g/ m3
                                                                                                    lag=4.6(0.6, 8.8).

                                                                                                    Cardiac deaths:
                                                                                                    8.33(2.14, 14.9)

                                                                                                    Respiratory deaths:
                                                                                                    13.9(3.25,25.6)
                                                PM1() at 2-day
                                                                                                    Total percent excess deaths.
                                                                                                    PM,,, = 2.0 (-1.0, 5.1) per 50 ,ug/m3
                                                                                                    PM25= 11.5(0.2, 24.1) per
                                                                                                    25 ueJm3
                                                                                     forPM10.25 andPM,,,, but not PM2 5 (possibly due to     PM10_25 = 1.3 (-0.6, 3.5) per
                                                                                     low range of PM25 concentrations and reduced sample
                                                                                     size for PM2 5 data)
                                                                                                    25, _
                                                                                                    Cardio deaths:
                                                                                                    PM,,, = 6.1 (2.0, 10.3) per 50 Aig/m3
                                                                                                    PM2S = 8.6(-6.4, 25.8) per
                                                                                                    25 Aig/m3
                                                                                                    PM,,,25 = 2.6(0.7, 4.5) per
                                                                                                    25 Mg/rn3
                                                                                                    Respiratory deaths:
                                                                                                    PMlo=-2.0(-11.4,8.4)per
                                                                                                                                        PM2S= 13.3 (-43.1,32.1) per
                                                                                                                                        25 Aig/m3
                                                                                                                                        PMlo.25 = -1.3(-6.2,4.0)per
                                                                                                                                        25 Aig/m3

-------
                 TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
 O
 o
Reference,
Location, Years,
PM Index, Mean or
Median, IQR in ,ug/m3.
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods' lags, smoothing, and covanates.
Results and Comments. Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
 ON

 oo
 H
 6
 o
 2
 o
o
 G
 O
 H
 W
 O
 !*
 O
 H—I
 H
          United States (cont'd)

          Fairley(1999).
          Santa Clara County, CA
          1989-1996.PM25 (13);
          PMIO(34);PM1M5(11);
          COH (0.5 unit); NO3 (3.0);
          SO4(1.8)
Schwartz et al. (1999).
Spokane, WA
1989-1995
PMU,: "control" days:
42 Aig/m'; dust-storm
days:  263

Popeetal. (1999a).
Ogden, Salt Lake City, and
Provo/Orem, UT
1985-1995
PM|0 (32 for Ogden; 41 for
SLC; 38 for P/O)
                           Total, cardiovascular, and respiratory deaths were
                           regressed on PMIO, PM25, PM1().25, COH, nitrate,
                           sulfate, O3, CO, NO2, adjusting for trend, season, and
                           mm and max temperature, using Poisson GAM
                           model. Season-specific analysis was also conducted.
                           The same approach was also used to re-analyze
                           1980-1986 data (previously analyzed by Fairley,
                           1990).
Effects of high concentration of coarse crustal
particles were investigated by comparing death counts
on 17 dust storm episodes to those on non-episode
days on the same day of the  years in other years,
adjusting for temperature, dewpoint, and day-of-
week, using Poisson regression.

Associations between PMH1 and total, cardiovascular,
and respiratory deaths studied in three urban areas in
Utah's Wasatch Front, using Poisson GAM model
and adjusting for seasonably, temperature, humidity,
and barometric pressure Analysis was conducted
with or without dust (crustal coarse particles) storm
episodes, as identified on  the high "clearing index"
days, an index of air stagnation.
                                                  PM2 5 and nitrate were most significantly associated
                                                  with mortality, but all the pollutants (except PMH,.25)
                                                  were significantly associated in single poll, models.
                                                  In 2 and 4 poll, models with PM25 or nitrate, other
                                                  pollutants were not significant. The RRs for
                                                  respiratory deaths were always larger than those for
                                                  total or cardiovascular deaths.  The difference in risk
                                                  between season was not significant for PM2 5. The
                                                  1980-1986 results were similar, except that COH was
                                                  very significantly associated with mortality.
                                                                                      No association was found between the mortality and
                                                                                      dust storm days on the same day or the following day.
Salt Lake City (SLC), where past studies reported little
PM|0-mortality associations, had substantially more
dust storm episodes. When the dust storm days were
screened out from analysis and PM10 data from
multiple monitors were used, comparable RRs were
estimated for SLC and Provo/Orem (P/O).
                                                   Total mortality per 25 ,ug/m3 PM, 5
                                                   at 0 d lag: 8% in one pollutant
                                                   model; 9-12% in 2 pollutant model;
                                                   12% in 4-pollutant model.  Also,
                                                   8% per 50 //g/m3 PM10 in one
                                                   pollutant model and 2% per
                                                   Cardiovascular mortality:
                                                   PM10 = 9% per 50 f^g/m3
                                                   PM25= 13%per25//g/m3
                                                   PM|0.25 = 3% per 25 /ug/m3

                                                   Respiratory mortality:
                                                   PM]0=ll%per50^g/m3
                                                   PM25 = 7%per25^g/m3
                                                   PM,(K!5=16%per25^g/m3

                                                   0% (-4.5, 4.7) for dust storm days
                                                   at 0 day lag (50 ^g/m3 PM,0)
                                                   (lagged days also reported to have
                                                   no associations).
Ogden PMI(,
Total (Od)= 12.0% (4.5, 20.1)
CVD(0-4d)=1.4%(-8.3, 12.2)
Resp. (0-4 d) = 23.8 (2.8, 49.1)
SLC PM,0
Total (0 d) = 2.3% (0.47)
CVD (0-4 d) = 6.5% (2.2, 11.0)
Resp. (0-4 d) = 8.2 (2.4, 15.2)
Provo/Ovem PMI(1
Total (Od)= 1.9% (-2.1,6.0)
CVD (0-4 d) = 8.6% (2.4, 15.2)
Resp. (0-4 d) = 2.2% (-9.8, 15.9)
Note: Above % for PM3 5 and
PM,,,.25 all per 25 ,ug/m3; all PM,,,
% per 50 ^g/m3.

-------
 03
 g.
 O
 O
       TABLE  6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES

Reference,
Location, Years,
PM Index, Mean or
Median, IQR in
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods:  lags, smoothing, and covanates.
Results and Comments. Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
 ON
 T)
 H
 6
 O
 2
 O
 H
O
 c
 O
 w
 O
 &
 O
 h-H
 H
 W
          United States (cont'd)

          Schwartz and Zanobetti
          (2000).
          Chicago
          1988-1993.
          PMH). Median = 36 ,ug/m3.
          Lippmann et al. (2000).
          Detroit, MI.1992-1994.
          PM10 = 31;
          PM25 = 18;
          PMIM5=13.
For 1985-1990 period
TSP, PMI(1, TSP-PM,,,,
Sulfate from TSP
(TSP-SO4-)
Total (non-accidental), m-hospital, out-of-hospital
deaths (median =• 132, 79, and 53 per day,
respectively), as well as heart disease, COPD, and
pneumonia elderly hospital admissions (115, 7, and
25 per day, respectively) were analyzed to investigate
possible "harvesting" effect of PMH). GAM Poisson
models adjusting for temperature, relative humidity,
day-of-week, and season were applied in baseline
models using the average of the same day and
previous day's PMH). The seasonal and trend
decomposition techniques called STL was applied to
the health outcome and exposure data to decompose
them into different time-scales (i.e, short-term to
long-term), excluding the long, seasonal cycles (120
day window).  The associations were examined with
smoothing windows of 15, 30, 45, and 60 days

For 1992-1994 study period, total (non-accidental),
cardiovascular, respiratory, and other deaths were
analyzed using GAM Poisson  models, adjusting for
season, temperature, and relative humidity.  The air
pollution variables analyzed were: PM10, PM25,
PM|0.25) sulfate, H+, 03, SO2, NO2, and CO.

For earlier 1985-1990 study period, total non-
accidental, circulatory, respiratory, and "other"
(non-circulatory or respiratory non-accidental)
mortality were evaluated versus noted PM indices and
gaseous pollutants.
                                                                             The effect size estimate for deaths outside of the
                                                                             hospital is larger than for deaths inside the hospital.
                                                                             All cause mortality shows an increase in effect size at
                                                                             longer time scales. The effect size for deaths outside of
                                                                             hospital increases more steeply with increasing time
                                                                             scale than the effect size for deaths inside of hospitals.
PM|0, PM25, and PM|()_25 were more significantly
associated with mortality outcomes than sulfate or H*.
PM coefficients were generally not sensitive to
inclusion of gaseous pollutants. PMU), PM25, and
PM|o.25 effect size estimates were comparable per
same distributional increment (5th to 95"1 percentile).

Both PM,,, (lag 1 and 2 day) and  TSP (lag 1  day) but
not TSP-PM,,, or TSP-SO4" significantly associated
with respiratory mortality for 1985-1990 period. The
simultaneous inclusions of gaseous pollutants with
PM10 or TSP reduced PM effect size by 0 to 34%.
Effect size estimates for total, circulatory, and "other"
categories were smaller than for respiratory mortality.
                                                    Mortality RR estimates per
                                                    50 Mg/m3 increase of mean of lag 0-
                                                    and 1-days PM,(I: total deaths 4.5
                                                    (3.1, 6.0); in-hospital 3.9 (2.1, 5.8);
                                                    out-of-hospital 6.3 (4.1, 8.6).
                                                    For total deaths, the RR
                                                    approximately doubles  as the time
                                                    scale changes from 15 days to 60
                                                    days. For out-of-hospital deaths, it
                                                    triples fromlS days to 60 days time
                                                    scale.
Total mortality percent excess
deaths:
PM,,,(1 d) = 4.4%(-1.0, 10.1)
PM25(Od) = 3.1%(-0.6, 7.0)
PM,0.2S(1 d) = 4.0% (-1.2, 9.4)
PM10(1 d) = 6.9%( 1.3, 15.7)
PM25(1 d) = 3.2% (-2.3, 8.9)
PM,,,.25(1 d) = 7.8%(0.0, 16.2)
Respiratory mortality:
PM,,, (0 d) = 7.8% (-10.2, 29.5)
Circulatory mortality:
PM25(od) = 2.3%(-10.3, 16,6)
PM,,,2S(2d) = 7.4%(-9.1,26.9)
Note: All above PM,,, per
50 /^g/rn3; all PM2
per 25 jUg/m3.

-------
g.

K>
O
O
       TABLE 6-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
Reference,
Location, Years,
PM Index, Mean or
Median, IQR in Afg/m3.
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, and covariates.
Results and Comments. Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
0\
to
O
Tl
H
6
o
z
o
H
O
G
O
w
o
x»
n
^H
H
tn
          United States (cont'd)

          Chock etal (2000).
          1989-1991
          Pittsburgh, PA
          PM10 (daily)
          PM2 5 (every 2 days)
Klemm and Mason (2000).
Atlanta, GA
1998-1999
PM25mean=19.9;
PM25/PMIO=065.
Nitrate, EC, OC, and
oxygenated HC

Gwynn et al. (2000).
Buffalo, NY. 1988-1990.
PM10 (24);
COH (0.2/1000ft);
SO4= (62 nmoles/m3)
Schwartz (2000c).
Boston, MA. 1979-1986.
                           Study evaluated associations between daily mortality
                           and several air pollution variables (PM,,,, PM25, CO,
                           O3, NO2, SO2) in two age groups (<75 yr., >75 yr.) in
                           Pittsburgh, PA, during 3-yr. period. Poisson
                           regression used, including filtering of data based on
                           cubic B-spline basis functions, with adjustments for
                           seasonal trends, day-of-week effects, temp., dew
                           point.  Single- and multi-pollutant models run for 0,
                           1, 2, and 3 day lags. PM25/PMI() = 0.67.
Reported "interim" results for 1 yr period of
observations regarding total mortality in Atlanta, GA
during 1998-1999. Generalized additive model used
to assess effects of PM, 5 vs PM l(,.2 5, and for nitrate,
EC, OC and oxygenated HC components.
Total, circulatory, and respiratory mortality and
unscheduled hospital admissions were analyzed for
their associations with H+, S04=, PMI(), COH, O3,
CO, SO2, and NO,, adjusting for seasonal cycles, day-
of-week, temperature, humidity, using. Poisson and
negative binomial GAM models

Non-accidental total, pneumonia, COPD, and
ischemic heart disease mortality were examined for
possible "harvesting" effects of PM. The mortality,
air pollution, and weather time-series were separated
into seasonal cycles (longer than 2-month period),
midscale, and short-term fluctuations using STL
algorithm. Four different midscale components were
used (15, 30, 45,  and 60 days) to examine the extent
of harvesting. GAM Poisson regression analysis was
performed using deaths, pollution, and weather for
each of the four midscale periods.
Issues of seasonal dependence of correlation among
pollutants, multi-collinearity among pollutants, and
instability of coefficients emphasized.  Single- and
multi-pollutant non-seasonal models show significant
positive association between PM10 and daily mortality,
but seasonal models showed much multi-collinearity,
masking association of any pollutant with mortality.
Also, based on data set half the size for PM10, the PM25
coefficients were highly unstable and, since no
consistently significant associations found in this small
data  set stratified by age group and season, no
conclusions drawn on relative role of PM2S vs. PM,,).^.

No significant associations were found for any of the
pollutants examined, possibly due to a relatively short
study penod (1 -year). The coefficient and t-ratio were
larger for PM2 5 than for PMHU5.
For total mortality, all the PM components were
significantly associated, with H+ being the most
significant, and COH the least significant predictors.
The gaseous pollutants were mostly weakly associated
with total mortality


For COPD deaths, the results suggest that most of the
mortality was displaced by only a few months.
For pneumonia, ischemic heart disease, and total
mortality, the effect size increased with longer time
scales.
                                                                                                     Total mortality percent increase per
                                                                                                     25 jUg/m3 for aged <75 yrs:
                                                                                                     PM2 5 = 2.6% (2.0, 7.3)
                                                                                                     PM1,,.25 = 0.7%(-1.7, 3.7)

                                                                                                     Total mortality percent increase per
                                                                                                     25 ,ug/m3 for aged >75 yrs:
                                                                                                     PM25=1.5%(-3.0, 6.3)
                                                                                                     PM1,,.25=1.3%(-1.3, 3.8)
Total mortality percent increase per
25 Aig/m3 for:
PM25 = 4.8%(-32, 13.4)
PM1,,.25=1.4%(-11.3, 159)
                                                                                                                                           12% (2.6, 22.7) per 50 ,ug/m3 PMI(I
                                                                                                                                           at 2-day lag.
Total mortality percent increase per
25 |Ug/m3 increase in PM, 5
5.3(1.8, 9.0) for short-term
fluctuations; 9.6(8.1, 11.1) for the
60 day window.

-------
                TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
o
o
Reference,
Location, Years,
PM Index, Mean or
Median, IQR in
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, and covariates.
Results and Comments. Design Issues, Uncertainties,
Quantitative Outcomes
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
H
6
o
z
o
H
o
HH
H
W
          United States (cont'd)

          Lipfert et al. (2000a).
          Philadelphia (7 county
          Metropolitan area),
          1992-1995. Harvard PM
          measurements: PM25
          (17.3); PM.o (24.1);
          PM10.25(6.8),  sulfate
          (53.1 nmol/m3);H+
          (8.0 nmol/m3)
Laden et. al. (2000)
Six Cities (means):
Watertown, MA (16.5);
Kingston-Harnman, TN
(21.1); St. Louis, MO
(19.2);Steubenville, OH
(30.5); Portage, WI (11.3);
Topeka, KS (12.2).
1979-1988?. 15 trace
elements in the dichot
PM25:Si, S, Cl, K, Ca, V,
Mn, Al, Ni, Zn, Se, Br, Pb,
Cu, and Fe.
Levy (1998).
King County, WA.
1990-1994.
PMHI Nephelometer (30);
(0.59 bsp unit)
12 mortality variables, as categorized by area, age,
and cause, were regressed on 29 pollution vanables
(PM components, O3, SO2, NO2> CO, and by sub-
areas), yielding 348 regression results.  Both
dependent and explanatory variables were pre-filtered
using the 19-day-weighted average filter prior to OLS
regression. Covanates were selected from filtered
temperature (several lagged and averaged values),
indicator variables for hot and cold days and day-of-
week using stepwise procedure. The average of
current and previous days' pollution levels were used.

Total (non-accidental), ischemic heart disease,
pneumonia, and COPD (mean daily total deaths for
the six cities:  59, 12, 55, 3, 11, and 3, respectively in
the order shown left) A factor analysis was
conducted on the 15 elements in the fine fraction of
dichot samplers to obtain five common factors;
factors were rotated to maximize the projection of the
single "tracer" element (as in  part identified from the
past studies conducted on these data) for each factor;
PM2, was regressed on the identified factors scores so
that the factor scores could be expressed in the mass
scale. Using GAM Poisson models adjusting for
temperature, humidity, day-of-week, season, and
time, mortality was regressed on the factor scores in
the mass scale. The mean of the same-day and
previous day (increasing the sample size from 6,211
to 9,108 days) mass  values were used.  The city-
specific regression coefficients were combined using
inverse variance weights.

Out-of-hospital deaths (total,  respiratory, COPD,
ischemic heart disease, heart failure, sudden cardiac
death screening codes, and stroke) were related to
PM10, nephelometer (0.2 - 1.0 ^m fine particles), SO2,
and CO, adjusting for day-of-week, month of the
year, temperature and dewpoint, using Poisson
regression.
                                                                              Significant associations were found for a wide variety
                                                                              gaseous and particulate pollutants, especially for peak
                                                                              O3. No systematic differences were seen according
                                                                              particle size or chemistry. Mortality for one part of the
                                                                              metropolitan area could be associated with air quality
                                                                              from another, not necessarily neighboring part.
Three sources of fine particles were defined in all six
cities with a representative element for each source
type: Si for soil and crustal material; Pb for motor
vehicle exhaust; and Se for coal combustion sources.
In city-specific analysis, additional sources (V for fuel
oil combustion, Cl for salt, etc.) were considered. Five
source factors were considered for each city, except
Topeka with the three sources. Coal and mobile
sources account for the  majority of fine particles in
each city.  In all of the metropolitan areas combined,
46% of the total fine particle mass was attributed to
coal combustion and 19% to mobile sources. The
strongest increase in daily mortality was associated
with the mobile source factor. The coal combustion
factor was positively associated with mortality in all
metropolitan areas, with the exception of Topeka. The
crustal factor from the fine particles was not associated
with mortality.
                                                                                        Nephelometer data were not associated with mortality.   Total mortality percent excess:
                                                    The fractional Philadelphia
                                                    mortality risk attributed to the
                                                    pollutant levels: "average risk" was
                                                    0.0423 for25,ug/m3 PM2S; 0.0517
                                                    for 25 //g/m3 PM,0.25; 0.0609 for
                                                    50 Mg/m3 PMIO, using the Harvard
                                                    PM indices at avg. of 0 and 1 d
                                                    lags.
                                                                                                                                           Total mortality percent excess
                                                                                                                                           overall: 4.0 (2.8, 5.3), 2.7 (0.6,
                                                                                                                                           5.0) with each 25 ptg/m3 increase in
                                                                                                                                           the two-day mean of coal
                                                                                                                                           combustion fine PM factor; 8.7
                                                                                                                                           (4.2, 13.4) with each 25 ^g/m3
                                                                                                                                           increase in  the two-day mean of
                                                                                                                                           mobile source fine PM factor; -5.7
                                                                                                                                           (-13.7, 3.2) with each 25 Mg/m3
                                                                                                                                           increase in  the two-day mean of the
                                                                                                                                           crustal source fine PM factor.
 Cause-specific death analyses suggest PM associations
 with ischemic heart disease deaths.  Associations of
 mortality with SO2 and CO not mentioned  Mean daily
 death counts were small (e.g., 7.7 for total,  1.6 for
 ischemic heart disease). This is an apparently
 preliminary analysis.
 5.6% (- 2.4, 1.43) per 50 ^g/
 PM10 at avg. of 2 to 4 d lag; 7.2%
 (-6.3, 22.8) with SO2 CO. 1.8%
 (-3.5,7.3)per25/ug/m3PMl; -1.0
 (-8.7,. 7.7) with SO2 and CO.

-------
                TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
O
O
Reference,
Location, Years,
PM Index, Mean or
Median, IQR i
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods, lags, smoothing, and covariates.
Results and Comments.  Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
to
to
Tl
H
s
o
c
3
W
O
5*>
O
H
tn
          United States (cont'd)

          Maretal. (2000).
          Phoenix, AZ. 1995-1997.
          PM10, andPM25, and
          PM,(I.25(TEOM), with
          means = 46.5, 13.0, and
          33.5, respectively; and
          PM25(DFPSS), mean =
          12.0.
Clyde et al. (2000).
Phoenix, AZ. 1995-1998.
PMlo,andPM25, (from
TEOM), with means =
454, and 13.8. PM,,,.,,
computed as PM|0-PM25.
Total (non-accidental) and cardiovascular deaths
(mean = 8.6 and 3.9, respectively) for only those who
resided in the zip codes located near the air pollution
monitor were included. GAM Poisson models were
used, adjusting for season, temperature, and relative
humidity. Air pollution variables evaluated included:
O3, SO2, NO2, CO, TEOM PM)0, TEOM PM25,
TEOM PM10.24, DFPSS PM25, S, Zn, Pb, soil, soil-
corrected K (KS), nonsoil PM, OC, EC, and TC. Lags
0 to 4 days evaluated.  Factor analysis also conducted
on chemical components of DFPSS PM25 (Al, Si,  S,
Ca, Fe, Zn, Mn, Pb, Br, KS, OC, and EC); and factor
scores included in mortality regression.

Elderly (age  >65 years) non-accidental mortality for
three regions of increasing size in Phoenix urban area
analyzed to evaluate influence of spatial uniformity of
PMHI and PM25 All-age accidental deaths for the
metropolitan area also examined as a "control".
GAM Poisson models adjusting for season
(smoothing splines of days), temperature, specific
humidity, and lags 0- to 3-d of weather variables  PM
indices for lags 0-3 d considered. Bayesian Model
Averaging (BMA) produces posterior mean relative
risks by weighting each model (out of all possible
model specifications examined) based on support
received from the data
                                                                            Total mortality was significantly associated with
                                                                            CO and NO2 and weakly associated with SO2, PM10,
                                                                            PMU).2 5, and EC. Cardiovascular mortality was
                                                                            significantly associated with CO, NO2, SO2, PM2 5,
                                                                             PM,,,, PM,,
            , OC and EC. Combustion-related
                                                                             factors and secondary aerosol factors were also
                                                                             associated with cardiovascular mortality. Soil-related
                                                                             factors, as well as individual variables that are
                                                                             associated with soil were negatively associated with
                                                                             total mortality.
The BMA results suggest that a weak association was
found only for the mortality variable defined over the
region with uniform PM25, with a 0.91 probability that
RR is greater than 1.  The other elderly mortality
variables, including the accidental deaths ("control"),
had such probabilities in the range between 0.46 to
0 77 Within the results for the mortality defined over
the region with uniform PM, 5, the results suggested
that effect was primarily due to coarse particles rather
than fine; only the lag 1 coarse PM was consistently
included in  the highly ranked models.
Total mortality percent excess:
5.4 (0.1, 11.1) for PM10 (TEOM)
50 Mg/m3 at lag 0 d; 3.0 (-0.5, 6.6)
for PM(1,W5) (TEOM) 25 /ug/m3 at
lag Od; 3.0 (-0.7, 6.9) for PM25
(TEOM) 25 Aig/m3 at lag 0 d.
Cardiovascular mortality RRs: 9.9
(1.9, 18.4) for PM10 (TEOM)
50 yug/m3 at lag 0 d; 18.7(5.7,33.2)
for PM2 5 (TEOM) 25 ,ug/m3 at lag
1 d; and 6.4 (1.4, 11.7)PM10
(TEOM) 25 //g/m3 PM(I(1.25) at lag 0
d.

Posterior mean RRs and 90%
probability intervals per changes of
25 ^g/m3 in all lags of fine and
coarse PM for elderly mortality for
uniform PMU, region  1.06(1+,
1.11).

-------
E

H
^C

^f^
H
U
W
fe
fe
W
^4
H

J
«<
H
OS
0
S
W
j
C cd O
O C 0
a B -a
3 M C
O 1> cd
g 0 M
i> w .5
S 2 «
3 O
«J « O
g g g
I S "
O r- W
^ ° «
0 | -
c S -3
2 8jf
•& o t5
1° £
l«{.f
-^"^
3 "O O
-^ G *5
C/3 TO P^


^1=
- c :s-
2 s c
« s 5
o" . x" — .
U C 4) »
c o -a c
JJ 43 C a












































TS
•**
0

8
«

I



                 •-
                                              SP c
                                              ~i -" •—-
                                                                          n.
                                                ce       .*-a-* o




                                                                  .







                                                       c

                                                                    — J "O 01
                    oCic-O-   0>   — O — !^  •—    --f  -T-«c_a— — J "









                 3>OoS^=Hgffl3c.!i>cg  rtx.gaSggWo'^ooos
                               "
                 o
                 o
                                                    '„" tu c
March 2001
6-23
DRAFT-DO NOT QUOTE OR CITE

-------
                 TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
O
O
Reference,
Location, Years,
PM Index, Mean or
Median, IQR in /^g/
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods' lags, smoothing, and covanates.
Results and Comments.  Design Issues, Uncertainties,
Quantitative Outcomes.
                                                   PM Index, lag, Excess
                                                   Risk% (95% LCL, UCL),
                                                   Co-pollutants.
ON
NJ
•n
H
6
O
2
O
O
c
O
w
O
*)
n
H
          United States (cont'd)

          Gamble (1998).  Dallas,
          TX. 1990-1994.
          PMI(1 (25)
          Ostro(1995).
          San Bernardino and
          Riverside Counties, CA,
          1980-1986.
          PM, 5 (estimated from
          visual range).
          Mean = 32.5.
Kelsalletal. (1997).
Philadelphia, PA
1974-1988.
TSP (67)
Moolgavkar and Leubeck
(1996). Philadelphia, PA
1973-1988. TSP (68)
Relationships of total, respiratory, cardiovascular,
cancer, and remaining non-accidental deaths to PM10,
O3, NO2, SO2, and CO evaluated, adjusting for
temperature,  dewpoint, day-of-week, and seasonal
cycles (trigonometric terms) using Poisson regression.

Study evaluated total, respiratory, cardiovascular, and
age > = 65 deaths (mean = 40.7, 3.8, 18.7, and 36.4
per day, respectively).  PM2 5 estimated based on
airport visual range and previously published
empirical formula. Autoregressive OLS (for total)
and Poisson (for sub-categones) regressions used,
adjusting for season (sme/cosme with cycles from 1
yr to 0.75 mo; prefiltermg with 15-day moving ave.;
dichotomous variables  for each year and month;
smooth function of day and temp.), day-of-week,
temp, and dewpoint.  Evaluated lags 0, 1, and 2 of
estimated PM25> as well as moving averages of 2, 3,
and 4 days and O3

Total, cardiovascular, respiratory, and by-age
mortality regressed on TSP, SO,, NO,, O3, and CO,
adjusting for temporal trends and weather, using
Poisson GAM model
A critical review paper, with an analysis of total daily
mortality for its association with TSP, SO,, NO2, and
O,, adjusting for temporal trends, temperature, and
also conducting analysis by season, using Poisson
GAM model.
                                                                             O3 (avg  of 1-2 day lags), NO2 (avg.. 4 -5 day lags), and
                                                                             CO (avg. of lags 5- 6 days) were significantly
                                                                             positively associated with total mortality.  PM10 and
                                                                             SO2 were not significantly associated with any deaths.
                                                                             The results were dependent on season. No PM25 -
                                                                             mortality association found for the full year-round
                                                                             period. Associations between estimated PM2 5 (same-
                                                                             day) and total and respiratory deaths found during
                                                                             summer quarters (April - Sept.).  Correlation between
                                                                             the estimated PM2 5 and daily max temp, was low (r =
                                                                             0.08) dunng the summer quarters. Ozone was also
                                                                             associated with mortality, but was also relatively highly
                                                                             correlated with temp, r = 0.73).  Moving averages of
                                                                             PM2 5 did not improve the associations.
                                                   -3.6% (-12.7, 6.6) per
                                                   PM|(, at 0 lag (other lags also
                                                   reported to have no associations)
                                                   Percent excess deaths per 25 ,ug/m3
                                                   of estimated PM2 5, lag 0: Full
                                                   year:  0.3 (-0.6, 1.2) for total; 2.1
                                                   (-0.3, 4.5) for respiratory; and 0.7
                                                   (-0.3, 1.7) for circulatory. Summer
                                                   quarters:  1.6 (0.03, 3.2) for total;
                                                   5.5 (1.1, 10.0) for respiratory; and 0
                                                   (-1.0, 1.0) for circulatory.
TSP, SO2, O3, and 1 -day lagged CO individually         Total mortality excess risk: 3.2%
showed statistically significant associations with total    (0, 6.1) per 100 /^g/m3 TSP at 0 day
mortality No NO2 associations unless SO2 or TSP was   lag.
also considered. The effects of TSP and SO2 were
diminished when both pollutants were included.

RR results presented as figures, and seasonal difference   Total mortality excess nsk  ranged
                           ,
noted.  TSP, SO2, O3 - mortality associations varied
across season. TSP associations were stronger in
summer and fall. NO2 was the most significant
predictor
                                                   = 0 (winter) to =4% (summer) per
                                                   1 00 A'g/m3 TSP at 1 day lag.

-------
 o
 Of
 to
 o
 o
       TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES

Reference,
Location, Years,
PM Index, Mean or
Median, 1QR in ,ug/m3.
Study Description. Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods:  lags, smoothing, and covanates.
Results and Comments.  Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
 o
 Z!
 o
 H
O
 c
 o
 H
o
          United States (cont'd)

          Neaset. al. (1999).
          Philadelphia. 1973-1980.
          TSP mean = 77.2.
          Schwartz (2000d).
          Philadelphia. 1974-1988.
          TSP.  Mean = 70 /j.g/m} for
          warm season (Apnl
          through August) and
                   for cold season.
Canada

Burnett et al. (1998a).
11 Canadian cities.
1980-1991. No PM index
data available on consistent
daily basis.
                           Total, age over 65, cancer, and cardiovascular deaths
                           analyzed for association with TSP. Conditional
                           logistic regression analysis with case-crossover
                           design conducted.  Average values of current and
                           previous days' TSP used. Case period is the 48-hr
                           period ending at midnight on day of death. Control
                           periods are 7, 14, and 21 days before and after the
                           case period.  Other covariates included temperature
                           on the previous day, dewpoint on the same day, an
                           indicator for hot days (> 80°F), an indicator for
                           humid days (dewpoint > 66°F), and interaction of
                           same-day temp, and winter season

                           Total  (non-accidental) deaths analyzed. GAM
                           Poisson models adjusting for temperature, dewpoint,
                           day-of-week, and season applied to each of 15 warm
                           and cold seasons. Humidity-corrected extinction
                           coefficient, derived from airport visual range, also
                           considered as explanatory variable. In the second
                           stage, resulting 30 coefficients were regressed on
                           regression coefficients of TSP on SO2. Results of
                           first stage analysis combined using inverse variance
                           weighting.
Total non-accidental deaths were linked to gaseous
air pollutants (N02, O3, SO2, and CO) using GAM
Poisson models adjusting for seasonal cycles, day-of-
week, and weather (selected from spline-smoothed
functions of temperature, dewpoint, relative humidity
with 0, 1, and 2 day lags using forward  stepwise
procedure). Pollution variables evaluated at 0,  1, 2,
and up to 3-day lag averages thereof.  No PM index
included in analyses because daily PM measurements
not available.  City-specific models containing all
four gaseous pollutants examined.  Overall risks
computed by averaging risks across cities.
                                                  In each set of the six control periods, TSP was
                                                  associated with total mortality. A model with four
                                                  symmetric reference periods 7 and 14 days around the
                                                  case period produced a similar result. A model with
                                                  only two symmetric reference periods of 7 days around
                                                  the case produced a larger estimate.  A larger effect was
                                                  seen for deaths in persons > 65 years of age and for
                                                  deaths due to pneumonia and to cardiovascular disease.
                                                  Cancer mortality was not associated with TSP.
                                                  When TSP controlled for, no significant association
                                                  between SO2 and daily deaths. SO2 had no association
                                                  with daily mortality when it was poorly correlated with
                                                  TSP. In contrast, when SO2 was controlled for, TSP
                                                  was more strongly associated with mortality than when
                                                  it was less correlated with SO2. However, all of the
                                                  association between TSP and mortality was explained
                                                  by its correlation with extinction coefficient.
NO2 had 4.1% increased risk per mean concentration;
O3 had 1.8%; SO2 had 1.4%, and CO had 0.9% in
multiple pollutant regression models.  A 0.4%
reduction in excess mortality was attributed to
achieving a sulfur content of gasoline of 30 ppm in five
Canadian cities. Daily PM data for fine and coarse
mass and sulfates available on varying (not daily)
schedules allowed ecologic comparison of gaseous
pollutant risks by mean fine particle indicators mass
concentrations.
                                                   Odds Ratio (OR) for all cause
                                                   mortality per 100 Mg/m3 increase in
                                                   48-hr mean TSP was 1.056 (1.027,
                                                   1.086). The corresponding number
                                                   for those  aged 65 and over was
                                                   1.074(1.037, 1.111), and 1.063
                                                   (1.021, 1.107) for cardiovascular
                                                   disease.
                                                   Total mortality excess risk
                                                   estimates combined across
                                                   seasons/years:  9.0 (5.7, 12.5) per
                                                   100,ug/m3TSP.
Found suggestion of weak negative
confounding of NO2 and SO2
effects with fine particles and weak
positive confounding of particle
effects with O3. No quantitative
RR or ER estimates reported for
PM indicators.

-------
                TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
O
o
Reference,
Location, Years,
PM Index, Mean or
Median, IQR in jug/
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, and covanates.
Results and Comments.  Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
K)
ON
T)
H
6
o
z
o
H
O
c
o
H
W
O
*>
o
HH
H
          Canada (cont'd)

          Burnett et al. (2000).
          8 largest Canadian cities.
          1986-1996. All city mean
          PM10 25.9; PM,5 13.3;
          Firths 12.6;sulfate2.6.
Burnett et al. (1998b).
Toronto, 1980-1994.
TSP (60); COM (0.42),
SO4= (9.2 ,ug/m3);
PM,,, (30, estimated);
PM25 (18, estimated)
Total non-accidental deaths linked to PM indices
(PM,0, PM25, PM|0.25, sulfate, 47 elemental
component concentrations for fine and coarse
fractions) and gaseous air pollutants (NO2,  O3, SO2,
and CO). Each city's mortality, pollution, and
weather variables separately filtered for seasonal
trends and day-of-week patterns. The residual series
from all the cities then analyzed in a GAM  Poisson
model. The weather model was selected from spline-
smoothed functions of temperature, relative humidity,
and maximum change in barometric pressure within a
day, with 0 and 1 day lags using forward stepwise
procedure. Pollution effects were examined at lags 0
through 5 days.  To avoid unstable parameter
estimates in multi-pollutant models, principal
components were also used as predictors in the
regression models.

Total, cardiac, and other nonaccidental deaths (and by
age groups) were regressed on TSP, COH,  SO4=, CO,
NO2, SO2, O3, estimated PM,,, and PM25 (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 dewpomt using
Poisson GAM model.
                                                                             O3 was weakly correlated with other pollutants and
                                                                             other pollutants were "moderately" correlated with
                                                                             each other (the highest was r = 0.65 for NO2 and CO).
                                                                             The strongest association with mortality for all
                                                                             pollutants considered were for 0 or 1 day lags. PM2S
                                                                             was a stronger predictor of mortality than PMH).2 5. The
                                                                             estimated gaseous pollutant effects were generally
                                                                             reduced by inclusion of PM25 or PM10, but not PM,o.25.
                                                                             Sulfate, Fe, Ni, and Zn were most strongly associated
                                                                             with mortality. Total effect of these four components
                                                                             was greater than that for PM2 5 mass alone.
Essentially all pollutants were significant predictors of
total deaths in single pollutant models, but in two
pollutant models with CO, most pollutants' estimated
RRs reduced (all PM indices remained significant).
Based on results from the co-pollutant models and
various stepwise regressions, authors noted that effects
of the complex mixture of air pollutants could be
almost completely explained by the levels of CO and
TSP.
                                                   Percentage increase in daily filtered
                                                   non-accidental deaths associated
                                                   with increases of 50 A^g/m3 PMH)
                                                   and 25
             PM2 5 or PM1
                                                                               at
                                                   lagl  day: 3.5 (1.0, 6.0) for PM,,,;
                                                   3.0 (1.1, 5.0) for PM25; and  1.8
                                                   (-0.7, 4.4) for PMi,,.25. In the
                                                   multiple pollutant model with
                                                   PM25, PM|0.25, and the 4 gaseous
                                                   pollutants, 1.9 (0.6, 3.2) for PM25;
                                                   and 1.2 (-1.3, 3.8) for PM(10.25I.
Total mortality percent excess:
2.3% (0.8, 3.8) per 100 ^g/m3 TSP,
3.5% (1 ..8, 5.3) per 50 //g/m3 PM,,,,
4.8% (3.3, 6 4) per 25 ,ug/m3 PM, 5.
0 day lag for TSP and PM,,,, Avg
of 0 and 1 day forPM,5.

-------
                TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
O
o
Reference,
Location, Years,
PM Index, Mean or
Median, IQR in /ig/m3.
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods:  lags, smoothing, and covariates.
Results and Comments.  Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
ON
b
o
o
H
O
o
H
tfl
O
i»
n
HH
H
W
          Canada (cont'd)

          Goldberg et al. (2000)
          Montreal, Quebec
          1984-95 Mean
          TSP = 53.1
          (65-120.5)^g/m3
          PM25 = 3.3 (0.0 -30.0)
          Ozkaynak et al. (1996).
          Toronto,  1970-1991.
          TSP (80); COH (0.42
          /1000ft).
                           Study aimed to shed light on population subgroups
                           that my be susceptible to PM effects. Linked data on
                           daily deaths with other health data (physician visits,
                           pharmaceutical Rx, etc.) to identify individuals with
                           presenting health conditions. PMHI and PM2 5
                           measured by dichotomous sampler 1 in 6 days until
                           1992 (2 stations), then daily through 1993.  PM
                           missing days interpolated from COH, ext. coefficient,
                           sulfates. Used quasi likelihood estimation in GAM's
                           to assess PM associations with total and cause-
                           specific mortality; and, also, in subgroups by age
                           and/or preexisting health conditions. Adjusted for
                           CO, NO2, NO, O3 and SO2 in 2-pollutant and all-
                           pollutant models.
                           Total, cardiovascular, COPD, pneumonia, respiratory,
                           cancer, and the remaining mortality senes were
                           related to TSP, SO2, COH, NO2, O3, and CO,
                           adjusting for seasonal cycles (by high-pass filtering
                           each senes) temperature, humidity, day-of-week,
                           using OLS regression. Factor analysis of multiple
                           pollutants was also conducted to extract automobile
                           related pollution, and mortality senes were regressed
                           on the resulting automobile factor scores.
                                                  Significant associations found for all-cause (total non-
                                                  accidental) and cause-specific (cancer, CAD,
                                                  respiratory disease, diabetes) with PM measures.
                                                  Results reported for PM25, COH and sulfates. All three
                                                  PM measures associated with increases in total, resp.,
                                                  and "other nonaccidental", and diabetes-related
                                                  mortality.  No PM associations found with digestive,
                                                  accidental, renal or neurologic causes of death. Also,
                                                  mainly in 65+ yr group, found consistent associations
                                                  with increased total mortality among persons who had
                                                  cancer, acute lower resp. diseases, any cardiovascular
                                                  disease, chronic CAD and congestive heart failure
                                                  (CHF).
                                                  TSP (0 day lag) was significantly associated with total
                                                  and cardiovascular deaths. NO2 (0-day lag) was a
                                                  significant predictor for respiratory and COPD deaths.
                                                  2-day lagged O3 was associated with total, respiratory,
                                                  and pneumonia deaths. Factor analysis showed factor
                                                  with high loadings for NO2) COH, and CO (apparently
                                                  representing automobile factor) as significant predictor
                                                  for total, cancer, cardiovascular, respiratory, and
                                                  pneumonia deaths.
                                                   Percent excess mortality per
                                                   25 Aig/m3 estimated PM2 5:
                                                   Total deaths (3 d ave.) = 4.4% (2.5,
                                                   6.3)
                                                   CV deaths (3 d ave.) = 2.6% (-01,
                                                   5.5)
                                                   Resp deaths (3 d ave.) = 16.0%
                                                   (9.7, 22 8)
                                                   Coronary artery (3 d ave.) = 3.4%
                                                   (-0.2,  7.1)
                                                   Diabetes (3 d ave.) = 15.7% (4.8,
                                                   27.9)
                                                   Lower Resp Disease (3  d ave.) =
                                                   9.7% (4.5, 15.1)
                                                   Airways disease (3 d ave.) = 2.7%
                                                   (-0.9,6.4)
                                                   CHF (3 d ave.) = 8.2%  (3.3, 13.4)

                                                   Total mortality excess risk: 2.8%
                                                   per  100 ,ug/m3 TSP at 0 day lag.

-------
o
N>
o
o
       TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES

Reference,
Location, Years,
PM Index, Mean or
Median, IQR in yug/m3.
Study Description:  Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, and covariates.
Results and Comments. Design Issues, Uncertainties,
Quantitative Outcomes.
                                                                                                                                           PM Index, lag, Excess
                                                                                                                                           Risk% (95% LCL, UCL),
                                                                                                                                           Co-pollutants.
K>
00
 \-/
 ?
 ^n
 H
 6
 o
 z
 3
0
 o
 H
 W
 O
 o
 >—*
 H
 tn
          Europe

          Katsouyanni et al. (1997).
          12 European (APHEA)
          cities. 1975-1992 (study
          years different from city to
          city). Median Black
          Smoke (BS) levels ranged
          from 13 in London to 73 in
          Athens and Kracow.
          Touloumietal. (1997).
          6 European (APHEA)
          cities  1977-1992 (study
          years different from city to
          city). Median Black Smoke
          (BS) levels ranged from
          14^6 in London to 84.4 in
          Athens.

          Zmirouetal. (1998).
          10 European (APHEA)
          cities. 1977-1992 (study
          years different from city to
          city). Median Black Smoke
          (BS) levels ranged from
          13 in London to 73 in
          Kracow.
                           Total daily deaths regressed on BS or SO2 using
                           Poisson models, adjusting for seasonal cycles, day-of-
                           week, influenza epidemic, holidays, temp., humidity
                           Final analysis done with autoregressive Poisson
                           models to allow for overdispersion and
                           autocorrelation. Pollution effects examined at
                           0 through 3 day lags and multi-day averages thereof.
                           When city-specific coefficients tested to be
                           homogeneous, overall estimates obtained by
                           computing variance-weighted means of city-specific
                           estimates (fixed effects model).  When significant
                           heterogeneity present, source of heterogeneity sought
                           by examining a predefined list of city-specific
                           variables, including annual and seasonal means of
                           pollution and weather variables, number of
                           monitoring sites, correlation between measurements
                           from different sites, age-standardized mortality,
                           proportion of elderly people, smoking prevalence,
                           and geographic difference (north-south, east-west).
                           A random effects model was fit when heterogeneity
                           could not be explained.

                           The results of the short-term effects of ambient NO2
                           and/or O, on daily deaths from all causes (excluding
                           accidents) were discussed to provide a basis of
                           comparison with estimated S02 or BS effects in the
                           APHEA cities.  Poisson models,  lag/averaging of
                           pollution, and the computation of combined effects
                           across the cities were done in the same way as done
                           by Katsouyanni et al. (1997), as described above.

                           Cardiovascular, respiratory, and digestive mortality
                           series mlO European cities analyzed to examine
                           cause-specificity of air pollution  The mortality series
                           were analyzed for associations with PM (BS,  except
                           TSP in Milan and Bratislava; PM,3 in Lyon),  NO,,
                           O,, and SO2.  Poisson models, lag/averaging of
                           pollution, and computation of combined effects
                           across the cities done in the same way as by
                           Katsouyanni et al. (1997), above.
                                                  Substantial variation in pollution levels (winter mean
                                                  SO2 ranged from 30 to 330 //g/m3), climate, and
                                                  seasonal patterns were observed across cities.
                                                  Significant heterogeneity was found for the effects of
                                                  BS and SO2, but only the separation between western
                                                  and central eastern European cities resulted in more
                                                  homogeneous subgroups.  Significant heterogeneity for
                                                  SO2 remained in western cities. Cumulative effects of
                                                  prolonged (two to four days) exposure to air pollutants
                                                  resulted in estimates comparable with the one day
                                                  effects. The effects of both SO2 and BS were stronger
                                                  during the summer and were independent.
                                                  Significant positive associations found between daily
                                                  deaths and both NO2 and O3.  Tendency for larger
                                                  effects of NO, in cities with higher levels of BS. When
                                                  BS included in the model, pooled estimate for O3 effect
                                                  only slightly reduced, but coefficient for NO, reduced
                                                  by half.  Authors speculated that short-term effects of
                                                  NO2 on mortality confounded by other vehicle-derived
                                                  pollutants

                                                  The cardiovascular and respiratory mortality series
                                                  were associated with BS and SO2 in western European
                                                  cities, but not in the five central European cities  NO2
                                                  did not show consistent mortality associations.  RRs for
                                                  respiratory causes were at least equal to, or greater than
                                                  those for cardiovascular causes  No pollutant exhibited
                                                  any association with digestive mortality.
                                                    Total mortality excess deaths per
                                                    25 pig/m3 increase in single day BS
                                                    for western European cities:  1.4
                                                    (1.0, 1.8), and 2 (1, 3) per 50 ^g/m3
                                                    PMKi increase. In central/eastern
                                                    Europe cities, corresponding figure
                                                    was 0.3 (0.05, 0.5) per 25 //g/m3
                                                    BS.
                                                    NO2 and/or O3 estimates only.
                                                   Pooled cardiovascular mortality
                                                   percent excess deaths per 25 ,ag/rn'
                                                   increase in BS for western
                                                   European cities' 1.0 (0 3, 1 7); for
                                                   respiratory mortality, it was 2.0
                                                   (0.8, 3.2) in single lag day models
                                                   (the lags apparently varied across
                                                   cities).

-------
                TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
O
o
Reference,
Location, Years,
PM Index, Mean or
Median, IQR in ,ug/m3.
Study Description Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, and covariates.
Results and Comments.  Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
K>
H
6
O
Z
O
H
o
H
tfl
o
J«
o
h- H
H
W
          Europe (cont'd)

          Bremneretal. (1999).
          London, UK, 1992-1994.
          BS(13),PM1(,(29).
          Prescott et al. (1998).
          Edinburgh, UK, 1981-
          1995.  PM,0(21,byTEOM
          only for 1992-1995); BS
          (8.7).
          Rooneyetal. (1998).
          England and Wales, and
          Greater London, UK
          PMUI (56, dunng the worst
          heat wave; 39, July-August
          mean)
Wordley et al. (1997).
Birmingham, UK,
1992-1994.
PM10 (apparently beta-
attenuation, 26)
                           Total, cardiovascular, and respiratory (by age)
                           mortality series were regressed on PM10, BS, O3, NO2,
                           CO, and SO2, adjusting for seasonal cycles, day-of-
                           week, influenza, holidays, temperature, humidity, and
                           autocorrelation using Poisson model.
Both mortality (total, cardiovascular, and respiratory)
and emergency hospital admissions (cardiovascular
and respiratory), in two age groups (<65 and >= 65),
were analyzed for their associations with PM1(I, BS,
SO2, N02, 03, and CO, using Poisson regression
adjusting for seasonal cycles, day-of-week,
temperature, and wind speed.

Excess deaths, by age, sex , and cause, during the
1995 heat wave were estimated by taking the
difference between the deaths dunng heat wave and
the 31-day moving averages (for 1995 and 1993-94
separately). The pollution effects, additively for O3>
PM10, and NO2, were estimated based on the
published season-specific coefficients from the 1987-
1992 study (Anderson et al., 1996).

Mortality data were analyzed for COPD, pneumonia,
all respiratory diseases, all circulatory diseases, and
all causes.  Mortality associations with PM1(), 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.
All effect size estimates (except O3) were positive for
total deaths (though not significant for single lag
models).  The effects of O3 found in 1987-1992 were
not replicated, except in cardiovascular deaths
Multiple day averaging (e.g., 0-1, 0-2 days) tend to
give more significant effect size estimates. The effect
size for PM10 and BS were similar for the same
distributional  increment.

Among all the pollutants, BS was most significantly
associated with all cause, cardiovascular, and
respiratory mortality series. In the subset in which
PMU| data were available, the RR estimates for BS and
PM,0 for all cause elderly mortality were comparable.
Other pollutants' mortality associations were generally
inconsistent.

Air pollution levels at all the locations rose during the
heat wave.  8.9% and 16.1% excess deaths were
estimated for  England and Wales, and Greater  London,
respectively.  Of these excess deaths, up to 62% and
38%, respectively for these locations, may be
attributable to combined pollution effects.
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 PM,,, to below 70 /^g/m3 was estimated to be
"small" (0.2% for total deaths), but the PM10 level
above 70 ^g/m3 occurred only once during the study
period.
                                                                                                     1.9% (0.0, 3.8) per 25 pig/m3 BS at
                                                                                                     lagl day; 1.3% (-1.0, 3.6) per
                                                                                                     50 /^g/m3 PM,0 at lag 1 d for total
                                                                                                     deaths.  Resp. deaths (3 d) = 4.9%
                                                                                                     (0.5, 9.4). CVD deaths (1 d) =
                                                                                                     3.0% (0.3, 5.7).
                                                                                                                                3.8 (1.3, 6.4) per 25 A^g/m3 increase
                                                                                                                                in BS for all cause mortality in age
                                                                                                                                65+ group, avg. of 1 -3 day lags.
                                                                                                                                 2.6% increase for PM10 in Greater
                                                                                                                                 London during heat wave.
 5.6% (0.5, 11.0) per 50 ^g/m3 PM1()
 at 1 d lag for total deaths. COPD (1
 d lag) deaths = 27.6 (5.1,54.9).
 Circulatory (1  d) deaths = 8.8 (1.9,
 17.1)

-------
3,
O
O
       TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES

Reference,
Location, Years,
PM Index, Mean or
Median, IQR in jUg/m3.
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, and covariates
Results and Comments.  Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
 O
o
z;
o
H
O
c
o
H
m
O
&
n
i— t
H
m
          Europe (cont'd)

          Hoek et al. (2000). The
          Netherlands, 1986-1994.
          PM,0 (median 34); BS
          (median  10).
          P6nkaetal.(1998).
          Helsinki, Finland,
          1987-1993.
          TSP (median 64); PM1()
          (median 28)
Peters et al. (1999a)
A highly polluted coal
basin area in the Czech
Republic and a rural area in
Germany, northeast
Bavana districts. 1982-
1994. TSP: mean = 121.1
and 51.6, respectively, for
these two regions. PM]0
and PM2 5 were also
measured in the coal basin
during  1993-1994 (mean =
65.9 and 51.0,
respectively).
                           Total, cardiovascular, COPD, and pneumonia
                           mortality series were regressed on PM10, BS, sulfate,
                           nitrate, O3, SO2, CO, adjusting for seasonal cycles,
                           day-of-week, influenza, temperature, and humidity
                           using Poisson GAM model. Deaths occurring inside
                           and outside hospitals were also examined.

                           Total and cardiovascular deaths, for age groups < 65
                           and 65 +, were related to PM10, TSP, SO2, NO2, and
                           O3, using Poisson model adjusting for temperature,
                           relative humidity, day-of-week, temporal patterns,
                           holiday and influenza epidemics.
Non-accidental total and cardiovascular deaths (mean
= 18.2 and 12.0 per day, for the Czech and Bavaria
areas, respectively). The APHEA approach (Poisson
model with sine/cosine, temperature as a quadratic
function, relative humidity, influenza, day-of-week
as covariates), as well as GAM Poisson models were
considered. Logarithm of TSP, SO2, NO2, O3, and
CO (and PM,0and PM25 for 1993-1994) were
examined at lags 0 through 3 days.
                                                 Particulate air pollution was not more consistently
                                                 associated with mortality than were the gaseous
                                                 pollutants SO2 and NO2.  Sulfate, nitrate, and BS were
                                                 more consistently associated with total mortality than
                                                 was PMHI. The RRs for all pollutants were larger in the
                                                 summer months than in the winter months.

                                                 No pollutant significantly associated with mortality
                                                 from all cardiovascular or CVD causes in 65+ year age
                                                   0.9(0.1, 1.7)per50,ug/m3PMl();
                                                   1.0 (0.5, 1.5) per 25 ^g/m3 BS; 3.2
                                                   (0.6, 5.9) per 25 ptg/m3 sulfate; and
                                                   4.1 (1.4, 6.9) per 25 ^g/m3 nitrate,
                                                   all at 1 day lag.
                                                   18.8% (5.6, 33.2) per 50 Mg/rn3
                                                   PM|,, 4 day lag (other lags negative
                                                 group. Only in age <65 year group, PM10 associated     or zero).
                                                 with total and CVD deaths with 4 and 5 d lags,
                                                 respectively. The "significant" lags were rather
                                                 "spiky". O3 was also associated with CVD mortality
                                                 <65 yr. group with inconsistent signs and late and
                                                 spiky lags (neg. on d 5 and pos. on d 6).
In the coal basin (i.e , the Czech Republic polluted
area), on the average, 68% of the TSP was PMHI, and
most of PMi,, was PM25 (75%).  For the coal basin,
associations were found between the logarithm of TSP
and all-cause mortality at lag 1 or 2 days.  SO2 was also
associated with all-cause mortality with slightly lower
significance. PMH) and PM25 were both associated
with all-cause mortality in 1993-1994 with a lag of 1-
day  NO2, O3 and CO were positively but more weakly
associated with mortality than PM indices or SO2.
In the Bavarian region, neither TSP nor SO2 was
associated with mortality, but CO (at lag 1-day) and O3
(at lag 0-day) were associated with all-cause mortality.
Total mortality excess deaths per
100 jUg/m3 increase in TSP for the
Czech region: 3.8 (0 8, 6.9) at lag
2-day for 1982-1994 period.  For
period 1993-1994, 9.5 (1.2, 18.5)
per 100 Atg/m3 increase in TSP at
lag 1-day, and 4.8 (0.7, 9.0) per
50 pig/m3 increase in PM10, and 1.4
(- 0.5, 3.4) per 25 ,ug/m3 PM2 5

-------
                 TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
 O
 O
Reference,
Location, Years,
PM Index, Mean or
Median, IQR in ^g/
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods'  lags, smoothing, and covariates.
                                                                                       Results and Comments. Design Issues, Uncertainties,
                                                                                       Quantitative Outcomes.
                                                   PM Index, lag, Excess
                                                   Risk% (95% LCL, UCL),
                                                   Co-pollutants
 T1
 H
 6
 o
 o
 H
/O
 O
 H
o
HH
H
          Europe (cont'd)

          Hoeketal. (1997).
          Rotterdam, the
          Netherlands, 1983-1991.
          TSP (median 42); BS
          (median 13).
          Kotesovec et al. (2000).
          Northern Bohemia, Czech
          Republic, 1982-1994. TSP
          (121.3).
          Zanobetti et al. (2000a).
          Milan, Italy.  1980-1989.
          TSP mean =  142.
Anderson et al. (1996).
London, UK, 1987-1992.
BS(15)
                           Total mortality (also by age group) was regressed on
                           TSP, Fe (from TSP filter), BS, O3, SO2, CO, adjusting
                           for seasonal cycles, day-of-week, influenza,
                           temperature, and humidity using Poisson GAM
                           model.
Total (excluding accidents and children younger than
1 yr), cause specific (cardiovascular and cancer), age
(65 and less vs. otherwise), and gender specific
mortality series were examined  for their associations
with TSP and S02 using logistic model, adjusting for
seasonal cycles, influenza epidemics, linear and
quadratic temperature terms.  Lags 0 through 6 days,
as well as a 7 day mean values were examined.

The focus of this study was to quantify mortality
displacement using GAM distributed lag models.
Non-accidental total deaths were regressed on smooth
function of TSP distributed over the same day and the
previous 45 days using penalized splines for the
smooth terms and seasonal cycles, temperature,
humidity, day-of-week, holidays, and influenza
epidemics. The mortality displacement was modeled
as the initial positive increase, negative rebound (due
to depletion), followed by another positive
coefficients period, and the sum of the three phases
were considered as the total cumulative effect.

Total, cardiovascular, and respiratory mortality series
were regressed on BS, O3, NO2, and SO2, adjusting
for seasonal cycles, day-of-week, influenza, holidays,
temperature, humidity, and autocorrelation using
Poisson model.
Daily deaths were most consistently associated with
TSP TSP and O3 effects were "independent" of SO2
and CO. Total iron (from TSP filter) was associated
"less consistently" with mortality than TSP was. The
estimated RRs for PM indices were higher in warm
season than in cold season.

For the total mortality, TSP, but not SO2, was
associated.  There were apparent differences in
associations were found between men and women.
For example, for age below 65 cardiovascular mortality
was associated with TSP for men but not for women
                                                                            TSP was positively associated with mortality up to
                                                                            13 days, followed by nearly zero coefficients between
                                                                            14 and 20 days, and then followed by smaller but
                                                                            positive coefficients up to the 45lh day (maximum
                                                                            examined). The sum of these coefficients was over
                                                                            three times larger than that for the single-day estimate.
Both O3 (0 day lag) and BS (1 day lag) were significant
predictors of total deaths. O3 was also positively
significantly associated with respiratory and
cardiovascular deaths.  The effect size estimates per
the same distributional increment (10% to 90%) were
larger for O3 than for BS. These effects were larger in
warm season.  SO2 and NO, were not consistently
associated with mortality.
                                                                                                     5.5 (1.1, 9.9) per 100 Mg/m1 TSP at
                                                                                                     1 day lag.
                                                                                                                               Total mortality percent excess
                                                                                                                               deaths per 100 ^g/m3 increase in
                                                                                                                               TSP at 2 day lag was 3.4 (0.5, 6.4).
                                                   Total mortality percent increase
                                                   estimates per IQR increase in TSP.
                                                   2.2 (1.4, 3.1) for single-day model;
                                                   6.7 (3.8, 9.6) for distributed lag
                                                   model
2.8% (1 .4, 4.3) per 25 ^g/m3
1 -d lag for total deaths.
CVD(1  d) = 1.0(-l. 1,3.1).
Resp. (1  d)=l.l (-27,50).
BS at

-------
1
o
o
       TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES

Reference,
Location, Years,
PM Index, Mean or
Median, IQR in
Study Description. Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, and covanates.
                                                                            Results and Comments.  Design Issues, Uncertainties,
                                                                            Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
ON
UJ
K)
 z
 3
O
 G
 O
 H
          Europe (cont'd)

          Michelozzi et al. (1998).
          Rome, Italy,
          1992-1995.  TSP("PM]3"
          beta attenuation, 84).
          Garcia-Aymerich et al.
          (2000). Barcelona, Spain.
          1985-1989. Black Smoke
          no data distribution was
          reported)
          Rahlenbeck and Kahl
          (1996). East Berlin,
          1981-1989. "SP"(beta
          attenuation, 97)
Rossi etal. (1999). Milan,
Italy, 1980-1989
TSP ("PM13" beta
attenuation, 142)
                           Total mortality was related to PM13, SO2, NO2, CO,
                           and O3, using Poisson GAM model , adjusting for
                           seasonal cycles, temperature, humidity, day-of-week,
                           and holiday. Analysis of mortality by place of
                           residence, by season, age, place of death (in or out
                           of hospital), and cause was also conducted

                           Daily total (mean = 1.8/day), respiratory, and
                           cardiovascular mortality counts of a cohort (9,987
                           people) with COPD or asthma were associated with
                           black smoke (24-hr),  SO2 (24-hr and 1-hr max), NO2
                           (24-hr and 1-hr max), O3 (1-hr max), temperature,
                           and relative humidity. Poisson regression models
                           using APHEA protocol were used. The resulting RRs
                           were compared with those of the general population.

                           Total mortality (as well as deviations from long-wave
                           cycles) was regressed on SP and SO2, adjusting for
                           day-of-week, month,  year, temperature, and relative
                           humidity, using OLS, with options to log-transform
                           pollution, and w/ and w/o days with pollution above
                           Specific causes of death (respiratory, respiratory
                           infections, COPD, circulatory, cardiac, heart failure,
                           and myocardial infarction) were related to TSP, SO2,
                           and NO2, adjusting for seasonal cycles, temperature,
                           and humidity, using Poisson GAM model.
                                                 PM|3 and NO2 were most consistently associated with
                                                 mortality. CO and O3 coefficients were positive, SO2
                                                 coefficients negative  RR estimates higher in the
                                                 warmer season.  RRs similar for in- and out-of hospital
                                                 deaths.
                                                 Daily mortality in COPD patients was associated with
                                                 all six pollution indices.  This association was stronger
                                                 than in the general population only for daily 1 -hr max
                                                 of SO2, daily 1-hr max and daily means of NO2. BS
                                                 and daily means of SO2 showed similar or weaker
                                                 associations for COPD patients than for the general
                                                 population.
                                                                            Both SP and SO2 were significantly associated with
                                                                            total mortality with 2 day lag in single pollutant model.
                                                                            When both pollutants were included, their coefficients
                                                                            were reduced by 33% and 46% for SP and SO2,
                                                                            respectively.
                                                 All three pollutants were associated with all cause
                                                 mortality.  Cause-specific analysis was conducted for
                                                 TSP only.  Respiratory infection and heart failure
                                                 deaths were both associated with TSP on the
                                                 concurrent day, whereas the associations for
                                                 myocardial infarction and COPD deaths were found for
                                                 the average of 3 to 4 day prior TSP.
                                                                                                                               1.9% (0.5, 3.4)per50/ug/m-1 PMI3
                                                                                                                               at 0 day lag.
                                                                                                                               Total mortality percent increase per
                                                                                                                               25 ,ug/m3 increase in avg. of 0-3
                                                                                                                               day lags of BS:  2.76 (1.31, 4.23) in
                                                                                                                               general population, and 1.14 (  4.4,
                                                                                                                               6.98) in the COPD cohort.
                                                                                                    6.1% per 100 /^g/m1 "SP" at 2 day
                                                                                                    lag.
                                                                                                                                         3.3% (2.4, 4.3) per lOO^g/m'TSP
                                                                                                                                         at 0 day lag.
 O
 HH
 H
 m

-------
                 TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS  STUDIES
O
o
Reference,
Location, Years,
PM Index, Mean or
Median, IQR in //g/m3,
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, and covanates.
Results and Comments. Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
H
6
o
z
o
H
/O
d
o
H
 O
 h-4
 H
 W
          Europe (cont'd)

          Sunyer et al. (2000).
          Barcelona, Spain.
          1990-1995.
          BS means: 43.9 for case
          period, and 43.1 for control
          period.
Tobias and Campbell
(1999).
Barcelona, Spain.
1991-1995.
Black Smoke (BS)
(no data distribution was
reported).


Alberdi Odriozola et al.
(1998). Madrid, Spam,
1986-1992.  "TSP"(beta
attenuation, 47 for average
of 2 stations)

Diaz etal. (1999).
Madrid, Spain 1990-1992.
TSP (no data distribution
was reported).
Those who were over age 35 and sought emergency
room services for COPD exacerbation between 1985
and 1989 and had died during 1990-1995 were
included in analysis. Total, respiratory, and
cardiovascular deaths were analyzed using a
conditional logistic regression analysis with a case-
crossover design, adjusting for temperature, relative
humidity, and influenza epidemics. Bi-directional
control period at 7 days was used.  Average of the
same and previous 2 days used for pollution exposure
period.  Data also stratified by potential effect
modifiers (e.g., age, gender, severity of ER visits,
number of ER visits, etc )

Study examined the sensitivity of estimated total
mortality effects of BS to different approaches to
modeling influenza epidemics: (1) with a single
dummy variable, (2) with three dummy variables;
(3) using daily number of cases of influenza. Poisson
regression used to model total daily mortality,
adjusting for weather, long-term trend, and season,
apparently following APHEA protocol.

Total, respiratory, and cardiovascular deaths were
related to TSP and SO,  Multivariate autoregressive
integrated moving average models used to adjust for
season, temperature, relative humidity, and influenza
epidemics.

Non-accidental, respiratory, and cardiovascular
deaths (mean = 62.4, 6.3, and 23.8 per day,
respectively). Auto-regressive Integrated Moving
Average (AfUMA) models fit to both depend, and
independ. variables first to remove auto-correlation
and seasonably (i.e., pre-whitening"), followed by
examining cross-correlation to find optimal lags.
Multivariate OLS  models thus included ARIMA
components, seasonal cycles (sine/cosine), V-shaped
temp., and optimal lags found for pollution and
weather variables.  TSP, SO2, NO,, and O3 examined.
Season-specific analyses also conducted.
                                                                              BS levels were associated with all cause deaths. The
                                                                              association was stronger for respiratory causes. Older
                                                                              women, patients admitted to intensive care units, and
                                                                              patients with a higher rate of ER visits were at greater
                                                                              risk of deaths associated with BS.
                                                                                        Using the reported daily number of influenza cases
                                                                                        resulted in a better fit (i.e., a lower AIC) than those
                                                                                        using dummy variables.  In the "better" model, the
                                                                                        black smoke coefficient was about 10% smaller than
                                                                                        those in the models with dummy influenza variables,
                                                                                        but remained significant. Lags not reported.
TSP (1-day lag) and SO2 (3-day lagged) were
independently associated with mortality.
TSP was significantly associated with non-accidental
mortality at lag 0 for year around and winter, but with
a 1 -day lag in summer.  A similar pattern was seen for
circulatory deaths  For respiratory mortality, a
significant association with TSP was found only in
summer (0-day lag).  SO2, NOx, and NO, showed
similar associations with non-accidental deaths at lag 0
day. O3' associations with non-accidental mortality
was U-shaped, with inconsistent lags (1, 4, and 10).
                                                    Percent increase per 25 ,ug/m3
                                                    increase in 3-day average BS:  14.2
                                                    (1.6, 28.4) for all causes; 9.7
                                                    (- 10.2, 34.1) for cardiovascular
                                                    deaths; 23.2 (3.0, 47.4) for
                                                    respiratory deaths.
                                                    Total mortality excess deaths per 25
                                                    /ug/m3 increase in BS:  1.37 (0.20,
                                                    2.56) for model using the daily case
                                                    of influenza;  1.71 (0.53, 2.91) for
                                                    model with three influenza dummy
                                                    variables.
4.8% (1 .8, 7.7) per
at lag 1 day.
                                                                                                                                                                       TSP
For non-accidental mortality,
excess deaths was 7.4%
(confidence bands not reported; p <
0.05) per 100 ,ug/m3 TSP at 0 day
lag.

-------
                 TABLE  6-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
O
o
Reference,
Location, Years,
PM Index, Mean or
Median, IQR i
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, and covariates.
Results and Comments. Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
O
H
6
o
O
c
o
w
o
*l
o
H
W
Europe (cont'd)

Wichmann et al., (2000).
Erfurt, Germany.
1995-1998.
Number counts (NC) &
mass concentrations (MC)
of ultrafine particles in
three size classes, 0.01 to
0.1 /urn, and fine particles
in three size classes from
0.1 to 2.5 nm diameter,
using Spectrometryll
Mobile Aerosol
Spectrometry (MAS).
MAS MC PM2.5-0.01
(mean 25.8, median  18.8,
IQR 19.9). Filter
measurements of PMH>
(mean 38.2, median  31.0,
IQR 27.7) and PM25 (mean
26.3, median 20.2, IQR
185). MAS NC2 5-0.01
(mean 17,966 per cu.cm,
median 14,769,  IQR
13,269).
Latin America

Cifuentes et al. (2000).
Santiago, Chile.
1988-1996.
PM25(64.0), andPMl()25
(47.3).
Total non-accidental, cardiovascular, and respiratory
deaths (mean 4.88, 2.87, 1.08 per day, respectively)
were related to particle mass concentration and
number counts in each size class, and to mass
concentrations of gaseous co-pollutants NO2, CO,
SO2, using GAM regression models adjusted for
temporal trends, day of week, weekly national
influenza rates, temperature and relative humidity.
Data were analyzed by season, age group, and cause
of death separately. Single-day lags and polynomial
distributed lag models (PDL) were used.  Particle
indices and pollutants were fitted using linear, log-
transformed, and LOESS transformations. Two-
pollutant models with a particle  index and a gaseous
pollutant were fitted. The "best" model as used by
Wichmann et  al. (2000) was that having the highest t-
statistic, since other criteria such as log-likelihood for
nested models and AIC for non-nested models could
not be applied due to different numbers of
observations in each model.  There should be little
difference between these approaches and resulting
differences in  results should be small in practice.
Sensitivity analyses included stratifying data by
season, winter year, age, cause of death, or
transformation of the pollution variable (none,
logarithmic, non-parametric smooth).
Non-accidental total deaths (56.6 per day) were
examined for associations with PM2 5, PMH)_2 5, 03,
CO, SO2, and NO2. Data analyzed using GAM
Poisson regression models, adjusting for temperature,
seasonal cycles. Single and two pollutant models with
lag days from 0 to 5, as well as the 2- to 5-day
average concentrations evaluated.
Loss of stat. power by using a small city with a small
number of deaths was offset by advantage of having
good exposure representation from single monitoring
site.  Since ultrafine particles can coagulate into larger
aggregates in a few hours, ultrafine particle size and
numbers can increase into the fine particle category,
resulting in some ambiguity. Significant associations
were found between mortality and ultrafine particle
number concentration (NC), ultrafine particle mass
concentration (MC), fine particle mass concentration,
or SO2 concentration. The correlation between
MCO.01-2.5 and NCO.01-0.1 is only moderate,
suggesting it may be possible to partially separate
effects of ultrafine and fine particles.  The most
predictive single-day effects are either immediate (lag 0
or 1) or delayed (lag 4 or 5 days), but cumulative
effects characterized by PDL are larger than single-day
effects. The significance of SO2 is robust, but hard to
explain as a true causal factor since its concentrations
are very low.  Age is an important modifying factor,
with larger effects at ages < 70 than > 70 years.
Respiratory mortality has a higher RR than  cardio-
vascular mortality. A large number of models were
fitted, with some significant findings of association
between mortality and particle mass or number indices.
Both PM size fractions associated with mortality, but
different effects found for warmer and colder months.
PM25 and PM,0.25 both important in whole year,
winter, and summer.  In summer, PM1(,.25 had largest
effect size estimate. NO2 and CO also associated with
mortality, as was O3 in warmer months.  No consistent
SO2-mortahty associations.
                                                                                                                                           Total mortality excess deaths:
                                                                                                                                           Filter PM10 (0-4 d lag) = 6.6 (0.7,
                                                                                                                                           12.8)per50^g/m3.  Filter PM25
                                                                                                                                           (0-1 d) = 3.0 (-1.7, 7.9). MCfor
                                                                                                                                           PM00,.256.2%(1.4,  11.2) for all
                                                                                                                                           year; by season,
                                                                                                                                           Winter = 9.2% (3.0,  15.7)
                                                                                                                                           Spring = 5.2% (-2.0, 12.8)
                                                                                                                                           Summer = -4.7% (-18.7, 11 7)
                                                                                                                                           Fall = 9.7% (1.9, 18.1)

                                                                                                                                           For ultrafine PM, NC 0.01-0 1  (0-4
                                                                                                                                           d lag):
                                                                                                                                           All Year =8.2% (0.3, 16.9)
                                                                                                                                           Winter = 9.7% (0.3,  19.9)
                                                                                                                                           Spring =10.5% (-1.4, 23.9)
                                                                                                                                           Summer = - 13.9% (-29 8, 5.7)
                                                                                                                                           Fall = 12.0% (2.1,22.7)
Percent excess total deaths per
25 ^g/m3 increase in the average of
previous two days for the whole
year: 1.8(1 3, 2.4) for PM25 and
2.3 (1.4, 3 2) for PM(,0.25) in single
pollutant models.

-------
03
3
O
o
                TABLE 6-1 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
Reference,
Location, Years,
PM Index, Mean or
Median, 1QR in jUg/
Study Description: Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, and covariates.
                                                                                       Results and Comments. Design Issues, Uncertainties,
                                                                                       Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
ON
OJ
 >»•
 £
 Tj
 H
 6
 o
 o
 H
O
 c
 o
 H
o
HH
H
          Latin America (cont'd)

          Castillejos et al. (2000).
          Mexico City.
          1992-1995.
          PM1()(44.6),PM25(27.4),
          andPMl().25(17.2).
Borja-Aburto et al. (1998)
Mexico-City,
1993-1995.
PM25(mean: 27)


Borja-Aburto et al. (1997).
Mexico-City,
1990-1992.
TSP (median. 204)
          Loomisetal. (1999).
          Mexico-City, 1993-1995.
          PM25(mean: 27 4 ^g/m3)

          Pereiraetal. (1998). Sao
          Paulo, Brazil, 1991-1992.
          PM10 (beta-attenuation, 65)
Non-accidental total deaths, deaths for age 65 and
over, and cause-specific (cardiac, respiratory, and the
other remaining) deaths were examined for their
associations with PM,,,, PM25, PMH).25, O3, and NO2.
Data were analyzed using GAM Poisson regression
models, adjusting for temperature (average of 1-3 day
lags) and seasonal cycles. Individual pollution lag
days from 0 to 5, and average concentrations of
previous 5 days were considered.

Total, respiratory, cardiovascular, other deaths, and
age-specific (age >= 65) deaths were related to PM2 5,
O3, and NO2, adjusting for 3-day lagged temperature
and periodic cycles, using Poisson GAM model.
Total, respiratory, cardiovascular, and age-specific
(age >= 65) deaths were related to O3, TSP, and CO,
adjusting for minimum temperature (temperature also
fitted seasonal cycles) using Poisson models.  The
final models were estimated using the iteratively
weighted and filtered least squares method to account
for overdispersion and autocorrelation

Infant mortality (avg.  = 3/day) related to PM, 5, O3,
and NO2, adjusting for temperature and smoothed
time, using Poisson GAM model.

Intrauterine mortality associations with PM,,,, NO,,
SO2, CO, and O, investigated using Poisson
regression adjusting for season and weather. Ambient
CO association with blood carboxyhemoglobin
sampled from umbilical cords of non-smoking
pregnant mothers studied in separate time period.
                                                                                       All three particle size fractions were associated
                                                                                       individually with mortality.  The effect size estimate
                                                                                       was largest for PM1(,_25, The effect size estimate
                                                                                       was stronger for respiratory  causes than for total,
                                                                                       cardiovascular, or other causes of death.  The results
                                                                                       were not sensitive to additions of O3 and NO2.  In the
                                                                                       model with simultaneous inclusion of PM25 and
                                                                                       PM,,,.2 5, the effect size for PM|0_2 5 remained about the
                                                                                       same, but the effect size for  PM25 became negligible.

                                                                                       PM2 5, O3, and NO2 were associated with mortality with
                                                                                       different lag/averaging periods (1 and 4 day lags; 1-2
                                                                                       avg.;  1-5 avg., respectively). PM25  associations were
                                                                                       most consistently significant SO2 was available, but
                                                                                       not analyzed because of its "low" levels.

                                                                                       O3, SO2, and TSP were all associated with  total
                                                                                       mortality in separate models, but in multiple pollutant
                                                                                       model, only TSP remained associated with mortality.
                                                                                       CO association weak.
                                                                              Excess infant mortality associated with PM2 5, NO2, and
                                                                              O3 in the same average/lags. NO2 and O3 associations
                                                                              less consistent in multi-pollutant models.

                                                                              NO2, SO2, and CO were all individually significant
                                                                              predictor of the intrauterine mortality.  NO2 was most
                                                                              significant in multi-pollutant model. PM10 and O3 were
                                                                              not significantly associated with the mortality
                                                                              Ambient CO levels were associated with and
                                                                              carboxyhemoglobin of blood sampled from the
                                                                              umbilical cords.
                                                                                                                                 Total mortality percent increase
                                                                                                                                 estimates per increase for average
                                                                                                                                 of previous 5 days: 9.5 (5.0,  14.2)
                                                                                                                                 for 50 ,ug/m3 PM,,,; 3 7 (0, 7.6) for
                                                                                                                                 25 ^g/m3 PM25; and 10.5 (6.4,
                                                                                                                                 14.8)for25//g/m3PM(l|,.25).
For total excess deaths, 3.4% (0.4,
6 4) per 25 ^g/m3 PM2 5 for both
0 and 4 d lags. For respiratory
(4d) = 6.4(-2.6, 16.2); for
CVD(4d) = 5.6(-01,11.5)

Total deaths:
6% (3.3, 8.3) per 100 ^g/m3 TSP at
0 d lag.
CVD deaths:
5.2% (0.9, 9.9).
Resp. deaths:
9.5% (1.3, 18.4).

Infant mortality excess risk:  18.2%
(6.4, 30.7) per 25 /ug/m3 PM2S at
avg. 3-5 lag days.

Intrauterine mortality excess risk:
4 !%(-!.8, 10.4)per50Mg/m3
PM,(1atOday lag.

-------
o

K>
O
O
       TABLE 6-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES

Reference,
Location, Years,
PM Index, Mean or
Median, IQR in /ug/m3.
Study Description:  Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, and covariates.
Results and Comments. Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
ON
u»
ON
          Australia

          Morgan etal. (1998).
          Sydney, 1989-1993.
          Nephelometer (0.30
          bscat/104m). Site-specific
          conversion: PM2 5 = 9;
          PM10=  18

          Simpson etal. (1997).
          Brisbane, 1987-1993.
          PM,0 (27, not used in
          analysis). Nephelometer
          (0.26 bscat/104m, size
          range: 0.01-2 /^m).
                           Total, cardiovascular, and respiratory deaths were
                           related to PM (nephelometer), O3, and NO2, adjusting
                           for seasonal cycles, day-of-week, temperature,
                           dewpoint, holidays, and influenza, using Poisson
                           GEE to adjust for autocorrelation.


                           Total, cardiovascular, and respiratory deaths (also
                           by age group) were related to PM (nephelometer), O3,
                           SO2, and NO2, adjusting for seasonal cycles, day-of-
                           week, temperature, dewpoint, holidays, and influenza,
                           using Poisson GEE to adjust for autocorrelation.
                           Season-specific (warm and cold) analyses were also
                           conducted.
                                                  PM, O3, and NO2 all showed significant associations
                                                  with total mortality in single pollutant models.
                                                  In multiple pollutant models, the PM and O3 effect
                                                  estimates for total and cardiovascular deaths were
                                                  marginally reduced, but the PM effect estimate for
                                                  respiratory deaths was substantially reduced.

                                                  Same-day PM and O3 were associated most
                                                  significantly with total deaths.  The O3 effect size
                                                  estimates for cardiovascular and respiratory deaths
                                                  were consistently positive (though not significant), and
                                                  larger in summer.  PM's effect size estimates were
                                                  comparable for warm and cold season for
                                                  cardiovascular deaths, but larger in warm season for
                                                  respiratory deaths.  NO2 and SO2 were not associated
                                                  with mortality.
                                                    4.7% (1.6, 8.0)per25,ug/m3
                                                    estimated PM2 5 or 50 ,ug/m3
                                                    estimated PM10 at avg. of 0 and 1
                                                    day lags.
                                                    (Note: converted from
                                                    nephelometry data)

                                                    3.4% (0.4, 6.4) per 25 ,ug/m3 1-h
                                                    PM2 5 increment at 0 d lag; and
                                                    7.8% (2.5,13.2) per 25 jug/m3 24-h
                                                    PM2 < increment.
          Asia
 Tl
 H
 a
 o
 x
 o
 H
O
 G!
 O
 H
          Hong etal. (1999).
          Inchon, South Korea,
          1995-1996 (20 months).
          PM,,, mean = 71.2.
Lee etal. (1999).
Seoul and Ulsan, Korea,
1991-1995. TSP(beta
attenuation, 93 for Seoul
and 72 for Ulsan)
Non-accidental total deaths, cardiovascular, and
respiratory deaths were examined for their
associations with PM,,,, O3, SO2, CO, and NO2.
Data were analyzed using GAM Poisson regression
models, adjusting for temperature, relative humidity,
and seasonal cycles. Individual pollution lag days
from 0 to 5, as well as the average concentrations of
previous 5 days were considered.

Total mortality series was examined for its
association with TSP, SO2, and O3, in Poisson GEE
(exchangeable correlation for days in the same year),
adjusting for season, temperature, and humidity.
A greater association with mortality was seen with the
5-day moving average and the previous day's exposure
than other lag/averaging time. In the models that
included a 5-day moving average of one or multiple
pollutants, PMH) was a significant predictor of total
mortality, but gaseous pollutants were not significant.
PM,,, was also a significant predictor of cardiovascular
and respiratory mortality.

All the pollutants were significant predictors of
mortality in single pollutant models. TSP was not
significant in multiple pollutant models, but SO2
and 03 remained significant.
                                                                                                                                 Percent excess deaths (t-ratio) per
                                                                                                                                 50 Atg/m3 increase in the 5-day
                                                                                                                                 moving average of PMH): 4.1 (0.1,
                                                                                                                                 8 2) for total deaths; 5.1 (0.1, 10.4)
                                                                                                                                 for cardiovascular deaths; 14.4
                                                                                                                                 (-3.2, 35.2) for respiratory deaths.
5.1% (3.1,7.2) for Seoul, and-
0.1% (-3.9, 3.9) for Ulsan, per
100^g/m3 TSP at avg. of 0, 1, and 2
day lags.
 n
 )-H
 H
 tn

-------
o
O
o
     TABLE 6-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES

Reference,
Location, Years,
PM Index, Mean or
Median, IQR in /ug/m3,
Study Description. Outcomes, Mean outcome rate,
and ages. Concentration measures or estimates.
Modeling methods:  lags, smoothing, and covariates.
Results and Comments. Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess
Risk% (95% LCL, UCL),
Co-pollutants.
CTs
T!
6
o
2
o
H
O
O
H
W
O
&
O
HH
H
W
          Asia (cont'd)

          Lee and Schwartz (1999).
          Seoul, Korea. 1991-1995.
          TSPmean = 925.
Xuetal. (2000).
Shenyang, China, 1992.
TSP (430).
Ostroetal. (1998).
Bangkok, Thailand,
1992-1995
PM10 (beta attenuation, 65)
Total deaths were analyzed for their association with
TSP, SO2, and O,.  A conditional logistic regression
analysis with a case-crossover design was conducted.
Three-day moving average values (current and two
past days) of TSP and SO2, and 1 -hr max O3 were
analyzed separately. The control periods are 7 and
14 days before and/or after the case period. Both
unidirectional and bi-directional controls (7 or
7 and 14 days) were examined, resulting in six sets
of control selection schemes. Other covariates
included temperature and relative humidity.

Total (non-accidental), CVD, COPD, cancer and
other deaths examined for their associations with
TSP and SO2,usmg Poisson (GAM, and Markov
approach to adjust for mortality serial dependence)
models, adjusting for seasonal cycles, Sunday
indicator, quintiles of temp, and humidity. Ave.
pollution values of concurrent and 3 preceding days
used.
Total (non-accidental), cardiovascular, respiratory
deaths examined for associations with PM]0 (separate
measurements showed =50% of PMH, was
PM25),using Poisson GAM model adjusting for
seasonal cycles, day-of-week, temp., humidity.
                                                                            Among the six control periods, the two unidirectional
                                                                            retrospective control schemes resulted in odds ratios
                                                                            less than 1; the two unidirectional prospective control
                                                                            schemes resulted in larger odds ratios (e.g., 1.4 for
                                                                            50 ppb increase in SO2); and bi-directional control
                                                                            schemes resulted in odds ratios between those for
                                                                            um-directional schemes. SO2 was more significantly
                                                                            associated with mortality than TSP.
Total deaths were associated with TSP and SO2 in both
single and two pollutant models. TSP was significantly
associated with CVD deaths, but not with COPD. SO2
significantly associated with COPD, but not with CVD
deaths. Cancer deaths not associated with TSP or SO2.
All the mortality series were associated with PM10 at
vanous lags. The effects appear across all age groups.
No other pollutants were examined.
                                                   OR for non-accidental
                                                   mortality per 100 ,ug/m3
                                                   increase in 3-day average
                                                   TSP was 1.010(0.988,
                                                   1.032).
Percent total excess deaths
per 100 ,ug/m3 increase in
0-3 day ave. of TSP = 1.75
(0.65, 2.85), with SO2 = 1.31
(0.14,2.49)
COPD TSP = 2.6 (-0.58,
5.89); with SO2 = 0.76
(-2.46,4.10).
CVD TSP = 2 15(0.56,
3.71);withSO2= 1.95(1.19,
3.74).
Cancer TSP = 0.87 (-1.14,
2.53), with SO, = 1.07
(-1.05,3.23). "
Other deaths TSP = 3.52
(0.82, 6.30); with SO2 = 2.40
(-0.51,5.89).

Total mortality excess risk:
5.1% (2.1, 8.3) per 50 Mg/m3
PMU)at3 d lag (0 and 2 d
lags also significant).
CVD (3d ave.) = 8.3 (3.1,
13.8)
Resp. (3d ave.) = 3.0 (-8.4,
15.9)

-------
S           TABLE 6-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES

CT       Reference,
N>       Location, Years,            Study Description. Outcomes, Mean outcome rate,                                                      PM Index, lag, Excess
O       PM Index, Mean or          and ages.  Concentration measures or estimates.        Results and Comments. Design Issues, Uncertainties,     Risk% (95% LCL, UCL),
1—1       Median, IQR in ,ug/m3.       Modeling methods: lags, smoothing, and covanates.    Quantitative Outcomes.                              Co-pollutants.

         Asia (cont'd)

         Cropper et al. (1997)        Total (by age group), respiratory and CVD deaths      TSP was significantly associated with all mortality       2.3% (significant at 0.05, but
         Delhi, India,  1991-1994      related to TSP, SO2, and NOx, using GEE Poisson      series except with the very young (age 0-4) and the       SE of estimate not reported)
         TSP (375)                 model (to control for autocorrelation), adjusting for     "very old" (age >=65).  The results were reported to      per 100 //g/m3 TSP at 2 day
                                   seasonal cycles (trigonometric terms), temperature,      be unaffected by addition of SO2 to the model. The      lag.
                                   and humidity. 70% deaths occur before age 65 (in      authors note that, because those who are affected are
                                   U.S., 70% occur after age 65).                      younger (than Western cities), more life-years are likely
             	to be lost per person from air pollution impacts.	
oo
Tl
H
6
o
2
o
H
O
O
H
m
O
w
o
H

-------
  1           As can be seen in Table 6-1, with a few exceptions, most all of the newly reported analyses
  2     continue to show statistically significant associations between short-term (24-h) PM exposures
  3     indexed by a variety of ambient PM measurements and increases in daily mortality in numerous
  4     U.S. and Canadian cities, as well as elsewhere around the world. Also, the effects estimates from
  5     the newly reported studies generally comport well with those derived from the earlier 1996 PM
  6     AQCD assessment, with the newly reported PM risk estimates generally falling within the range
  7     of ca. 1 to 8% increase in excess deaths per 50 //g/m3 PM10 and ca. 2 to 6% increase per 25 Aig/m3
  8     PM2 5. Several newly available PM epidemiology studies which conducted time-series analyses
  9     in multiple cities are of particular interest, as discussed below.
 10
 11     6.2.2.3  New Multi-City Studies
 12          The new multi-city studies are of particular interest here due to their evaluation of a wide
 13     range of PM exposures and large numbers of observations holding promise of providing more
 14     precise effects estimates then most smaller scale independent studies of single cities. Another
 15     major advantage of the multi-city studies, over meta-analyses for multiple "independent" studies,
 16     is the consistency in data handling and model specifications, which eliminates variation due to
 17     study design. Further, unlike regular meta-analysis, they clearly do not suffer from potential
 18     omission of negative studies due to "publication bias".  Furthermore, geographic patterns of air
 19     pollution effects can be systematically evaluated in multiple-city analyses. Thus, the results from
 20     multi-city studies can provide especially valuable evidence regarding the consistency and/or
 21      heterogeneity, if any, of PM-health effects relationships across geographic locations. Also, many
 22     of the cities included in these multi-city studies were ones for which no time-series analyses had
 23     been previously reported.
 24
 25      6.2.2.3.1  U.S. Multi-City Studies
 26     U.S. PM10 20-Cities and 90-Cities NMMAPS Analyses
 27           The National Morbidity, Mortality, and Air Pollution Study (NMMAPS) focused on time-
28      series analyses of PM10 effects on mortality during 1987-1994 in the 90 largest U.S. cities (Samet
29      et al., 2000a,b), in the 20 largest U.S. cities in more detail (Dominici et al., 2000), and PM10
30      effects on emergency hospital admissions in 14 U.S. cities (Samet et al., 2000a,b). These
31      NMMAPS analyses are marked by extremely sophisticated statistical approaches addressing

        March 2001                               6-39       DRAFT-DO NOT QUOTE OR CITE

-------
 1      issues of measurement error biases, co-pollutant evaluations, regional spatial correlation, and
 2      synthesis of results from multiple cities by hierarchical Bayesian meta-regressions and
 3      meta-analyses. These analyses provide extensive new information of much importance in being
 4      among that most highly relevant to the setting of U.S. PM standards, because no other study has
 5      examined as many U.S. cities in such a consistent manner. NMMAPS used only one consistent
 6      PM index (PM10) across all cities; death records were collected in a uniform manner; and
 7      demographic variables were uniformly addressed. Both the 20 and 90 cities analyses studies
 8      employ multi-stage models (see Table 6-1) in which heterogeneity in individual cities'
 9      coefficients in the first stage GAM Poisson models were evaluated in the second  stage models
10      with city or region specific explanatory variables.
11           In both the 20 and 90 cities studies, the combined estimates of PM10 coefficients were
12      positively associated with mortality at all the lags examined (0, 1, and 2 day lags), although the
13      1-day lag PMIO resulted in the largest overall combined estimate.  Figure 6-1 shows the  estimated
14      percent excess total deaths per 10 /ag/m3 PM,0 at lag 1 day in the 90 largest cities, as well as
15      (weighted average) combined estimates for U.S. geographic regions depicted in Figure 6-2.  The
16      majority of the coefficients were positive for the various  cities listed along the left axis of
17      Figure 6-1.  See Appendix 6-A for names of cities designated by the abbreviations, e.g.,
18      seat = Seattle. The estimates for the individual cities were first made independently, without
19      borrowing information from other cities. The cities were then grouped into the 7 regions seen in
20      Figure 6-2 (based  on characteristics of the  ambient PM mix typical of each region, as delineated
21      in the 1996 PM AQCD).  The bolded segments represent the posterior means and 95% posterior
22      intervals of the pooled regional effects under the more conservative prior A for the heterogeneity
23      across both regions and cities within regions. The solid circles and squares denote, respectively,
24      the overall regional means without and with borrowing information from other regions, ("overall
25      1" = the regional mean without other regions, "overall 2" = with information from other regions).
26      The triangles and bolded segments at the bottom of Figure 6-1 display combined  estimates of
27      nationwide  overall effects of PM10 for all cities overall, and for all cities minus those in the
28      Northeast (overall-north).
29           Note that there appears to be some regional-specific variation in the overall combined
30      estimates, shown as "overall 1" and "overall 2" for the two sets of modeling assumptions and
31      specifications used in analyses combining data from all the cities in a given region. This can be

        March 2001                               6-40        DRAFT-DO NOT QUOTE OR CITE

-------
            % Increase in Mortality per 10 |jg/m3 Increase in PM
                                                              10



oakl







spok
overall 1








overall 1
la





overall 1
mmn

desm
tope
overall 1
chic
det
clev
pitt
buff


indi
lOUi







overall 1
ny
phtl








syra
nor
artv
overall 1





tamp
orla
jckv
birm
nash

ral
batr
Itrk
gmb
shre

overall 1
overall- north
overall
i i





z





•








•








1 ft
	 ^-Q 	


— <





	 O
















	 e 	 ^










	 e—

—



	 o
•

1 i


o
ft










fc "


f^ *J


*r




JP



^

• 0





f^
o



0 '



»°
Q 	 	
Qu
A



„
0 n
o ....
	 2_e 	







	 ^-e- 	
— e 	 	 —
	 e — 	





o 	


5sr
                                                                      northwest
                                                                     southwest
                                                                     southcal
                                                                     uppermidwest
                                                                     industmidwest
                                                                     northeast
                                                                     southeast
                                                                     overall
Figure 6-1. Estimated excess risks for PM mortality (1 day lag) for the 90 largest U.S. cities
           as shown in the original NMAPS report. From Samet et al. (2000a,b).
March 2001
6-41
DRAFT-DO NOT QUOTE OR CITE

-------
                  Northwest
       Southern
       California
                                    Upper
                                  Midwest
Industrial
 Midwest
                                                                         Northeast
                            Southwest
                                           Southeast
       Figure 6-2.  Map of the United States showing the 90 cities (the 20 cities are circled) and
                  the seven regions considered in the NMMAPS geographic analyses. Regions:
                  Northwestern; Southern California; Southwest; Upper Midwest; Industrial
                  Midwest; Northeast; Southeast.
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
discerned more readily in Figure 6-3 (which depicts overall region-specific excess risk estimates
for day 0 and 2 day lags, as well as for lag 1 day).  For example, the coefficients for the Northeast
are generally higher than for other regions (the Northeast combined estimate, 4.5% excess total
deaths per 50 /ug/m3 increase in PM10, was about twice that for the 90-cities overall). The overall
national combined estimate (i.e., at lag 1 day, 2.3% excess total deaths per 50 /ug/m3 increase in
PM,0) for the 90 cities is consistent with the range of estimates reported in the 1996 PM AQCD.
     In the 90 cities study, the weighted second-stage regression included five types of county-
specific variables: (1) mean weather and pollution variables; (2) mortality rate (crude mortality
rate); (3) socio-demographic variables (% not graduating from high school and median household
income); (4) urbanization (public transportation); (5) variables related to measurement error
(median of all pair-wise correlations between monitors). Some of these variables were
apparently correlated (e.g., mean PM10 and NO2, household income and education) so  that the
       March 2001
                                      6-42
DRAFT-DO NOT QUOTE OR CITE

-------

CO
I
o 2-

I— O
&s
.£•0-
».£
t 0)
is
CD 0 —
- b
0 C
V)
CD

b "1 ~~
*
A -£•
^ x X
X X X X X X ,
1 1 1 1 1 1
• i i r i i
i i i i i i
i i i i ill
i i i i | | 1 |
, . . i 1 + : + +|t | fj+ + f |t " j| j
T i | i i , , , , ±
i i i i i i T

i i i r i i
i i i i i i
i i i i i i
O-'-CMO-'-CNO'-CMOv- CNO^CMO^CM O^~
QlOiCTOJOOiOOOJOlCwOJCDCBCSD)1-"^ D) O)
COn3CO(OCOCOfOfOcOCOfOCO(OttJCOCO™
-------
 1      PM2 5 and, thus, of PM,0 when it is not dominated by the coarse particle fraction.  The
 2      investigators concluded that the PM10 effect on mortality "did not appear to be affected by other
 3      pollutants in the model".
 4
 5      U.S. 10-Cities Studies
 6           In another set of multi-city analyses, Schwartz (2000a,b), Schwartz and Zanobetti (2000),
 7      Zanobetti and Schwartz (2000), and Braga et al. (2000) analyzed 1987-1995 air pollution and
 8      mortality data from ten U.S. cities (New Haven, CT; Pittsburgh, PA; Birmingham, AL; Detroit,
 9      MI; Canton, OH; Chicago, IL; Minneapolis-St. Paul, MN; Colorado Springs, CO; Spokane, WA;
10      and Seattle, WA.) or subsets (4 or 5 cities) thereof.  The selection of these cities was based on the
11      availability of daily (or near daily) PM10 data. The main results of the study were presented in the
12      Schwartz (2000a) paper and the other studies noted above focused on each of several specific
13      issues, including: potential confounding, effect modification, distributed lag, and threshold.
14      In this section, the results for the Schwartz (2000a)  main analyses and that of Braga et al. (2000)
15      on confounding are discussed, and results for analyses of other specific issues are discussed later
16      in appropriate sections.  For each of the  10 cities, daily total (non-accidental) mortality was fitted
17      using a GAM Poisson model adjusting for temperature, dewpoint, barometric pressure, day-of-
18      week, season, and time. Deaths stratified by location of death  (in or outside hospital) were also
19      examined. The data were also analyzed by season (November through April as heating season).
20      In the second stage, the PM10 coefficients were modeled as a function of city-dependent
21      covariates including co-pollutant to PM10 regression coefficient (to test potential confounding),
22      education, unemployment rate, poverty level, and percent non-white. Threshold effects were also
23      examined. The inverse variance weighted averages of the ten cities' estimates were used to
24      combine results.  PM10was significantly associated  with total deaths, and the effect size estimates
25      were the same in summer and winter. Adjusting for other pollutants did not substantially change
26      the PM10 effect size estimates.  The socioeconomic variables did not modify the estimates.  The
27      effect size estimates for the deaths outside hospital were substantially greater than for inside
28      hospital.  The combined percent  excess death estimate for total mortality was 3.4% (95% CI:
29      2.7-4.1) per 50 Mg/m3 increase in PM10, but was larger for days with PM10 < 50 /ug/m3.
30           Braga et al. (2000) evaluated potential confounding of the reported PM-mortality
31      associations by effects of respiratory epidemics, using data from a subset of 5 of the  10 cities

        March 2001                                6-44         DRAFT-DO NOT QUOTE OR CITE

-------
  1      evaluated by Schwartz (2000a).  When adjustments were made for respiratory epidemics, small
  2      decreases in PM10 effects were seen in the cities evaluated. The overall estimated percent excess
  3      deaths per 50 /ug/m3 PM,0 for the five cities was 4.3% (CI 3.0, 5.6) without control for respiratory
  4      epidemics, but slightly decreased to 4.0% (CI 2.6, 5.3) with control for epidemics.
  5
  6      U.S. 3-Cities Study
  7           Moolgavkar (2000a) evaluated associations between short-term measures of major air
  8      pollutants and daily deaths in three large U.S. metropolitan areas (Cook Co., IL, encompassing
  9      Chicago; Los Angeles Co., CA; and Maricopa Co., AZ, encompassing Phoenix) during a 9-year
 10      period (1987-1995). Generalized additive models (GAM) were used in a standard manner to
 11      conduct time-series Poisson regression analyses independently for each of the three cities
 12      (allowing comparison of results across them not due to methodological differences), but no
 13      combined analyses were attempted to derive overall PM effects estimates. Total non-accidental
 14      deaths and cause-specific deaths from cardiovascular disease (CVD), cerebrovascular disease
 15      (CrD), and chronic obstructive lung disease (COPD), and associated conditions were analyzed in
 16      relation to 24-h readings for PM, O3, CO, NO2, SO2 averaged over all monitors in a given county.
 17      Daily readings were available for each of the gaseous pollutants in all three countries, as were
 18      PM10 values for Cook County. However, PM10 values were only available every sixth day in
 19      Maricopa and Los Angeles Counties; as were PM25 values in Los Angeles Co.  PM values were
 20      highest in the winter and fall in Los Angeles Co., in the fall in Maricopa Co., and in summer in
 21      Cook Co., whereas the gases (except for O3) were highest in winter in all three counties (O3 was
 22      highest in summer in all three). The PM indices were moderately correlated (r = 0.30 to 0.73)
 23      with CO, NO2, and SO2 in Cook Co. and Los Angeles Co., but poorly correlated (r < 0.22) with
 24      those gases in Maricopa Co.  Ozone was very poorly (r < 0.20) or negatively correlated with PM
 25      or the other gases in each location (except for Cook Co., r = 0.36 for O3 vs PM10). Total
26      non-accidental, CVD, and COPD deaths were all highest during winter in all three counties, but
27      CrD deaths were relatively constant from season to season (no season-specific analyses reported).
28           Controlling for temperature and relative humidity effects in separate analyses for each
29      mortality endpoint for each of the three countries, varying patterns of results were found from
30      one location to another, as noted in Table 6-1. In general, although PMIO in each of the three
31      counties (and PM2 5 in Los Angeles) and each of the gaseous pollutants (except O3) were all

        March 2001                               6-45        DRAFT-DO NOT QUOTE OR CITE

-------
 1      statistically significantly associated with total non-accidental mortality at one or more lag times
 2      (0 to 5 days) in single pollutant models, the PM effect estimates tended to be reduced and non-
 3      significant in many of the multi-pollutant (PM plus one other gas or PM  plus all others) analyses.
 4      In contrast, effect estimates for several of the gases (CO, SO2, and NO2) tended to be more robust
 5      than those for PM in multi-pollutant models, with their estimates remaining statistically
 6      significant (although usually somewhat attenuated) at one or more lag times when included in
 7      multi-pollutant models with PM10 or PM2 s. Similarly,  a somewhat analogous varying pattern of
 8      results was observed for the cause-specific mortality analyses (discussed further below in Section
 9      6.2.2.5).  That is, although PM10 or PM2 5 were statistically significantly related to CVD and
10      COPD-related (and to CrD only in Maricopa Co., lag 5) mortality in single pollutant models,
11      their coefficients were typically markedly reduced and  became non-significant in multi-pollutant
12      analyses with one or more of the gases included in the model. Moolgavkar (2000a) concluded
13      that, while direct effects of individual components of air pollution cannot be ruled out, individual
14      components can best be thought of as indices of the overall air pollution  mix; and he noted
15      considerable heterogeneity of air pollution effects across the three geographic areas evaluated.
16
17      6.2.2.3.2 Canadian Multi-City Study Analyses
18      Urban Air Pollution Mix and Daily Mortality in 11  Canadian Cities
19           The number of daily deaths for non-accidental causes during 1980-1991 were obtained for
20      11 Canadian cities and linked to concentrations of ambient gaseous air pollutants using relative
21      risk regression models for longitudinal count data (Burnett et al., 1998a). The GAM Poisson
22      models used evaluated daily mortality versus O3, NO2,  SO2 and CO (including adjustments for
23      seasonal cycles, day-of-week effects, and weather effects), but no PM indices were included in
24      their analyses because daily PM measurements  were not available.  However, data were available
25      for fine and coarse PM mass from dichot samples, and  sulfates, on variable schedules somewhat
26      more frequently than once per six  days in Montreal, Toronto, and Windsor (with smaller
27      numbers in the other cities). This  allowed an ecologic  comparisons of gaseous pollutant risks by
28      mean fine particle concentration (their Figure 1).  These comparisons suggested a weak negative
29      confounding of NO2 and SO2 effects with fine particles, and a weak positive confounding of
30      particle effects with O3.
31

        March 2001                              6-46        DRAFT-DO NOT QUOTE OR CITE

-------
  1      Eight Largest Canadian Cities Study
  2           Burnett et al. (2000) analyzed various PM indices (PM10, PM2 5, PMIO_2 5, sulfate, COH, and
  3      47 elemental component concentrations for fine and coarse fractions) and gaseous air pollutants
  4      (NO2, O3, SO2, and CO) for association with total mortality in the 8 largest Canadian cities:
  5      Montreal, Ottawa-Hull, Toronto, Windsor, Winnipeg, Calgary, Edmonton, and Vancouver. This
  6      study differs from (Burnett et al., 1998a), including fewer cities but more recent years of data
  7      (1986-1996 vs.  1980-1991) and detailed analyses of particle mass components by size and
  8      elemental composition. Each city's mortality, pollution, and weather variables were separately
  9      filtered for seasonal trends and day-of-week patterns.  The residual series from all cities were
 10      then combined and analyzed in a GAM Poisson model. The weather model was selected from
 11      spline-smoothed functions of temperature, relative humidity, and maximum change in barometric
 12      pressure within a day and with 0 and  1 day lags, using forward stepwise procedures. Pollution
 13      effects were examined at lags 0 through 5 days. To avoid unstable parameter estimates in multi-
 14      pollutant models, principal components were also used as predictors in the regression models.
 15           Ozone was weakly  correlated with other pollutants, and other pollutants were "moderately"
 16      correlated with each other (the highest was r = 0.65 for NO2 and CO). The strongest association
 17      with mortality for all pollutants considered were for 0 or 1 day lags.  PM2 5 was a stronger
 18      predictor of mortality than PM10_2 5. The gaseous pollutant effects estimates were generally
 19      reduced by inclusion of PM25 or PM10, but not PM10_25, where strength of prediction is measured
 20      by the t value or statistical significance of the excess risk. In addition to the results implicating
 21      the fine particle fraction (PM2 5) most clearly, other findings on fine particle components were
 22      also of interest.  Specifically, sulfate,  Fe, Ni, and Zn were most strongly associated with
 23      mortality. The total effect of these four components was greater than that for PM2 5 mass alone,
 24      the authors suggesting that the characteristics of the complex chemical mixture in the fine
 25      fraction may be a better predictor of mortality than the mass index alone.
 26
27      6.2.2.3.3 European Multi-City APHEA Study Analyses
28           The Air Pollution and Health: a European Approach (APHEA) project is a multi-center
29      study of short-term effects of air pollution on mortality and hospital admissions during the period
30      1975-1992, using data from 15 European cities with a wide range of geographic, climatic,
31      sociodemographic, and air quality patterns.  The obvious strength of this approach is to be able to

        March 2001                               6-47        DRAFT-DO NOT QUOTE OR CITE

-------
 1      evaluate potential effect modifiers in a consistent manner. It should be noted that PM indices
 2      measured in those cities were mostly black smoke (BS), except for:  Paris, Lyon (PM13);
 3      Bratislava, Cologne, and Milan (TSP); and Barcelnoa (BS and TSP). As discussed below, there
 4      have been three papers published that presented either a meta-analysis or pooled summary
 5      estimates of these multi-city mortality results: (1) Katsouyanni et al. (1997)-SO2 and PM results
 6      from 12 cities; (2) Touloumi et al. (1997)-ambient oxidants (O3 and NO2) results from six cities;
 7      and (3) Zmirou et al. (1998)- cause-specific mortality results from 10 cities (see Section 6.2.2.5).
 8
 9      APHEA Sulfur Dioxide and Particulate Matter Results for 12 Cities
10          The Katsouyanni et al. (1997) analyses evaluated data from the following cities: Athens,
11      Barcelona, Bratislava, Cracow, Cologne, Lodz, London, Lyons, Milan, Paris, Poznan, and
12      Wroclaw.  In the western European cities, an increase of 50 /ug/m3 in SO2 or BS was associated
13      with a 3% (95% CI = 2.0, 4.0) increase in daily mortality;  and the corresponding figure was 2%
14      (95% CI = 1.0, 3.0)  for estimated PMIO (they used conversion:  PMIO = TSP*0.55). In the
15      central/eastern European cities, the increase in mortality associated with a 50 /^g/m3 change was
16      0.8% (CI = -0.1, 2.4) for SO2 and 0.6% (CI = 0.1, 1.1) per 50 Mg/m3 change in BS. Estimates of
17      cumulative effects of prolonged (two to four days) exposure to air pollutants were comparable to
18      those for one day effects.  The effects of both pollutants (BS, SO2) were stronger during the
19      summer and were mutually independent. Regarding the contrast between the western and
20      central/eastern Europe results, the authors speculated that this could be due to: difference in
21      exposure representativeness; difference in pollution toxicity or mix; difference in proportion of
22      sensitive sub-population; and model fit for seasonal control. Bobak and Roberts (1997)
23      commented that  the heterogeneity between central/eastern and western Europe could  be due to
24      the difference in mean temperature. However, Katsouyanni and Touloumi (1998) noted that,
25      having examined the source of heterogeneity, other factors could apparently explain the
26      difference in estimates as well as  or better than temperature.
27
28      APHEA Ambient Oxidants (Ozone and Nitrogen  Dioxide) Results for Six Cities
29          Touloumi et al. (1997) reported on additional APHEA data analyses, which evaluated
30      (a) short-term effects of ambient oxidants on daily deaths from all causes (excluding accidents),
31      and (b) impacts on effect estimates for NO2 and  O3 of including a PM measure (BS) in

        March 2001                               6-48        DRAFT-DO NOT QUOTE OR CITE

-------
  1      multi-pollutant models. Six cities in central and western Europe provided data on daily deaths
  2      and NO2 and/or O3 levels. Poisson autoregressive models allowing for overdispersion were
  3      fitted.  Significant positive associations were found between daily deaths and both NO2 and O3.
  4      Increases of 50 /ug/m3 in NO2 (1-hour maximum) or O3 (1-hour maximum) were associated with
  5      a 1.3% (95% CI 0.9-1.8) and 2.9% (95% CI 1.0-4.9) increase in the daily mortality, respectively.
  6      There was a tendency for larger effects of NO2 in cities with higher levels of BS:  when BS was
  7      included in the model, the pooled estimate for the O3 effect was only slightly reduced, but the
  8      coefficient for NO2 was reduced by half (but remained significant). The authors speculated that
  9      the short-term effects of NO2 on mortality might be confounded by other vehicle-derived
 10      pollutants (e.g., airborne ambient  PM indexed by BS measurements). Thus, while this study
 11      reports only relative risk levels for NO2 and O3 (but not for BS), it illustrates the importance of
 12      confounding of NO2 and PM effects and the relative limited confounding of O3 and PM effects.
 13
 14      6.2.2.3.4 Comparison of Effects Estimates from Multi-City Studies
 15           In summary, based on pooled analyses of data combined across multiple cities, the percent
 16      excess (total, non-accidental) deaths estimated per 50 yUg/m3 increase in PM10 in the above
 17      multi-city studies were:  (1) 2.3% in the 90 largest U.S. cities (4.5% in the Northeast region);
 18      (2) 3.4% in 10 U.S. cities; (3) 3.5% in the 8 largest Canadian cities; and (4) 2.0% in western
 19      European cities (using PM10 =  TSP*0.55).  These combined estimates are  all consistent with the
 20      range of PM10 estimates previously reported in the 1996 PM AQCD.
 21
 22      6.2.2.4  The Role of Particulate  Matter Components
 23           Delineation of the roles of specific ambient PM components in contributing to associations
 24      between short-term PM  exposures and mortality requires  evaluation of several factors, e.g., size,
 25      chemical composition, surface characteristics, and presence of gaseous co-pollutants. While
26      possible combinations of interactions among these factors can in theory be limitless, the actual
27      data tend to cover definable ranges of aerosol characteristics and co-pollutant environments due
28      to typical source characteristics (e.g., fine particles tend to be  combustion products in most
29      cities).  Newly available studies conducted in the last few years have begun to provide more
30      extensive information on the issue of PM component roles; their results are discussed below in


        March 2001                               6-49        DRAFT-DO NOT QUOTE OR CITE

-------
  1      relation to three topics:  (1) PM particle size (e.g., PM2 5 vs. PM10.2 5); (2) chemical components;
  2      and (3) source oriented evaluations.
  3
  4      6.2.2.4.1  Particulate Matter Particle Size Evaluations
  5           Numerous new studies published since the 1996 PM AQCD substantiate associations
  6      between PM2 5 and increased total mortality. Consistent with the 1996 PM AQCD findings,
  7      effect size estimates from the new studies generally fall within the range of 2 to 6% excess total
  8      mortality per 25 /u.g/m3 PM2 5, with many being statistically significant at p<0.05.
  9           With regard to the relative importance of fine and coarse particles, at the time of the 1996
10      PM AQCD, there was only one acute mortality study (Schwartz et al., 1996a), in which this issue
11      was examined. That study suggested that fine particles, but not coarse particles, were associated
12      with daily mortality. A recent study (Klemm and Mason 2000) to reconstruct the data and to
13      replicate the analyses essentially reproduced the original investigators' results.  Since the 1996
14      PM AQCD, several new studies used size-fractionated PM data to investigate the relative
15      importance of fine (PM2 5) vs. coarse (PM10_2 5) particles.
16           In Table 6-2, synopses of these studies with regard to the relative importance of the two
17      size fractions, as well as some characteristics of the data, are provided.  The average levels of
18      PM2 5  ranged from about 13 to 20 Mg/m3 in the U.S. cities, but much higher average levels were
19      measured in Mexico City (27.4 //g/m3) and Santiago, Chile (64.0 /^g/m3).  As can be seen in
20      Table 6-2, in the northeastern U.S.  cities (Pittsburgh, Philadelphia, and Detroit) and Atlanta, GA,
21      there was more PM2 5 mass than PM|0.2 5 mass on the average, whereas in the western U.S.
22      (Phoenix, AZ; Coachella Valley, CA; Santa Clara County, CA) the average PM10_25 levels were
23      higher than PM2 5 levels.  It should  be noted that the three Phoenix studies in Table 6-2 use much
24      the same data set, with fine and coarse particle data from EPA's 1995-1997 platform study.
25      Seasonal differences in PM component levels should also be noted.  For example, in Santa Clara
26      County and in Santiago, Chile, the  winter PM2 5 levels averaged twice those during summer. The
27      temporal correlation between PM2 5 and PM10_2 5 ranged between 0.30 and 0.65. Such differences
28      in ambient PM mix characteristics  from season to season or from location to location
29      complicates assessment of the relative importance of PM25 and PM10.25.
30           To facilitate a quantitative overview of the effect size estimates and their corresponding
31      uncertainties from these studies, the percent excess risks are plotted in Figure 6-4. These

        March 2001                                6-50        DRAFT-DO NOT QUOTE OR CITE

-------
        TABLE 6-2. SYNOPSIS OF SHORT-TERM MORTALITY STUDIES THAT
             EXAMINED RELATIVE IMPORTANCE OF PM7 < VERSUS PM
                                                                     -2.5
                                                                                      MO-2.5
 Author, City
 Means (/zg/m3); ratio of
 PM25 to PM10; and
 correlation between PM2 5
 andPM,0.25.
 Results regarding relative importance of PM25 vs. PM,(
 and comments.
 Fairley(1999).
 Santa Clara County,
 CA
 Ostro et al. (2000).
 Coachella Valley, CA
 Clyde etal. (2000).
 Phoenix, AZ
 Mar et al. (2000).
 Phoenix, AZ
 1995-1997.
 Smith et al. (2000).
 Phoenix, AZ
 Lippmann et al.
 (2000). Detroit, MI
 1992-1994.
 Lipfertetal. (2000a).
 Philadelphia, PA
 1992-1995.
PM25 mean = 13;
PM25/PM10 = 0.38;
r = 0.51.
PM25 (Palm Springs and
Indio, respectively)
mean= 12.7, 16.8;
PM25/PM10 = 0.43,0.35;
r = 0.46, 0.28.

PM25mean = 13.8;
PM25/PMIO = 0.30;
r = 0.65.
PM25(TEOM)mean=13;
PM25/PM10 = 0.28;
r = 0.42.
Not reported, but likely
same as Clyde's or Mar's
data from the same
location.
PM2Jmean=18;
PM25/PM10=0.58;
r = 0.42.
PM25mean=17.3;
PM25/PMIO=0.72.
 Of the various pollutants including PM10, PM25, PMi0.25,
 sulfates, nitrates, COH, CO, NO2, and O3, strongest
 associations were found for ammonium nitrate and PM25.
 PM2 5 was significantly associated with mortality, but
 PM10.25 was not, separately and together in the model.
 Sulfate was a significant predictor of mortality in single
 pollutant model, but not when PM2 5 was included
 simultaneously. Winter PM2 5 level is more than twice that
 in summer.

 Total mortality was more significantly associated with PM2 5
 than with PMi0.25. Cardiovascular mortality was associated
 with PM,0.25more significantly than  with PM25, but their
 effect size estimates per IQR were similar.
Using Bayesian Model Averaging that incorporates model
selection uncertainty, with 29 covariates (lags 0- to 3-day),
effects of coarse particles (most consistent at lag 1 day)
were found to be stronger than that for fine particles. The
association was for mortality confined to the region where
fine particles (PM2 5) are expected to be uniform.

Total mortality was weakly (p < 0.10) associated with
PM10_25.  It was less strongly (p > 0.10) associated with
PM2 5. Cardiovascular mortality was both significantly
associated with PM25 (lags 1, 3, 4) and PM]0.25 (lag 0), with
similar effect size estimates.

In linear PM effect model, a statistically significant
mortality association found with PMI0.2 5, but not with PM25.
In models allowing for a threshold, evidence of a threshold
for PM2 5 (in the range of 20-25 Mg/m3) suggested, but not
for PM10.2 5.  Seasonal interaction in the PM,0.2 5 effect also
reported: the effect being highest in spring and summer
when anthropogenic concentration of PM10.25 is lowest.

Both PM25 and PM10.25 were positively associated  with
mortality outcomes to a similar extent. Simultaneous
inclusion of PM25 and PMI0.25 also resulted in comparable
effect sizes. Similar patterns were seen in hospital  admission
outcomes.

The authors conclude that no systematic differences were
seen according to particle size or chemistry. However,
when PM2 5 and PM,0.2 5 were compared, PM2 5 (at lag 1 or
average of lag 0 and 1) was more significantly (with larger
attributable risk estimates) associated with  cardiovascular
mortality than PM,0.25.
March 2001
                          6-51
             DRAFT-DO NOT QUOTE OR CITE

-------
         TABLE 6-2 (cont'd). SYNOPSIS OF SHORT-TERM MORTALITY STUDIES THAT
                     EXAMINED RELATIVE IMPORTANCE OF PM,, VS. PM
                                                                          2.5
                                                                                    l!0-2.5
        Author, City
Means (/^g/m3); ratio of
PM25toPM10;and
correlation between PM2 5
and PM10.25.
Results regarding relative importance of PM2 5 vs. PM10.2 5
and comments.
        Klemm and Mason
        (2000). Atlanta, GA
        Chock et al. (2000).
        Pittsburgh, PA
        Burnett et al. (2000)
        8 Canadian cities
        1986-1996

        Castillejos et al.
        (2000). Mexico City.
        1992-1995
        Cifuentes et al.
        (2000). Santiago,
        Chile
        1988-1996.
PM2 5 mean =19.9;
PM25/PM10=0.65.
Data distribution not
reported.
PM25/PMIO= 0.67.
PM25 mean=13.3;
PM25/PMIO=0.51;
r = 0.37.

PM25mean=27.4;
PM25/PM10=0.61;
r = 0.52.
PM25mean=64.0;
PM25/PMIO=0.58;
r = 0.52.
No significant associations were found for any of the
pollutants examined, possibly due to a relatively short study
period (1-year).  The coefficient and t-ratio were larger for
PM25thanforPMlo.25.

Seasonal dependence of correlation among pollutants, multi-
collinearity among pollutants, and instability of coefficients
were all emphasized in discussion and conclusion. These
considerations and small size of dataset stratified by age
group and season limit confidence in results finding no
consistently significant associations for any size fraction.

PM2 5 was a stronger predictor of mortality than PM]0.2 5.
For chemical species, sulfate ion, nickel, and zinc from the
fine fraction were most strongly associated with mortality.

Both PM25 and PM,0.25 were associated individually with
mortality, but the PM10.2 5 effect size was larger and more
significant. When both were included in the model, the
effect size of PM10.2 5 remained the same but that of PM2 5
was virtually eliminated.

Results were different for warmer and colder months. PM25
was more important than PM,0.2 5 in the whole year and in
winter, but not in summer. The mean of PM2 s was more
than twice higher in winter (82.4 /ug/m3) than in summer
(32.8), whereas the mean of PM10.2 5 was more comparable
for winter (49.9 /ug/m3) and for summer (42.9).	
1      excluded the Clyde et al. study, in which the model specification did not obtain RRs for PM2 5

2      and PMI0.2 5 separately, and the Smith et al. study, which did not present linear term RRs for

3      PM2 5 and PM,0_2 5. Note that, in most of the  original studies, the RRs were computed for

4      comparable distributional features (e.g., inter quartile range, mean, 5th-to-95th percentile, etc.).

5      However, the increments derived and their absolute values varied across studies; and therefore,

6      the RRs used in deriving the excess risk estimates delineated in Figure 6-4 were re-computed for

7      consistent increments of 25 Aig/m3 for both PM2 5 and PM10.2 5.  Note also that re-computing the

8      RRs per 25 yUg/m3 in some cases changed the relative effect size between PM2 5 and PM10_2 5, but

9      it did not affect the relative significance.
       March 2001
                         6-52
             DRAFT-DO NOT QUOTE OR CITE

-------
 .

K)
O
O
Tl
H
o
o
2
O
H
/O
W
O
&
n
h-H
H
W
                  Klemm etal  (2000) _
          Harvard 6 Cities (recomputed)

                  Burnett et al  (2000) _
                   8 Canadian Cities
                  Chock etal (2000) _
                     Pittsburgh, PA

             Klemm and Mason (2000)
                       Atlanta, GA
                  bpfertetal (2000)
                    Philadelphia, PA
                 Lippman et al (2000)
                        Detroit, Ml ~
    Mar et al (2000) _
        Phoenix, AZ

      Fairley(1999) _
    Santa Clara Co

   Ostroetal (2000)
 Coachella Valley, CA

Castillejos et al (2000) _
  Mexico City, Mexico

 Cifuentesetal (2000)
     Santiago, Chile
                             Percent excess death  (total unless otherwise noted) per
                                   25 ug/m3 increase in PM2.s (•) or PM10-2.5  (o).
                                -5   -4   -3   -2   -1
                                                                                         10   11   12  13   14   15
                            Lag CM-1 day
                            Lag 3 day
                            Lag 1 day
Lag 0 day
Lag 4 day
Lag 5 day MA i



Lag 2 day MA}
                                                                              }age<75

                                                                              }age > 75
                                                                       - cardiovascular
                                                                        mortality
                              . } All year
                                                                        } Winter
                                                                                                , } Summer
        Figure 6-4. Percent excess risks estimated per 25 //g/m1 increase in PM2 s or PM10.2 5 from new studies evaluating both PM2 5
                    and PM,0.2 5 data for multiple years. All lags = 1 day, unless indicated otherwise.

-------
 1           All of the studies found positive associations between both the fine and coarse PM indices
 2      and increased mortality risk, with most for PM2 5 and a few for PM10_2 5 being statistically
 3      significant (at p < 0.05).  Unfortunately, most of the studies did not have large enough sample
 4      sizes to separate out what often appear to be relatively small differences in effect size estimates;
 5      but several do show statistical distinctly larger and significant mortality associations with PM2 5
 6      than for non-significant PM10.2 5 effects.  For example, the Klernm et al. (2000) recomputation of
 7      the Harvard Six Cities time-series study reconfirmed the original Schwartz et al. (1996a) finding
 8      of PM2 5 being significantly associated (at p < 0.05) with excess mortality, whereas PM10_2 5 was
 9      not. Similar results were obtained by the other multi-city study, i.e., the 8 largest Canadian cities
10      study by Burnett et al. (2000), and by the Atlanta (Klemm et al. 2000), Santa Clara (Fairley et al.,
11      1999), and the Coachella Valley (Ostra et al., 2000) studies. There were two studies in which the
12      importance of PM2 5 and PM10.2 5 were considered to be similar or,  at least, not distinguishable:
13      Philadelphia, PA (Lipfert et al., 2000a) and Detroit, MI (Lippmann et al., 2000). The three
14      Phoenix studies obtained "mixed" results, in that the Smith et al. (2000) and Clyde et al. (2000)
15      analyses found PM10,2 5 to appear to be more important in explaining mortality than PM2 5 but
16      Mar et al. found both to be significant, as depicted in Figure 6-4. Also, the Mexico City analysis
17      by Castillejos et al. (2000) implicated PM,0_2 5 as the apparent more important fraction of PM10.
18      However, the Santiago, Chile study (Cifuentes et al., 2000) found  significant associations with
19      both fine and coarse fractions and interesting seasonal differences, as well. In Chock et al.'s
20      (2000) analysis of Pittsburgh, PA data, the  authors emphasized the lack of significant PM
21      associations; and no specific comments were made regarding the relative importance of PM2 5 vs.
22      PM10.25.
23           The Canadian 8-city study (Burnett et al., 2000) is noteworthy for a variety of reasons,
24      including the use of elemental composition and principal  components analyses to provide
25      additional information about the relative  importance of fine and coarse particles. The PM2 5
26      effect on mortality is greater than the PM10_2 5 effect for all gaseous-pollutant models in Table 5 of
27      Burnett et al. (2000) and in the principal  component model 1 in their Table 8, where both PM
28      size fractions and the four gaseous co-pollutants are used simultaneously. PM component
29      models from this study are discussed further below, in Section 6.2.2.4.2.
30           The Lippmann et al. (2000) results  for Detroit are also noteworthy in that additional PM
31      indices were evaluated besides those depicted in Figure 6-4 and the overall results obtained may

        March 2001                               6-54       DRAFT-DO NOT QUOTE OR CITE

-------
  1      be helpful in comparing fine- versus coarse-mode PM effects. In analyses of 1985 to 1990 data,
  2      PM-mortality relative risks and their statistical significance were generally in descending order:
  3      PM,0, TSP-SCV, and TSP-PM]0.  For the 1992-1994 period, relative risks for equivalent
  4      distributional increment (e.g., IQR) were comparable among PM10, PM2 5, and PM10_2 5 for both
  5      mortality and hospital admissions categories; and SO4= was more strongly associated with most
  6      outcomes than H+.  Consideration of the overall pattern of results led the authors to state that the
  7      mass of smaller size index could explain a substantial portion of the variation in the larger size
  8      indices. In these data, on average, PM2 5 accounted for 60% of PMIO (up to 80% on some days)
  9      and PM10 for 66% of TSP mass. Also, the temporal correlation between TSP and PM2 5 was r =
 10      0.63, and for PM2 5 vs. PM10 r = 0.90, suggesting that much of the apparent larger particle effects
 11      may well be mainly driven by temporally covarying smaller PM2 5 particles. The stronger
 12      associations for sulfates than FT, suggestive of non-acid fine particle effects, must be caveated by
 13      noting the very low H+ levels present (often circa non-detection limit).
 14           Three research groups have examined the same Phoenix, AZ data set using different
 15      methods. While these groups used somewhat different approaches, there is some consistency
 16      among their results in that PM10_2 5 appeared to emerge as possibly the more important predictor
 17      of mortality versus PM2 5. In the Clyde et al. (2000) analysis, PM-mortality associations were
 18      found only for the geographic area where PM2 5 was considered uniformly distributed, but the
 19      association was with PM10.2 5, not PM2 5. Based on the Bayes Information Criterion, the highly
 20      ranked models consistently included 1-day lagged PM10_2 5.  In the Mar et al. (2000) analysis, total
 21      mortality was significantly associated with CO and NO2 and weakly (p < 0.1) associated with
 22      PM10 and PM10_2 5 (and PM2 5 was more weakly associated), although cardiovascular mortality was
 23      significantly associated with both PM2 5 and PMIO_2 5 at p<0.05. Smith et al.'s (2000)  analyses
 24      found that, based on a linear PM effect, PM10.2 5 was significantly associated with total mortality,
 25      but PM2 5 was not.  The PM2 5 in Phoenix is mostly generated from motor vehicles, whereas
 26      PM10_2 5 consists mainly of two types of particles: (a) crustal particles from natural (wind blown
27      dust) and anthropogenic (construction and road dust) processes, and (b) organic particles from
28      natural biogenic processes (endotoxin and molds) and anthropogenic (sewage aeration)
29      processes.  Thus, the associations with PM,0_2 5 are not necessarily indicative of crustal particle
30      effects.
       March 2001                               6-55        DRAFT-DO NOT QUOTE OR CITE

-------
 1           The Castillejos et al. (2000) and Cifuentes et al. (2000) analyses also appear to implicate
 2      PM,0_2 5, as well as PM2 5, as importantly contributing to mortality in two non-U.S. locations,
 3      Mexico City and Santiago, Chile. The latter study also suggests possible seasonal differences in
 4      Santiago, the PM effects in summer being more than double those in winter at that South
 5      American location.
 6
 7      Crustal Particle Effects
 8           Since the 1996 PM AQCD, several studies have yielded interesting new information
 9      concerning possible roles of crustal wind-blown particles or crustal particles within the fine
10      particle fraction (i.e., PM2 5) in contributing to observed PM-mortality effects.
11           Schwartz et al. (1999), for example, investigated the association of coarse particle
12      concentrations with non-accidental deaths in Spokane, Washington, where dust storms elevate
13      coarse particle concentrations. During the 1990-1997 period, 17 dust storm days were identified.
14      The PM|0 levels during those  storms averaged 263 /ug/m3, compared to 39 ,ug/m3 for the entire
15      period. The coarse particle domination of PM,0 data on those dust storm days was confirmed by
16      a separate measurement of PM10 and PM, during a dust storm in August,  1996: the PM,0 level
17      was 187 yUg/m3, while PM, was only 9.5 /ug/m3. The deaths on the day of a dust storm were
18      contrasted with deaths on control days (n=95 days in the main analysis and 171 days in the
19      sensitivity analysis), which are defined as the same day of the year in other years when dust
20      storms  did not occur.  The relative risk for dust  storm exposure was estimated using Poisson
21      regressions, adjusting for temperature, dewpoint, and day of the week.  Various sensitivity
22      analyses considering different seasonal adjustment, year  effects, and lags, were conducted.  The
23      expected relative risk for these storm days with  an increment of 221 //g/m3 would be about 1.04,
24      based on PM,0 relative risk from past studies, but  the estimated RR for high PM10 days was found
25      to be only 1.00 (95% CI=0.95-1.05 for 50 /ug/m3 change) in this study.  Schwartz et al. concluded
26      that there was no evidence to  suggest that coarse (presumably crustal) particles were associated
27      with daily mortality.
28          Pope et al. (1999a) investigated PM10-mortality associations in three metropolitan areas
29      (Ogden, Salt Lake City, and Provo/Orem) in Utah's Wasatch Front mountain region during
30      1985-1995 period. While  the three metropolitan areas shared common weather patterns,
31      pollution levels and patterns among the three areas were  different due to different emission

        March 2001                               6-56       DRAFT-DO NOT QUOTE OR CITE

-------
  1     sources.  The authors ingeniously utilized an index of air stagnation (the clearing index which the
  2     National Weather Service computes from temperature, moisture and wind), to identify and screen
  3     obvious windblown dust days, days clearly identified as  with low stagnation index but high PM10.
  4     They found that Salt Lake City experienced substantially more episodes of wind-blown dust.
  5     They therefore conducted Poisson regression of mortality series using both unscreened and
  6     screened PM10 data. The effects of screening were most apparent in Salt Lake City results.
  7     Before screening no significant relationships were observed. After screening, the RRs per
  8     50 /ug/m3 increase in PM,0 for mortality in the three metropolitan areas were 1.12(1.045 - 1.20),
  9     1.023 (1.00 - 1.047), and 1.019 (0.979 - 1.06) for Ogden, Salt Lake City, and Provo/Orem,
 10     respectively.  These results suggest that the pollution episodes of wind-blown (crustal-derived)
 11     dusts were  less associated with mortality than were the episodes of (presumably) combustion-
 12     related particles.
 13          Ostro et al. (1999a) analyzed the Coachella Valley, CA data for 1989-1992.  This desert
 14     valley, where coarse particles of geologic origin comprise circa 50-60% of annual-average PMIO
 15     (> 90% during wind episodes throughout the year), includes the cities of Palm Springs and Indio,
 16     CA. Total, respiratory, cardiovascular, non-cardiorespiratory and age-over-50 deaths were
 17     analyzed. The correlation between gravimetric and beta-attenuation measurements, separated by
 18     25 miles, was high (r = 0.93); and the beta-attenuation data were used for analysis.  GAM
 19     Poisson models adjusting for temperature, humidity, day-of-week, season,  and time were used.
20     Seasonally stratified analyses were also conducted. Lags 0 through 3 days  (separately) of PM,0,
21      along with moving averages of 3 and 5 days, were evaluated, as were O3, NO2, and CO.
22     Associations were found between 2- or 3-day lagged PM10 and all mortality categories examined,
23      except non-cardiorespiratory.  Effect size estimates for total and cardiovascular deaths were
24      larger for warm season (May through October) than for all year (analogous to Cifuentes et al.
25      (2000) findings for Santiago, Chile). NO2 and CO were statistically significant predictors of
26      mortality in single pollutant models; but in multi-pollutant models, all gaseous pollutants
27      coefficients were reduced and non-significant, whereas PM10 coefficients remained the same and
28      significant.  Ostro et al. (2000) also conducted a follow-up study of the Coachella Valley data for
29      1989-1998,  using actual PM2 5  and PM10.2 5 data for the last 2.5 years but PM2 5 and PMIO_2 s
30      concentrations estimated for the other, earlier years. PM2 5, CO, and NO2 were significantly
31      associated with all-cause mortality, and PM,0 and PM10.2 5 with cardiovascular mortality (but not

        March 2001                               6-57        DRAFT-DO NOT QUOTE OR CITE

-------
 1      PM2 5, possibly due to the low range of concentrations and reduced sample size for PM2 5 data
 2      versus PM10 data).  Thus, although the cardiovascular mortality results hint at crustal particle
 3      effects possibly being important in this desert situation, the ability to discern more clearly the
 4      role of fine particles would likely be improved by analyses of more years of actual data for PM2 5.
 5           Laden et al. (2000) analyzed Harvard Six Cities study data and Mar et al. the Phoenix data
 6      to investigate the role of crustal particles in PM2 5 samples on daily mortality. More detailed
 7      discussion of this study is provided below in Section 6.2.2.4.3 on the source-oriented evaluation
 8      of PM, and only the basic result regarding crustal particles is mentioned here. The elemental
 9      abundance data (from X-ray fluorescence spectroscopy analysis of daily filters) were analyzed to
10      estimate the concentration of crustal particles in PM2 5 using factor analysis.  Then, they
11      estimated the association of mortality with fine crustal  mass using Poisson regression (regressing
12      mortality on factor scores for "crustal factor"), adjusting for time trends and weather. Neither
13      found a positive association between fine crustal mass  factor and mortality.
14           The above results, overall, mostly suggest that crustal particles (coarse or fine) per se are
15      not likely associated with daily mortality. However, as noted in the previous section, three
16      analyses of Phoenix, AZ data suggested that PM10.25 may be associated with mortality.  The
17      results from one of the three studies (Smith et al., 2000) suggest that coarse particle mortality
18      associations are stronger in spring and summer, when the anthropogenic portion of PM,0.2 5 is
19      lowest as determined by factor analysis.  However, during spring and summer, biogenic
20      processes (e.g., wind-blown endotoxins and molds) may contribute more to the PM,0.2 5 fraction
21      in the Phoenix area, clouding any attribution of observed PMi0.2 5 effects there to crustal particles,
22      per se.  Disentangling potential contributions of biogenically-derived organic particle
23      components from those of crustal materials in the PM]0_2 5 fraction in Mexico City and Santiago
24      poses further interesting challenges.
25
26      Ultrafine Particle Effects
27           The Wichmann et al.  (2000) study evaluated the attribution of PM effects to specific size
28      fractions, including both the number concentration (NC) and mass concentration (MC) of
29      particles in a given size range.  The study was carried out in the small German city of Erfurt
30      (pop. 200,000) in the former German Democratic Republic, by a team of scientists at the
31      Gessellschaft fur Strahlenforschung (GSF) and Ludwig Maximilian University in Germany.

        March 2001                               6-58         DRAFT-DO NOT QUOTE OR CITE

-------
  1     Erfurt was heavily polluted by particles and SO2 in the 1980s, and excess mortality was attributed
  2     to high levels of TSP by Spix et al. (1993).  Concentrations of PM and SO2 have markedly
  3     dropped since then.  The present study provides a much more detailed look at the health effects
  4     of ultrafme particles (diameter < 0.1 /urn) than earlier studies, and allows examination of effects
  5     related to number counts for fine and ultrafme particles, as well as to their mass.
  6           The Mobile Aerosol Spectrometer (MAS), developed by GSF, produces number and mass
  7     concentrations in three size classes of ultrafmes (0.01 to 0.1 //m) and three size classes of larger
  8     fine particles (0.1  /^m to 2.5 /urn). The mass concentration MCO.01-2.5 is well correlated with
  9     gravimetric PM2 5, and the number concentration NCO.01-2.5 is well correlated with total particle
 10     counts from a condensation particle counter (CPC). Mortality data were coded by cause of death,
 11      with some discrimination between underlying causes and prevalent conditions of the deceased.
 12     Some analyses looked at cardiovascular causes without respiratory, respiratory without
 13     cardiovascular, and both causes together as separate groups.  Age was used as a modifying factor,
 14     as was weekly data for all of Germany on influenza and similar diseases.  Daily mortality data
 15     were fitted using a Poisson Generalized Additive Model (GAM) with adjustments for weather
 16     variables, time trends, day of week, and particle indices. Two types of models were fitted, one
 17     using the best single-day lag for air pollution, and a second using the best polynomial distributed
 18     lag (PDL) model for air pollution.
 19          Winter PM generally  had the most significant positive effects on mortality, and fall PM
 20     effects were similar in magnitude, but less significant because of the smaller NC and MC in fall
 21      than in winter. Summer PM effects were consistently lower and not significant.  PDL models
 22      generally had larger and more significant PM effects than single-day lag models. Log-
 23      transformed pollution models occasionally provided better fits than untransformed pollutant
 24      models, particularly for number concentration indices in single-day lag models.  However, there
 25      were some nonlinear relationships that could not be adequately described by either parametric
 26      model, as shown by use  of LOESS models. The results cited in Table 6-1 are all for linear PDL
 27      models, to facilitate comparison.
28           Mass concentration was most often significantly associated with excess mortality in one-
29      pollutant models, with excess risks for MAS MCO. 1-2.5 being about 6.2% (CI1.4, 11.2) per
30      25 p:g/m3.  The non-significant estimate from filter PM2 5 was about 3% (CI -1.7, 7.9) per


        March 2001                               6-59        DRAFT-DO NOT QUOTE OR CITE

-------
 1      25 /ug/m3.  Filter PM10 estimates were also significant predictors of mortality overall, about 6.6%
 2      excess risk per 50 ^/g/m3 (CI 0.7 to 12.8) in PDL models.
 3           Mass concentrations for smaller fine particles were also often significant, with excess risk
 4      for MCQ.01-1.0 being ca. 5.1% (CI 0.2, 10.2) per 25 /ug/m3 in a linear PDL model.  Smaller-size
 5      components of MCO.01-1.0 were also significantly associated, or nearly so, with excess
 6      mortality.  The intermodal fraction MCI.0-2.5 was also significant in a PDL logarithmic model,
 7      4.7% (CI 1.05, 8.5) per IQR in log concentration. No results were reported for the effects of
 8      ultrafine mass concentrations in classes 0.01-0.3, 0.03-0.05, 0.05-0.1 Mg/m3.
 9           Number concentrations of ultrafine particles were also associated with excess mortality,
10      significantly or nearly so in smaller size classes.  The results for linear models are shown in
11      Table 6-3.  The table also shows how much the estimated excess risks are reduced,  sometimes
12      drastically, when co-pollutants (especially SO2 and NO2) are included in a two-pollutant model.
13      Number and mass concentrations of various ultrafine and fine particles in all size ranges are
14      rather well correlated with gaseous co-pollutants except for the intermodal size range MCI.0-2.5.
15      The correlations range from 0.44 to 0.62 with SO2, from 0.58 to 0.66 with NO2, and from  0.53 to
16      0.70 with CO. The mass correlations range from 0.53 to 0.62 with SO2, from 0.48 to 0.60 with
17      NO2, and from 0.56 to 0.62 with CO.  The large decreases in excess risk for number
18      concentration, particularly when NO2 is a co-pollutant with NCO.01-0.1, clearly involves a more
19      complex structure than simple correlation. The large decrease in excess risk when SO2 is  a co-
20      pollutant with MCO.01-2.5 is not readily explained, and is discussed in some detail in Wichmann
21      etal. (2000).
22           SO2 is a strong predictor of excess mortality in this study; and its estimated effect is  little
23      changed when different particle indicators are included in a two-pollutant model. The authors
24      noted: "... the [LOESS] smoothed dose response curve showed most of the association at the left
25      end, below 15 jUg/m3, a level at which effects were considered biologically implausible ...".
26      Replacement of sulfur-rich surface coal has reduced mean SO2 levels in Erfurt from 456 /^g/m3 in
27      1988 to 16.8 //g/m3 during 1995 to 1998 and to 6 Mg/m3 in 1998. The estimated concentration-
28      response functions for SO2 are very different in these time periods, comparing Spix et al. (1993)
29      with Wichmann et al. (2000) results.  Wichmann et al. concluded "These inconsistent results for
30      SO2 strongly suggested that SO2 was not the causal agent but an indicator for something else."
31      The authors offered no specific suggestions as to what the "something else" might be, but they

        March 2001                               6-60        DRAFT-DO NOT QUOTE OR CITE

-------
               TABLE 6-3.  EXCESS TOTAL MORTALITY RISKS ESTIMATED TO BE
          ASSOCIATED WITH VARIOUS AMBIENT PARTICLE SIZE-RELATED INDICES
PM Index
NCO.01-0.03
NCO.03-0.05
NCO.05-0.1
NCO.01-2.5
NCO.01-0.1




MCO.01-2.5

Co-Pollutant
None
None
None
None
None
SO2
NO2
CO
MCO.01-2.5
None
S02

Excess Risk, %
3.00°
3.80"
4.00a
6.891"
8.238b
4.758b
0.739b
3.594"
4.123"
6.194C
2.014C
Single-Pollutant Models
Lower 95% CL
-0.342
0.021
-0.307
0.662
0.252
-0.451
-3.951
-2.312
-1.437
1.409
-2.304

Upper 95% CL
6.455
7.722
8.493
13.504
16.860
10.239
5.658
9.856
9.996
11.205
6.523
         "Risks estimates for mortality associated with number concentrations (NC) in specified ranges. At actual
          interquartile range, respectively 8888, 2524, and 1525 particles/cm3.
         bAt standard increment 25,000 particles/cm3; winter IQR is 22,211 particles/cm3, annual IQR is 12,690 particles/cm3.
         cAt standard increment 25 A
-------
 1      cardiovascular or combined respiratory diseases were generally the next highest category. Other
 2      natural causes (i.e., neither respiratory nor cardiovascular) almost always had the lowest risk.
 3
 4      6.2.2.4.2  Chemical Components
 5           Nine new studies from U.S. and Canada examined specific chemical components of PM.
 6      Table 6-3 shows the chemical components examined in these studies, the mean concentrations
 7      for Coefficient of Haze (COH), sulfate, and H+, as well as the list of those that were found to be
 8      associated with increased mortality. There are several chemical components of PM whose
 9      associations with mortality can be compared across studies, including COH, sulfate, and H+.
10
11      Coefficient of Haze, Elemental Carbon, and Organic Carbon
12           COH is highly correlated with elemental carbon (EC) and is often considered as a good PM
13      index for motor vehicle sources (especially diesel), although other combustion processes such as
14      space heating likely also contribute to COH levels. Five studies (Table 6-4) examined COH;
15      and, in most cases, positive and significant associations with mortality outcomes were reported.
16      In terms of relative significance of COH in comparison to other PM components, COH was not
17      the most significant PM component in any of these studies. The average level of COH in these
18      studies ranged from 0.2 (Buffalo, NY) to 0.5 (Santa Clara County, CA) 1000 linear feet. The
19      correlation between COH and NO2 or CO in these studies (8 largest Canadian cities; Santa Clara
20      County, CA; and Buffalo, NY) were moderately high (r ~  0.7 to 0.8), suggesting a likely motor
21      vehicle contribution.  Some of the inconsistencies in the results across cities may be in part due
22      to the differences in COH levels.  For example, in Buffalo, NY (where COH was lowest), no
23      significant association were found for any pollutant, possibly due to small sample size (~ 1 year
24      of data). However, both EC and OC were significant predictors of cardiovascular mortality in
25      the Phoenix study, with their effect sizes per IQR being comparable to those for PM10, PM2 5, and
26      PM,0_2 5; there, EC and OC represented major mass fractions of PM2 5 (11% and 38%,
27      respectively) and correlated highly with PM2 5 (r = 0.84 and 0.89, respectively). They were also
28      highly correlated with CO and NO2 (r = 0.8 to 0.9), indicating their associations with an
29      "automobile" factor.  Thus, the COH and EC/OC results from the Mar et al. (2000) study suggest
30      that PM components  from motor vehicle sources are likely associated with  mortality.
31

        March 2001                               6-62        DRAFT-DO NOT QUOTE OR CITE

-------
  TABLE 6-4.  SUMMARY OF PARTICULATE MATTER CHEMICAL COMPONENTS
                            ANALYZED IN RECENT STUDIES
Author, City
Mean COH
(1000ft)
Mean SO4
G«g/m3)
Mean H+
(nmol/m3)
Other PM
components
analyzed
PM components
associated with
mortality. Comments.
  Burnett etal.(1998b)      0.42
  Toronto, Canada.
  1980-1994.
  Burnett et al. (2000).       0.26
  8 largest Canadian
  cities
  1986-1996.

  Fairley (1999).            0.5
  Santa Clara County,
  CA
  Gwynn et al. (2000).       0.2
  Buffalo, NY
  1988-1990
9.2
         TSP, estimated
         PM10andPM25,
2.6
1.8
5.9
36.4
         PM,0,PM2S,
         PMI0.25, and 47
         trace elements
        PM10,PM25,
        PMI0.25, and
        nitrate
PM,
TSP, COH, sulfate,
estimated PM]0 and
PM25. However, CO
together with TSP
explained most of the
association.

PMIO,PM25,COH,
sulfate, Zn, Ni, and Fe
significantly associated
with total mortality.

COH, sulfate, nitrate,
PM10, and PM25 were
associated with
mortality. PM2 5 and
nitrate most significant.

Sulfate, H+, PM,0, and
COH were associated
with total mortality.
COH was least
significant predictor.
Lipfert et al. (2000a). 0.28
Philadelphia, PA
1992-1995

Lippmann et al.
(2000). Detroit, MI
1992-1994

Klemm and Mason
(2000). Atlanta, GA
1998-1999





5.1 8.0 Nephelometry,
NH4+, TSP, PM10
PM2 5, and
PMI0.25
5.2 8.8 PM10PM25, and
PM,0.25


5.2 0.0 Nitrate, EC, OC,
oxygenated HC,
PM10, PM25, and
PM10.25




Essentially all PM
components were
associated with
mortality.
PMlo,PM25,andPM10.25
were more strongly
associated with mortality
outcomes than sulfate or
"Interim" results based
on one year of data. No
statistically significant
associations for any
pollutants. Those with
t-ratio of at least 1 .0
were: H+, PM]0, and
PM25,
March 2001
       6-63
         DRAFT-DO NOT QUOTE OR CITE

-------
            TABLE 6-4 (cont'd). SUMMARY OF PARTICULATE MATTER CHEMICAL
                        COMPONENTS ANALYZED IN RECENT STUDIES
Author, City
Mar et al. (2000).
Phoenix, AZ
1995-1997








Tsai et al. (2000).
Newark, Elizabeth,
and Camden, NJ
1981-1983




Hoeketal.(2000).
The Netherlands
1986-1994

Other PM
Mean COH Mean SO4" Mean H+ components
(1000ft) Cwg/m3) (nmol/m3) analyzed
S, Zn, Pb, soil-
corrected K,
reconstructed
soil, EC, OC, TC,
PM,0, PM25, and
PM.O-25





12.7 PM]5, PM25,
sulfates
cyclohexane-
solubles (CX),
dichloromethane-
solubles (DCM),
and acetone-
solubles (ACE).
3.8 PM,0, BS, and
(median) nitrate


PM components
associated with
mortality. Comments.
S, Pb, and soil were
negatively associated
with total mortality.
PMIO and PM|0,25 were
positively associated
with total mortality. Soil-
corrected K, non-soil
PM25, EC,OC,TC,
PM,0, PM25,andPM,0.25
were associated with
cardiovascular mortality.
PMI5, PM25, sulfate, CX
and ACE were
significantly associated
with total and/or
cardiovascular mortality
in Newark and/or
Camden.

Sulfate, nitrate, and BS
were more consistently
associated with total
mortality than PM10.
 1     Sulfate and Hydrogen Ion
 2          Sulfate and H+, markers of acidic components of PM, have been hypothesized to be
 3     especially harmful components of PM (Lippmann and Thurston, 1996). The newly available
 4     studies that examined sulfate are shown in Table 6-4; four of them also analyzed H+ data. The
 5     sulfate concentrations ranged from 1.8 /ug/m3 (Santa Clara County, CA) to 12.7 //g/m3 (three NJ
 6     cities). Aside from the west versus east coast contrast, the higher levels observed in Toronto and
 7     the three NJ cities are likely due to their study period coverage of the early 1980's, when sulfate
 8     levels were higher. Sulfate explained 25 to 30% of PM2 5 mass in  eastern U.S. and Canadian
 9     cities, but it was only 14% of PM2 5 mass in Santa Clara County, CA. The mean H+ level in the
10     Buffalo, NY study (36.4 nmol/m3) was much higher than the levels in Philadelphia, Detroit, or
11     Atlanta, in part because the Buffalo study covered the 1988 summer when summer-haze episodes
       March 2001                             6-64        DRAFT-DO NOT QUOTE OR CITE

-------
  1     occurred. The H+ levels measured in the other three cities were low, especially in Atlanta, GA
  2     (where the mean concentration was reported to be 0.0 /ug/m3).  Even the mean H+ concentration
  3     for Detroit, MI (the H+ was actually measured in Windsor, a Canadian city a few miles from
  4     downtown Detroit), 8.8 nmol/m3, was low compared to the reported detection limit of
  5     15.1 nmol/m3 (Brook et al., 1997) for the measurement system used in the study.  Note that the
  6     corresponding detection limit for sulfate was 3.6 nmol/m3 (or 0.34 yUg/m3) and the mean sulfate
  7     level for Detroit was 54 nmol/m3 (or 5.2 yUg/m3), so that the signal-to-noise ratio is expected to be
  8     higher for sulfate than for H+. Thus, the ambient levels and possible relative measurement errors
  9     for these data should be considered in interpreting the results of the studies listed in Table 6-4.
 10          Sulfate was a statistically significant (at p< 0.05) predictor of mortality, at least in  single
 11     pollutant models, in: Toronto, CN; the 8 largest Canadian cities; Santa Clara County, CA;
 12     Buffalo, NY; Philadelphia, PA; Newark, NJ;  and Camden, NJ; but not in Detroit, MI, Elizabeth,
 13     NJ, or Atlanta, GA. However, it should be noted that the relative significance across the cities is
 14     influenced by the sample size (both the daily mean death counts and number of days available),
 15     as well as the range of sulfate levels, and therefore should be interpreted with caution. Figure 6-5
 16     shows the excess risks (± 95% CI) estimated per 5 ,ug/m3  increase in 24-h sulfate reported in
 17     these studies. The largest estimate was seen for Santa Clara County, CA, but the very wide
 18     confidence band (possibly due to the small variance of the sulfate, since its levels were low)
 19     should be taken into account. Also, in  the Santa Clara County analysis, the sulfate effect was
 20     eliminated once PM2 5 was included in the model, perhaps being indicative of sulfate mainly
 21      serving as a surrogate for fine particles in general there. In any case, more weight should be
 22     ascribed to estimates from other studies with narrower confidence bands. In the rest of the
 23      studies, the effect size estimates mostly ranged from about 1 to 4% per 5 ,ug/m3 increase in 24-h
 24      sulfate.
 25           The relative significance of sulfate and H+ compared to other PM components varied from
 26      city to city, as seen in Table 6-4. Because each study included different combinations of
 27      co-pollutants that had different extents  of correlation with sulfate, and because multiple mortality
28      outcomes were analyzed, it is difficult to assess the overall importance of sulfate across the
29      available studies. However, it can generally be seen that the associations were stronger in cities
30      where the sulfate and H+ levels were relatively high. For example, the Gwynn et al., 2000
31      finding for Buffalo, NY data that H+ and sulfate were most significantly associated with total

        March 2001                                6-65        DRAFT-DO NOT QUOTE OR CITE

-------
                               Percent excess death (total mortality, unless otherwise noted)
                                              per 5  pg/m3 increase in sulfate
             Burnett etal. (1998b)
                Toronto, Canada

              Burnett et al. (2000)
                     8 Largest -
                 Canadian Cities
                  Fairley(1999)_
                Santa Clara, Co.

              Gwynn et al (2000) _
                    Buffalo, NY

         Klemm and Mason (2000) _
                    Atlanta, GA

              Lipfert et al. (2000a) _
                Philadelphia, PA

             Lippman et al. (2000) _
                     Detroit, Ml

                Tsai et al. (2000) _
                    3 NJ Cities
                                    -2
0
                    8
10
                         Newwark
                         	 Camden
                                                              Elizabeth
        Figure 6-5.  Excess risks estimated for sulfate per 5 /wg/m3 increase from the studies in
                    which both PM2 5 and PM10_2 5 data were available.
 1      mortality may be in part due to the high acid aerosol levels in that data.  Also, the fact that the

 2      Lippmann et al. (2000) finding for Detroit, MI data on H+ and sulfate being less significantly

 3      associated with mortality than the size-fractionated PM mass indices may be due to acidic

 4      aerosols levels being mostly below the detection limit in that data. In this case, it appears that the

 5      Detroit PM components show mortality effects even without much acidic input.

 6           In summary, assessment of new study results for individual chemical components of PM

 7      suggest that an array of PM components (mainly fine particle constituents) were associated with

 8      mortality outcomes, including: COH, EC, OC, sulfate, H+, and nitrate.  The discrepancies seen

 9      with regard to the relative significance of these PM components across studies may be in part due

10      to the difference in their concentrations. This issue is further discussed below as part of the

11      assessment of new studies involving source-oriented evaluation of PM components.
        March 2001
    6-66
DRAFT-DO NOT QUOTE OR CITE

-------
  1      6.2.2.4.3 Source-Oriented Evaluations
  2           Several new studies have conducted source-oriented evaluation of PM components.
  3      In these studies, daily concentrations of PM components (i.e., trace elements) and gaseous
  4      co-pollutants were analyzed using factor analysis to estimate daily concentrations due to
  5      underlying source types (e.g., motor vehicle emissions, soil, etc.), which are weighted linear
  6      combinations of associated individual variables.  The mortality outcomes were then regressed on
  7      those factors (factor scores) to estimate the impact of source types, rather than just individual
  8      variables. These studies differ in terms of: specific objectives/focus, the size fractions from
  9      which trace elements were extracted, and the way factor analysis was used (e.g., rotation).  The
 10      main findings from these studies regarding the source-types identified (or suggested)  and their
 11      associations with mortality outcomes are summarized in Table 6-5.
 12           The Laden et al. (2000) analysis of Harvard Six Cities data for  1979-1988 aimed to identify
 13      distinct source-related fractions of PM2 5 and to examine each fraction's association with
 14      mortality. Fifteen elements in the fine fraction samples were routinely found above their
 15      detection limits and included in the data analyses.  For each of the six cities, up to 5 common
 16      factors were identified from among the 15 elements, using specific rotation factor analysis.
 17      Using the Procrustes rotation (a type of oblique rotation), the projection of the single tracer for
 18      each factor was maximized. This specification of the tracer element was based on:
 19      (1) knowledge from previous source apportionment research; (2) the  condition that regression of
 20      total fine mass on that element must result in a positive coefficient; and (3) identifications of
 21       additional local source factors that positively contributed to total fine mass regression. Three
 22      source factors were identified in all six cities:  (1) a soil and crustal material factor with Si as a
 23      tracer; (2) a  motor vehicle exhaust factor with  Pb as a tracer; and, (3) a coal combustion factor
 24      with Se as a tracer. City-specific analyses also identified a fuel combustion factor (V), a  salt
 25      factor (Cl), and selected metal factors (Ni, Zn, or Mn). For each city, a GAM Poisson regression
 26      model, adjusting for trend/season, day-of-week, smooth function of temperature and dewpoint,
 27      was used to  estimate impacts of each source type (using absolute factor scores) simultaneously.
 28      Summary estimates across cities were obtained by combining the city-specific estimates,  using
29      inverse variance weights. The identified factors and their tracers are listed in Table 6-5.  The
30      results from  mortality regression analysis including these factors indicated that the strongest
31      increase in daily mortality was associated with the mobile source factor.  Also, the coal

        March 2001                                6-67        DRAFT-DO NOT QUOTE OR CITE

-------
          TABLE 6-5.  SUMMARY OF SOURCE-ORIENTED EVALUATIONS OF
             PARTICULATE MATTER COMPONENTS IN RECENT STUDIES
 Author, City
Source types identified (or suggested) and associated
variables
     Source types associated with mortality.
     Comments.
 Laden et. al., (2000)
 Harvard Six Cities
 1979-1988
 Mar et al. (2000).
 Phoenix, AZ
 1995-1997
Soil and crustal material: Si

Motor vehicle emissions: Pb

Coal combustion: Se

Fuel oil combustion:  V

Salt: Cl

Note: the trace elements are from PM2 5 samples


PM25 (from DFPSS) trace elements:

Motor vehicle emissions and re-suspended road dust:
Mn, Fe, Zn, Pb, OC, EC, CO, and NO2

Soil: Al, Si, and Fe

Vegetative burning: OC, and Ks (soil-corrected
potassium)

Local SO2 sources: SO2

Regional sulfate:  S
     Strongest increase in daily mortality
     associated with mobile source factor.
     Coal combustion factor was positively
     associated with mortality in all
     metropolitan areas, with exception of
     Topeka.  Crustal factor from fine particles
     not associated with mortality. Coal and
     mobile sources account for majority of
     fine particles in each city.
     PM, < factors results:  Soil factor and local
     SO2 factor were negatively associated with
     total mortality.  Regional sulfate was
     positively associated with total mortality
     on the same day, but negatively associated
     on the lag 3 day. Motor vehicle factor,
     vegetative burning factor, and regional
     sulfate factor were significantly positively
     associated with cardiovascular mortality.
                        PMI0.!5 (from dichot) trace elements:

                        Soil:  Al, Si, K, Ca, Mn, Fe, Sr, and Rb

                        A source of coarse fraction metals: Zn, Pb, and Cu

                        A marine influence: Cl
                                                 Factors from dichot PMIO,25 trace elements
                                                 not analyzed for associations with
                                                 mortality because of small sample size
                                                 (every-3rd day samples from June 1996).
 Tsai et al. (2000).
 Newark, Elizabeth,
 and Camden, NJ.
 1981-1983.
Motor vehicle emissions:  Pb, CO

Geological (Soil):  Mn, Fe

Oil burning:  V, Ni

Industrial: Zn, Cu, Cd (separately)

Sulfale/secondary aerosol: sulfate

Note: the trace elements are from PM,5 samples
     Oil burning, industry, secondary aerosol,
     and motor vehicles factors were associated
     with mortality.
 Ozkaynak et al.
 (1996).
 Toronto, Canada.
Motor vehicle emissions:  CO, COH, and NO2
     Motor vehicle factor was a significant
     predictor for total, cancer, cardiovascular,
     respiratory, and pneumonia deaths.	
March 2001
                             6-68
DRAFT-DO NOT QUOTE OR CITE

-------
  1     combustion factor was positively associated with mortality in all metropolitan areas, except for
  2     Topeka.  Lastly, S, Ni, and Pb were specific elements individually associated with mortality, but
  3     the crustal factor from fine particles was not.
  4          Mar et al. (2000) analyzed PM10, PM10_2 5, two measurements of PM2 5, and various
  5     sub-components of PM2 5 for their associations with total (non-accidental) and cardiovascular
  6     deaths in Phoenix, AZ during 1995-1997, using both individual PM components and factor
  7     analysis-derived factor scores. GAM Poisson models were used, adjusting for season,
  8     temperature, and relative humidity.  The evaluated air pollution variables included: O3, SO2,
  9     NO2, CO, TEOM PM10, TEOM PM2 5, TEOM PM10.2 5, DFPSS PM2 5, S, Zn, Pb, soil, soil-
 10     corrected K (Ks), nonsoil PM, OC, EC, and TC. Lags 0 to 4 days were evaluated.  As earlier
 11     noted, individual PM component results indicated that PM10.2 5 was more significantly associated
 12     with total mortality than PM2 5, although both TEOM PM2 5 and PM10.2 5 were significantly
 13     associated with cardiovascular mortality.  A factor analysis conducted on the chemical
 14     components of DFPSS PM2 5 (Al, Si, S, Ca, Fe, Zn, Mn, Pb, Br, Ks,  OC, and EC) identified
 15     factors for:  motor vehicle emissions/re-suspended road dust; soil; vegetative burning; local SO2
 16     sources; and regional sulfate (see Table 6-5). The results of mortality regression with these
 17     factors suggested that the soil factor and local SO2 factor were negatively associated with total
 18     mortality. Regional sulfate was positively associated with total mortality on the same day, but
 19     negatively associated on the lag 3 day. The motor vehicle factor, vegetative burning factor, and
20     regional sulfate factor were each significantly positively associated with  cardiovascular mortality.
21      The authors also analyzed elements from dichot PM10_2 5 samples, and identified soil, a source of
22     coarse fraction metals (industry), and marine influence factors. However, these factors were not
23     analyzed for their associations with mortality outcomes due to the short measurement period
24     (starting in June 1996 with every-3rd-day sampling).
25           It should be noted here that the Smith et al. (2000) analysis of Phoenix data also included
26     factor analysis on the elements from the coarse fraction and identified essentially the same
27      factors ("a source of coarse fraction metals" factor in Mar et al.'s study was called "the
28      anthropogenic elements" in Smith et al.'s study).  While Smith et al. did not relate these factors
29      to mortality (due to a small  sample size), they did show that the anthropogenic elements were
30      low in summer and spring, when the PM10_2 5 effect was largest.  These results suggest that the
31      PMjo-25 effects were not necessarily due to anthropogenic components of the coarse particles,

        March 2001                               6-69        DRAFT-DO NOT QUOTE OR CITE

-------
 1      with biogenically-generated coarse particles perhaps being key during the warmer months (as
 2      noted earlier above).
 3           Tsai et al. (2000) conducted an exploratory analysis of mortality in relation to specific PM
 4      source types for three New Jersey cities (Camden, Newark, and Elizabeth) using factor analysis -
 5      Poisson regression techniques.  During the three-year study period (1981-1983), extensive
 6      chemical speciation data were available, including nine trace elements, sulfate, and particulate
 7      organic matter. Total (excluding accidents and homicides), cardiovascular, and respiratory
 8      mortality were analyzed. Tsai et al. first conducted a factor analysis of trace elements and
 9      sulfate, identifying major source types: motor vehicle (Pb, CO); geological (Mn, Fe); oil burning
10      (V, Ni); industrial (Zn, Cu); and sulfate/secondary aerosols (sulfate). In addition to Poisson
11      regression of mortality on these factors, they also used an alternative approach in which the
12      inhalable particle mass (IPM, D50 < 15 //m) was first regressed on the factor scores of each of the
13      source types to apportion the PM mass; and then the estimated daily PM mass for each source
14      type was included in Poisson regression, so that RR could be calculated per mass concentration
15      basis for each PM source type.  They found that oil burning (V, Ni), various industrial sources
16      (Zn, Cd), motor vehicle (Pb, CO), and the secondary aerosols, as well as the individual PM
17      indices IPM, FPM (D50 < 3.5 /urn), and sulfates, were all associated with total and/or
18      cardiorespiratory mortality in Newark and Camden, but not in Elizabeth.  In Camden, the RRs for
19      the source-oriented PM were higher (~ 1.10) than those for individual PM indices (= 1.02).
20           Ozkaynak et al. (1996) analyzed 21 years of mortality and air pollution data in Toronto,
21      Canada. In addition to the usual simultaneous inclusion of multiple pollutants in mortality
22      regressions, they also conducted a factor analysis of all the air pollution and weather variables,
23      including TSP, SO2,  COM, NO2, O3, CO, relative humidity and temperature. The factor with the
24      largest variance contribution (~ 50%) had the highest factor loadings for CO, COH, and NO2,
25      which they considered to be representative of motor vehicle emissions, since this pollution
26      grouping was also consistent with the emission inventory information for that  city.  They then
27      regressed mortality on the factor scores (a linear combination of standardized scores for the
28      covariates), after filtering out seasonal cycles and adjusting for temperature and day-of-week
29      effects. The estimated impacts on mortality from motor vehicle pollution ranged from 1 to 6%,
30      depending on the outcomes.


        March 2001                                6-70        DRAFT-DO NOT QUOTE OR CITE

-------
  1           In summary, these studies suggest that a number of source-types are associated with
  2     mortality, including motor vehicle emissions, coal combustion, oil burning, and vegetative
  3     burning. The crustal factor from fine particles was not associated with mortality in the Harvard
  4     Six Cities data. In Phoenix data, where coarse particles were reported to be associated with
  5     mortality, the associations between the factors related to coarse particles (soil, marine influence,
  6     and anthropogenic elements) and mortality could not be evaluated due to the small sample size.
  7     However, the soil (i.e., crustal) factor from fine particles in the Phoenix data was negatively
  8     associated with mortality.  Thus, although some unresolved issues remain, mainly due to the lack
  9     of sufficient data, the source-oriented evaluation approach, using factor analysis, thus far seems
 10     to implicate fine particles of anthropogenic origin as being most important (versus crustal
 11     particles of geologic origin) in contributing to observed increased mortality risks.
 12
 13     6.2.2.5  New Assessments of Cause-Specific Mortality
 14           Consistent with similar findings described in the 1996 PM AQCD, most of the newly
 15     available studies summarized in Table 6-1 that examined non-accidental total, circulatory, and
 16     respiratory mortality categories (e.g., Samet et al., 2000a,b; Dominici et al., 2000; Moolgavkar,
 17     2000a; Gwynn et.al., 2000; Lippmann et al., 2000; Ostro et al., 1999a; Schwartz, 2000c) found
 18     significant PM associations with both cardiovascular and/or respiratory-cause mortality. Several
 19     (e.g.,  Ostro et al.,  1998; Fairley, 1999; Gwynn et al., 2000; Borja-Aburto et al., 1997; Wordley
 20     et al., 1997; Borja-Aburto et al., 1998; Prescott et al., 1998;) reported estimated PM effects that
 21      were generally higher for respiratory deaths than for circulatory or total deaths. Once again, the
 22     NMMAPS results for U.S. cities are among those of particular note here due to the large study
 23      size and the combined, pooled estimates derived for various U.S. regions.
 24          The Samet et al. (2000a,b) NMMAPS 90-cities analyses not only examined all-cause
 25      mortality (excluding accidents), but also evaluated cardiovascular, respiratory, and other
 26      remaining causes of deaths.  Results were presented for all-cause, cardio-respiratory, and "other"
27      mortality for lag 0, 1, and 2 days. The investigators commented that, compared to the result for
28      cardio-respiratory deaths showing 3.5% (CI 1.0, 5.9) increase per 50 yUg/m3 PM]0, there was less
29      evidence for non-cardio-respiratory deaths.  However, the estimates for "other" mortality, though
30      half those for cardio-respiratory mortality, were nevertheless positive, with fairly high posterior
31      probability (e.g., 0.84 at lag 0 day) that the overall effects were greater than 0 (estimated percent

        March 2001                                6-71        DRAFT-DO NOT QUOTE OR CITE

-------
  1      excess "other" deaths being -1.3 per 50 /^g/m3 PM10 at lag 0). Dominici et al. (2000) evaluated
  2      the 20 largest U. S. cities, a subset of the cities included in Samet et al.'s NMMAPS analyses.
  3      The pattern of PM10 effects on cardiovascular and respiratory mortality was similar to that
  4      discussed earlier for total mortality, with lag day 1  showing the largest estimates. In this case,
  5      the PM10 effect in these analyses was smaller and weaker for "other" causes. Regional model
  6      results suggested that PM10 effects in the western U.S. were larger than in the eastern or southern
  7      U.S. The PM coefficients were little affected by including gaseous pollutants in the model.
  8           The Lippmann et al. (2000) analyses of cause-specific mortality in Detroit also evaluated
  9      such mortality at various lags (0-3 days) in relation to several PM indices (PM]0, PM2 5, PM10.2 5,
10      sulfate, H+) and various gaseous pollutants (O3, SO2, NO2 and CO), with appropriate adjustment
11      for season, temperature, relative humidity, etc.  Significant effects for both cardiovascular and
12      respiratory mortality were more consistently found for the first three PM indices than for H+ or
13      sulfate. Effect size estimates tended to be highest for lag 1 day. It is notable here that, in the
14      Lippmann et al. (2000) analysis of Detroit mortality data, the "other" mortality category, also
15      showed statistically significant effect size estimates. The authors noted, however, that  the
16      "other" (non-circulatory and non-respiratory) mortality showed seasonal cycles and apparent
17      influenza peaks, suggesting that this series may have also been influenced by respiratory
18      contributing causes.
19           Another U.S. study, that of Moolgavkar (2000a), evaluated possible PM effects on cause-
20      specific mortality across a broad range of lag (0-5 days) times.  Moolgavkar reported that in
21      Poisson regression GAM analyses, controlling for temperature and relative humidity, varying
22      patterns of results were obtained for PM indices in evaluations of daily deaths related to
23      cardiovascular disease (CVD), cerebrovascular disease (CrD), and chronic obstructive lung
24      disease (COPD) in three large U.S. metropolitan areas. In Cook County (Chicago area), the
25      association of PM,0 with CVD mortality was statistically significant at a lag of 3 days based on a
26      single-pollutant analysis and remained significantly associated  with CVD deaths with a 3-day lag
27      in two pollutant models including one or another of CO, NO2, SO2, or O3. In joint analyses with
28      both O3 and SO2, however, the PM10 association became markedly reduced and non-significant.
29      Also, in Los Angeles single-pollutant analyses, PM10 and PM2 5 were significantly associated with
30      CVD mortality with lags of 2 and 1 days, respectively; but their coefficients were not robust to
31      inclusion of one or more gaseous pollutants. In Maricopa Co., AZ, PM,,, coefficients were large
        March 2001                                6-72        DRAFT-DO NOT QUOTE OR CITE

-------
  1      for several lags and significantly associated with CVD mortality lagged 1 day, as were each of
  2      the gaseous pollutants tested (except O3) at several different lag times; and PM10 coefficients
  3      seemed to be robust in 2-pollutant models including PMIO and NO2. As for cerebrovascular
  4      disease, Moolgavkar (2000a) reported that there was little evidence of association for PM with
  5      CrD deaths at any lag in any of the three counties analyzed.  With regard to COPD deaths, PM10
  6      was significantly associated with COPD mortality (lag 2 days) in Cook County.
  7           As for European findings of particular interest, Zmirou et al. (1998) presented cause-
  8      specific mortality analyses results for 10 of the 12 APHEA European cities. Using Poisson
  9      autoregressive models adjusting for trend, season, influenza  epidemics, and weather, each
 10      pollutant's relative risk was estimated in each city, and "meta-analyses" of city specific estimates
 11      was conducted.  The pooled excess risk estimates for cardiovascular mortality were 1.0% (0.3,
 12      1.7) per 25 ,ug/m3 increase in BS and 2.0% (0.5, 3.0) per 50 /ug/m3 increase in SO2 in western
 13      European cities. The pooled risk estimates for respiratory mortality in the same  cities were:
 14      2.0% (0.8, 3.2) and 2.5% (1.5, 3.4) for BS and SO2, respectively. However, significant
 15      associations were not found for the central/eastern European cities. Again, the investigators
 16      noted: (a) potential explanations for differences between the western and central/eastern
 17      European cities (smaller fraction of elderly population and likely larger measurement error
 18      related to exposure representativeness in the central/eastern European cities), and (b) lack of
 19      consistency in NO2- mortality associations. Also of note, Wichmann et al. (2000) found
 20      significant associations of elevated cardiovascular and respiratory disease mortality with various
 21      fine (and ultrafine) particle indices evaluated in Erfurt, Germany. "Other" natural causes (neither
 22      cardio- or respiratory-related) almost always had the lowest risk in those models evaluating
 23      cause-specific mortality.
 24           Seeking unique cause-specificity of effects associated with various pollutants has been
 25      difficult because the "cause specific" categories examined are typically rather broad (usually
 26      cardiovascular and respiratory) and overlap; also cardiovascular and respiratory conditions tend
 27      to cooccur. Examinations of more specific cardiovascular and respiratory sub-categories may be
28      necessary to test hypotheses about any specific mechanisms,  but smaller sample sizes for more
29      specific sub-categories may make a meaningful analysis difficult. The study by Rossi et al.
30      (1999), however, examined associations between TSP and detailed cardio-vascular and
31      respiratory cause-specific mortality in Milan,  Italy for a 9-year period (1980-1989). They found

        March 2001                                6-73        DRAFT-DO NOT QUOTE OR CITE

-------
  1      significant associations for respiratory infections (11% increase per 100 (Ug/m3 increase in TSP;
  2      95%CI: 5, 17) and for heart failure (7%; 95%CI:  3,11), both on the same day TSP. The
  3      associations with myocardial infarction (10%; 95%CI: 3, 18) and COPD (12%; 95%CI: 6, 17)
  4      were found for the average of 3 and 4 day TSP levels.  They noted the difference in lags between
  5      temperature effects (i.e., cold temp, at lag 1 day for respiratory infections; hot temp, at lag 0 for
  6      heart failure and myocardial infarction) and air pollution (TSP) effects. The immediate hot
  7      temperature effects and the lagged cold temperature effects for total and cardiovascular mortality
  8      have been reported in many of the past studies (e.g., Philadelphia, Chicago), but investigations of
  9      the differences in lags of PM effects for specific cardiovascular or respiratory categories have
10      rarely been conducted in time-series mortality studies.
11           A very recently published HEI report on an epidemiologic study conducted by Goldberg
12      et al. (2000) in Montreal, Canada also provides interesting new information regarding types of
13      medical conditions putting susceptible individuals at increased risk for PM-associated mortality
14      effects, and it highlights the potential importance of evaluating "contributing causes" in cause-
15      specific mortality analyses. First, the immediate causes of death, as listed on death certificates,
16      were evaluated in relation to various ambient PM indices (TSP, PM10, PM2 5, COH, sulfates,
17      extinction coefficients) lagged for 0 to 4 days, with results reported emphasizing effects at 3 day
18      lags for three main PM measures (COH, sulfate, estimated PM2 5).  Significant associations were
19      observed between all three measures and total nonaccidental deaths, respiratory diseases, and
20      diabetes, with an approximate 2% increase in excess nonaccidental mortality being observed per
21      9.5 ,ug/m3 interquartile increase in 3-day mean estimated PM2 5 exposure.
22           When underlying clinical conditions identified in decedents' medical records were then
23      evaluated in relation to ambient PM measures, all three measures (COH, sulfate, estimated PM2 5)
24      were associated with acute lower respiratory disease, congestive heart failure, and any
25      cardiovascular disease. Estimated PM2 5 and COH were also reported to be associated with
26      chronic coronary artery disease, any coronary artery disease, and cancer; whereas, sulfate was
27      associated with acute and chronic upper respiratory disease.  None of the three PM measures
28      were related to airways disease, acute coronary artery disease, or hypertension. These results
29      both tend to support previous findings identifying individuals with preexisting cardiopulmonary
30      diseases as being at increased risk for ambient PM effects and appear to implicate another risk


        March 2001                                6-74        DRAFT-DO NOT QUOTE OR CITE

-------
  1      factor, diabetes (which typically also involves cardiovascular complications as it progresses), as a
  2      possible susceptibility condition putting individuals at increased risk for ambient PM effects.
  3           Overall, then, the above assessment of newly available information provides interesting
  4      additional new information (beyond that in the 1996 PM AQCD) with regard to cause-specific
  5      mortality related to ambient PM. That is, a growing number of studies continue to report
  6      increased cardiovascular- and respiratory-related mortality risks as being significantly associated
  7      with ambient PM measures at one or another varying lag times. Largest effects estimates are
  8      most usually reported for 0-1 day lags (with some studies also now noting a second peak at
  9      3-4 day lags).  A few of the newer studies also report associations of PM metrics with "other"
 10      (i.e., non-cardiorespiratory) causes, as well.  However, at least some of these "other" associations
 11      may also be due to seasonal cycles that include relationships to peaks in influenza epidemics that
 12      may imply respiratory complications as a "contributing" cause to the "other" deaths.  Or, the
 13      "other" category may include sufficient numbers of deaths due to diabetes or other diseases
 14      which may also involve cardiovascular complications as contributing causes. Varying degrees of
 15      robustness of PM effects are seen in the newer studies, as typified by estimates in multiple
 16      pollutant models containing gaseous co-pollutants; many show little effect of gaseous pollutant
 17      inclusion on estimated PM effect sizes, some show larger reductions in PM effects to non-
 18      significant levels upon such inclusion, and a growing number also report significant associations
 19     of cardiovascular and respiratory effects with one or more gaseous co-pollutants. Thus, the
 20     newer studies both further substantiate PM effects on cardiovascular- and respiratory-related
 21      mortality, while also pointing toward possible significant contributions of gaseous pollutants to
 22      such cause-specific mortality, as well. The magnitudes of the PM effect size estimates are
 23      consistent with the range of estimates derived from the few earlier available  studies assessed in
 24      the 1996PM AQCD.
 25
26      6.2.2.6  Salient Points Derived from Summarization of Studies of Short-Term Particulate
27              Matter Exposure Effects on Mortality
28           The most salient key points to be extracted from the above discussion of newly available
29      information on short-term PM exposures mortality relationships can be summarized as follow:
30
31

        March 2001                                6-75         DRAFT-DO NOT QUOTE OR CITE

-------
  1      • PMJg effects estimates.  Since the 1996 PM AQCD, thus far, there have been more than 70 new
  2       time-series PM-mortality analyses published. Estimated mortality relative risks in these studies
  3       are generally positive, statistically significant, and consistent with the previously reported PM-
  4       mortality associations.  Of particular importance are several studies which evaluated multiple
  5       cities using consistent data analytical approaches. The NMMAPS analyses for the largest 90
  6       U.S. cities ([Samet et al., 2000a,b] thought to probably provide the most precise estimates for
  7       PM10 effects applicable to the U.S.) derived a combined  nationwide excess risk estimate of
  8       about 2.3% increase in total (non-accidental) mortality per 50 fj.g/m3 increase in PM10. The
  9       other multi-city analyses, as well as various single city analyses, also obtained PM,0 effect sizes
10       generally in the range of 1.5 to 8.5% per 50 /ug/m3 increase in PM,0, consistent with the broader
11       range of statistically significant estimates given in the 1996 PM AQCD. However, more
12       geographic heterogeneity is evident among the newer multi-city study results than was the case
13       among the fewer studies assessed in the 1996 PM AQCD.  In particular, in the NMMAPS
14       analysis of the 90 largest U.S. cities data, the risk estimates varied somewhat by U.S.
15       geographic region, with the estimate for the Northeast being the largest (4.6% per 50 ^g/m3
16       PMIO increase).  The  observed heterogeneity in the estimated PM risks across cities/regions
17       could not be explained with the city-specific explanatory variables, such as the mean levels of
18       pollution and weather, mortality rate, sociodemographic variables (e.g., median household
19       income), urbanization, or variables related to measurement error. Notable apparent
20       heterogeneity was also seen among effects estimates for  PM (and SO2) indices in the multi-city
21       APHEA study conducted in European cities.  The issue of heterogeneity of effects estimates is
22       discussed further below in Section 6.4.
23
24      • Confounding and effect modification by other pollutants. Numerous new short-term PM
25       exposure studies not only continue to report significant associations between various PM
26       indices and mortality, but also between gaseous pollutants (O3, SO2, NO2,  and CO) and
27       mortality as well. In  most of these studies, simultaneous inclusions of gaseous pollutants in the
28       regression models did not meaningfully affect the PM effect size estimates.  This was the case
29       for the NMMAPS 90 cities study with regard to the overall combined U.S. regional and
30       nationwide risk estimates derived for that study.  On the  other hand, some other non-multicity
31       studies found significant PM-mortality associations at one or more lag times to be reduced by

        March 2001                               6-76        DRAFT-DO NOT QUOTE OR CITE

-------
  1       inclusion of one or another gaseous pollutants (e.g., CO, NO2, SO2, but not typically O3).  Thus,
  2       there appears to be independent effects of short-term PM and gaseous pollutant exposures on
  3       mortality and/or significant confounding between PM and gaseous pollutants derived from
  4       common sources.  There is not sufficient evidence to clearly establish modifications of PM
  5       effects by other gaseous co-pollutants. The issue of confounding is discussed further in
  6       Section 6.4.
  7
  8     • Fine and coarse particle effects. Newly available studies provide generally statistically
  9       significant PM2 5 associations with mortality, with effect size estimates falling in the range
 10       reported in the 1996 PM AQCD. New results from Germany appear to implicate both ultrafme
 11       (nuclei-mode) and accumulation-mode fractions of urban ambient fine PM as being important
 12       contributors to increased mortality risks. As to the relative importance of fine and coarse
 13       particles, in the 1996 PM AQCD there was only one acute mortality study in which examined
 14       this issue. In that study, the authors suggested that fine particles (PM2 5), but not coarse
 15       particles (PM10_2 5), were associated with daily mortality. Now, more than ten studies have
 16       analyzed both PM2 5 and PM10.2 5 for their associations with mortality. While the results from
 17       some of these new studies (e.g., Santa Clara County, CA analysis [Fairley, 1999] and the
 18       largest 8 Canadian cities analysis [Burnett et al., 2000]) did suggest that PM25 was more
 19       important than PM10.2 5 in predicting mortality fluctuations, other studies (e.g., Phoenix, AZ
 20       analyses [Clyde et al., 2000; Mar et al., 2000; Smith et al., 2000]; Mexico City and Santiago,
 21        Chile studies [Castillejos et al., 2000; Cifuentes et al., 2000]) suggest that PMI0.2 5 may also be
 22       important in at least some locations. Seasonal dependence of size-related PM component
 23        effects observed in some of the studies complicates interpretations.  At least some of the
 24        reported coarse (PMI0.2 5) fraction particle effects seen most clearly during warmer seasons may
 25        be hypothesized to be due to biogenically-derived particles (molds, endotoxins, etc.) that tend
26        to be elevated during such seasons.
27
28      •  Chemical components ofPM. Several new studies have examined the role of specific
29        chemical components of PM.  The studies conducted in U.S. and Canadian cities showed
30        mortality associations with specific fine particle components of PM including H+, sulfate,
31        nitrate, as well as COH, but their relative importance varied from city to city, likely depending

        March 2001                               6-77        DRAFT-DO NOT QUOTE OR CITE

-------
 1       on their levels (e.g., no clear associations in those cities where H+ and sulfate levels were very
 2       low, i.e., circa non-detection limits).  The results of several studies that investigated the role of
 3       crustal particles, although somewhat mixed, do not appear overall to support associations
 4       between crustal particles and mortality (see also source-oriented evaluations below).
 5
 6      • Source-oriented evaluations.  Several studies conducted source-oriented evaluations of PM
 7       components using factor analysis. The results from these studies generally indicate that
 8       several combustion-related source-types are likely associated with mortality, including: motor
 9       vehicle emissions; coal combustion; oil burning; and vegetative burning. The crustal factor
10       from fine particles was not associated with mortality in the Harvard Six Cities data, and the soil
11       (i.e., crustal) factor from fine particles in the Phoenix data was negatively associated with
12       mortality. Thus, the source-oriented evaluations seem  to implicate fine particles of
13       anthropogenic origin as being most important as contributing to increased mortality and are
14       thus far generally non-supportive of increased mortality risks being related to short-term
15       exposures to crustal materials in U.S. ambient environments examined to date.
16
17      • Cause-specific mortality. Findings for new results concerning cause-specific mortality
18       comport well with those for total (non-accidental) mortality, the former showing generally
19       larger effect size estimates for cardiovascular, respiratory, and/or combined cardiorespiratory
20       excess risks than for total mortality risks.
21
22      • Lags.  In general, maximum effect sizes  for total mortality appear to be obtained with 0-1 day
23       lags, with some studies finding a second peak for 3-4 days lags. There is also some evidence
24       that, if effects distributed over multiple lag days are considered, the effect size may be  larger
25       than for any single maximum effect size lag day.
26
27      • Threshold.  Few new short-term mortality studies explicitly address the issue of thresholds.
28       One study that analyzed Phoenix, AZ data did report some limited evidence suggestive of a
29       possible  threshold for PM2 5 there. However, several different analyses of larger PM10  data sets
30       across multiple cities generally provide little or no support to indicate a threshold for PM10
31       mortality effects. Threshold issues are discussed further in Section 6.4.

        March 2001                               6-78        DRAFT-DO NOT QUOTE OR CITE

-------
  1      6.2.3  Mortality Effects of Long-Term Exposure to Ambient Particulate
  2             Matter
  3      6.2.3.1  Studies Published Prior to the 1996 Particulate Matter Criteria Document
  4      6.2.3.1.1 Aggregate Population Cross-Sectional Chronic Exposure Studies
  5           Mortality effects associated with chronic, long-term exposure to ambient PM have been
  6      assessed in cross-sectional studies and, more recently, in prospective cohort studies.  A number
  7      of older cross-sectional studies from the  1970s provided indications of increased mortality
  8      associated with chronic (annual average) exposures to ambient PM, especially with respect to
  9      fine mass or sulfate (SO4=) concentrations. However, questions unresolved at that time regarding
 10      the adequacy of statistical adjustments for other potentially important covariates (e.g., cigarette
 11      smoking, economic status, etc.) across cities tended to limit the degree of confidence that was
 12      placed by the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996) on  such purely
 13      "ecological" studies or on quantitative estimates of PM effects derived from these  studies.
 14      Evidence comparing the toxicities of specific PM components was relatively limited. The sulfate
 15      and acid components had already been discussed in detail in the previous PM AQCD (U.S.
 16      Environmental Protection Agency, 1986).
 17
 18      6.2.3.1.2 Semi-Individual (Prospective  Cohort) Chronic Exposure Studies
 19           Semi-individual cohort studies using subject-specific information about relevant covariates
 20      (such as cigarette smoking, occupation, etc.) have provided more certain findings of  long-term
 21      PM exposure effects than past purely "ecological studies" (Kiinzli  and Tager, 1997).  At the same
 22      time, these better designed cohort studies have largely confirmed the magnitude of PM effect
 23      estimates from past cross-sectional study results.
 24           Prospective cohort semi-individual  studies of mortality associated with chronic exposures
 25      to air pollution of outdoor origins have yielded especially valuable insights into the adverse
 26      health effects of long-term PM exposures. The extensive Harvard  Six-Cities Study (Dockery
 27      et al., 1993) and the American Cancer Society (ACS) Study (Pope  et al., 1995) agreed in their
28      findings of statistically significant positive associations between fine particles and  excess
29      mortality, although the ACS study did not evaluate the possible contributions of other air
30      pollutants. Neither study considered multi-pollutant models, although the Six-City study did
31      examine various gaseous and particulate matter indices (including total particles, PM2 5, SO4=, H+,

        March 2001                               6-79        DRAFT-DO NOT QUOTE OR CITE

-------
  1      SO2, and ozone), finding that sulfate and PM2 5 fine particles were best associated with mortality.
  2      The RR estimates for total mortality in the Six-Cities study (and 95 percent confidence intervals,
  3      CI) per increments in PM indicator levels were: RR=1.42 (CI=1.16-2.01) for 50 /^g/m3 PM10;
  4      RR=1.31 (CI=1.11 -1.68) for 25 ^g/m3 PM2 5; and RR=1.46 (CI=1.16-2.16) for 15 ^g/m3 SO4=.
  5      The estimates for total mortality derived from the ACS study were RR=1.17 (CI=1.09-1.26) for
  6      25 //g/m3 PM25 and 1.11 (CI=1.06-1.16) for 15 Aig/m3 SO4=. The ACS pollutant RR estimates are
  7      smaller than those from the Six-Cities study, although their 95% confidence intervals overlap.
  8      In some cases in these studies, the life-long cumulative exposure of the study cohorts included
  9      distinctly higher past PM exposures, especially in cities with historically higher PM levels (e.g.,
10      Steubenville, OH); but more current PM measurements were used to estimate the chronic PM
11      exposures.  In the ACS study, the pollutant exposure estimates were based on concentrations at
12      the start of the study (during 1979-1983). Also, the average age of the ACS cohort was 56,
13      which could overestimate the pollutant RR estimates and perhaps underestimate the life-
14      shortening associated with PM associated mortality. Still, although caution must be exercised
15      regarding the use of the reported quantitative risk estimates, the Six-Cities and ACS semi-
16      individual  studies provided consistent evidence of a significant mortality association with long-
17      term exposure to PM of ambient origins.
18           In contrast to the Six-Cities and ACS studies, early results from the Adventist Health Study
19      on Smog (AHSMOG) of California nonsmokers by Abbey et al. (1991) and Abbey et al. (1995a)
20      found no significant mortality effects of previous PM exposure in a relatively young cohort.
21      However, these analyses used TSP as the PM exposure metric, rather than more health relevant
22      PM metrics such as PM10 or PM2 5, included fewer subjects than the ACS study, and considered a
23      shorter follow-up time than the Six-Cities study (ten years vs.  15 years for the Six-Cities study).
24      Moreover,  the AHSMOG study included only non-smokers, indicated by the Six-Cities Study as
25      having lower pollutant RR's than smokers, suggesting that a longer follow-up time than
26      considered in the past (10 years) might be required to have sufficient power to detect significant
27      pollution effects than is required in studies that include smokers (such as the Six-Cities and ACS
28      studies). Thus, greater emphasis has been placed thus far on the Six-Cities and ACS studies.
29           Overall, the past chronic PM exposure studies collectively indicated that increases in
30      mortality are associated with long-term exposure to ambient airborne particles. Also, effect  size
31      estimates for total mortality associated with chronic PM exposure indices are much larger than

        March 2001                               6-80        DRAFT-DO NOT QUOTE OR CITE

-------
  1     those reported from daily mortality PM studies. This suggests that a major fraction of the
  2     reported mortality relative risk estimates associated with chronic PM exposure likely reflects
  3     cumulative PM impacts above and beyond those exerted by the sum of acute exposure events
  4     (i.e., assuming that the latter are fully additive over time). The 1996 PM AQCD (Chapter 12)
  5     reached several conclusions concerning four key questions about the prospective cohort studies,
  6     as directly quoted below:
  7
  8           (1) Have potentially important confounding variables been omitted?
  9                 "While it is not likely that the prospective cohort studies have overlooked plausible
 10           confounding factors that can account for the large effects attributed to air pollution, there
 11           may be some further adjustments in the estimated magnitude of these effects as individual
 12           and community risk factors are included in the analyses." These include individual
 13           variables such as education, occupational exposure to dust and fumes, and physical activity,
 14           as well as ecological (community) variables such as regional location, migration, and
 15           income distribution.  Further refinement of the effects of smoking status may also prove
 16           useful."
 17
 18           (2) Can the most important pollutant species be identified?
 19                 "The issue of confounding with co-pollutants has not been resolved for the
 20           prospective cohort studies . . . Analytical strategies that could have allowed greater
 21           separation of air pollutant effects have not yet been applied to the prospective cohort
 22           studies." The ability to separate the effects of different pollutants, each measured as a long-
 23          term average on a community basis, was clearly most limited in the Six Cities study.  The
 24          ACS study offered a much larger number of cities, but did not examine differences
 25          attributable to the spatial and temporal differences in the mix of particles and gaseous
 26          pollutants across the cities.  The AHSMOG study constructed time- and location-dependent
 27          pollution metrics for most of its subjects that might have allowed such analyses, but no
 28          results were reported.
29
30


        March 2001                                6-81         DRAFT-DO NOT QUOTE OR CITE

-------
  1           (3) Can the time scales for long-term exposure effects be evaluated?
  2                 "Careful review of the published studies indicated a lack of attention to this issue.
  3           Long-term mortality studies have the potential to infer temporal relationships based on
  4           characterization of changes in pollution levels over time.  This potential was greater in the
  5           Six Cities and AHSMOG studies because of the greater length of the historical air pollution
  6           data for the cohort. The chronic exposure studies, taken together, suggest that there may be
  7           increases in mortality in disease categories that are consistent with long-term exposure to
  8           airborne particles, and that at least some fraction of these deaths are likely to occur between
  9           acute exposure episodes.  If this interpretation is correct, then at least some individuals may
10           experience some years of reduction of life as a consequence of PM exposure."
11
12           (4) Is it possible to identify pollutant thresholds that might be helpful in health
13           assessments?
14                 "Model specification searches for thresholds have not been reported for prospective
15           cohort studies. . . . Measurement error in pollution variables also complicates the search
16           for potential threshold effects. . . . The problems that complicate threshold detection in the
17           population-based studies have a somewhat different character for the long-term studies."
18
19      6.2.3.2  Prospective Cohort Analyses of Chronic Particulate Matter Exposure Mortality
20              Effects Published Since the 1996 Particulate Matter Air Quality Criteria Document
21           Considerable progress has been made  towards addressing further the above issues.
22      For example, extensive reanalyses (Krewski et al., 2000) of the ACS  and Six-Cities Study
23      conducted under sponsorship by the Health Effects  Institute (HEI) indicate that the published
24      findings of the  original investigators (Dockery et al., 1993;  Pope et al., 1995) are based on
25      substantially valid data sets and statistical analyses. The HEI reanalysis project has demonstrated
26      that small corrections in input data have very little effect on the findings and that alternative
27      model specifications further substantiate the robustness  of the originally reported findings.
28      In addition,  some of the above key questions have been  further investigated by Krewski et al.
29      (2000) via sensitivity analyses for the Six City and ACS studies data sets, including consideration
30      of a much wider range of confounding variables.  Recently  published analyses of AHSMOG data
31      (Abbey et al., 1999; Beeson et al., 1998) also extend the ASHMOG findings showing some

        March 2001                                6-82         DRAFT-DO NOT QUOTE OR CITE

-------
  1      analytic outcomes different from earlier analyses reported out from the study. Additional new
  2      studies suggestive of possible effects of sub-chronic PM exposures on infant mortality (Woodruff
  3      et al., 1997; Bobak and Leon, 1998), are also discussed below.
  4
  5      6.2.3.2.1  Health Effects Institute Reanalyzes of the Six-Cities and ACS Studies
  6          The overall objective of the HEI "Particle Epidemiology Reanalysis Project" was to
  7      conduct a rigorous and independent assessment of the findings of the Six Cities (Dockery et al.,
  8      1993) and ACS (Pope et al., 1995) Studies of air pollution and mortality. The following
  9      description of approach, key results, and conclusions is largely extracted from the Executive
 10      Summary of the HEI final report (Krewski et al., 2000). The HEI-sponsored reanalysis effort
 11       was approached in two steps:
 12          • Part I:  Replication and Validation. The Reanalysis Team sought to test:  (a) if the
 13            original studies could be replicated via a quality assurance audit of a sample of the
 14            original data and; (b) if the original numeric results could be validated.
 15          • Part II:  Sensitivity Analyses. The Reanalysis Team tested the robustness of the original
 16            analyses to alternate risk models and analytic approaches.
 17          The Part I audit of the study population data for both the Six Cities and ACS Studies and of
 18      the air quality data in the Six Cities Study revealed the  data to be of generally high quality with
 19     few exceptions. In both studies, a few errors were found in the data coding for and exclusion of
 20      certain subjects; when those subjects were included in the analyses, they did not materially
 21      change the results from those originally reported. Because the air quality data used in the ACS
 22      Study could not be audited, a separate air quality database was constructed for the sensitivity
 23      analyses in Part II.
 24           The Reanalysis Team was able to replicate the original results for both studies using the
 25      same data and statistical methods as used by the Original Investigators.  The Reanalysis Team
 26      confirmed the original point estimates, as shown in Table 6-6. For  the Six Cities  Study, they
 27      reported the relative risk of mortality from all causes associated with an increase in fine particles
 28      of 20.0 Aig/m3 as 1.28, the same as the 1.28 per 20 /^g/m3 reported by the Original Investigators.
29      For the ACS Study, the relative risk of all-cause mortality associated with a 20 /ug/m3 increase in
 30      fine particles was  1.19 in the reanalysis, very close to the original 1.14 value.
31

        March 2001                                6-83        DRAFT-DO NOT QUOTE OR CITE

-------
               TABLE 6-6.  COMPARISON OF SIX CITIES AND AMERICAN CANCER
              SOCIETY STUDY FINDINGS FROM ORIGINAL INVESTIGATORS
                         AND HEALTH EFFECTS INSTITUTE REANALYSIS
Type of Health
Effect & Location Indicator
Original Investigators' Findings
Six City? PM25
Six City/1 PMI5/IO
ACS Stud/ PM2S
HEI reanalysis Phase I: Replication
Six City Reanalysisd PM2 5
PMI5
A CS Study Reanalysisd PM2 s
PM15 (dichot)
PM,S (SSI)
Mortality Risk
Total mortality
Relative Risk (95% CI)
1.28(1.09-1.51)
1.18(1.06-1.32)
1.14(1.07-1.21)
Relative Risk (95% CI)
1.28 (1.09-1.51)
1.19(1.06-1.34)
1.19(1.08-1.21)
1.04 (1.01-1.07)
1.02(0.99-1.04)
per Increment in PMa
Cardiopulmonary mortality
Relative Risk (95% CI)
1.40(1.12-2.01)
e
1.25 (1.14-1.36)

1.41 (1.13-1.76)
1.20 (1.02-1.41)
1.33 (1.19-1.47)
1.07 (1.02-1.12)
1.06(1.03-1.09)
        " Estimates calculated on the basis of differences between the most-polluted and least-polluted cities, using
         increments of 20 ,ug/m3 increase for PM,0, PMI5 and PM2 s
        "Dockeryetal. (1993)
        LPope et al. (1995)
        ''Krewski et al. (2000)
        'Data presented only by smoking subgroup
 1           The Part II sensitivity analysis applied an array of different models and variables to
 2     determine whether the original results would remain robust to different analytic assumptions and
 3     model specifications. The Reanalysis Team first applied the standard Cox model used by the
 4     Original Investigators and included variables in the model for which data were available from
 5     both original studies, but had not been used in the published analyses (e.g. physical activity, lung
 6     function, marital status).  The Reanalysis Team also designed models to include interactions
 7     between variables. None of these alternative models produced results that materially altered the
 8     original findings.
 9           Next, for both the Six Cities and ACS Studies, the Reanalysis Team investigated the
10     possible effects of fine particles and sulfate on a range of potentially susceptible subgroups of the
11     population.  These analyses did not find differences in PM-mortality associations among
12     subgroups based on various personal characteristics (e.g., including gender, smoking status,
13     exposure to occupational dusts and fumes, and marital status). However, estimated effects of
14     fine particles did vary with educational level; the association between an increase in fine particles
       March 2001                               6-84        DRAFT-DO NOT QUOTE OR CITE

-------
  1     and mortality tended to be higher for individuals without a high school education than for those
  2     with more education. The Reanalysis Team postulated that this finding could be attributable to
  3     some unidentified socioeconomic effect modifier.  The authors concluded, "The Reanalysis
  4     Team found little evidence that questionnaire variables had led to confounding in either study,
  5     thereby strengthening the conclusion that the observed association between fine particle air
  6     pollution and mortality was not the result of a critical covariate that had been neglected by the
  7     Original Investigators." (Krewski et al., 2000, pp. 219-220).
  8           In the ACS study, the Reanalysis Team tested whether the relationship between ambient
  9     concentrations and mortality was linear. They found some indications of both linear and
 10     nonlinear relationships, depending upon the analytic technique used, suggesting that the shapes
 11     of the concentration-response relationships warrant additional research in the future.
 12           One of the criticisms of both original studies has been that neither analyzed the effects of
 13     change in pollutant levels over time. In the Six Cities Study, for which such data were available,
 14     the  Reanalysis Team tested whether effect estimates changed when certain key risk factors
 15     (smoking, body mass index, and air pollution) were allowed to vary over time. In general, the
 16     reanalysis results did not change when smoking and body mass index were allowed to vary over
 17     time.  The Reanalysis Team did find for the Six Cities Study, however, that when the general
 18     decline in fine particle levels over the monitoring period was included as a time-dependent
 19     variable, the association between fine particles and all-cause mortality was reduced (RR = 1.22,
 20     CI 1.03, 1.45). This would be expected, since the most polluted cities would be expected to have
 21     the  greatest decline as pollution controls were applied.  Despite this adjustment, the PM2 5 effect
 22     estimate continued to be positive and statistically significant.
 23          To test the validity of the original ACS air quality data, the Reanalysis Team constructed
 24     and applied its own air quality dataset from available historical data.  In particular, sulfate levels
 25     with and without adjustment were found to differ by about 10% for the Six Cities Study. Both the
 26     original ACS Study air quality data and the newly constructed dataset contained sulfate levels
 27     inflated by approximately 50% due to artifactual sulfate. For the Six Cities Study, the relative
 28     risks of mortality were essentially unchanged with adjusted or unadjusted sulfate. For the ACS
29     Study, adjusting for artifactual sulfate resulted in slightly higher relative risks of mortality from
30     all causes and cardiopulmonary disease compared with unadjusted data, while the relative risk of
31      mortality from lung cancer was lower after the data had been adjusted. Thus,  the Reanalysis

        March 2001                               6-85        DRAFT-DO NOT QUOTE OR CITE

-------
  1      Team found essentially the same results as the original Harvard Six-Cities and ACS studies, even
  2      after using independently developed pollution datasets and after adjusting for sulfate artifact.
  3           Because of the limited statistical power to conduct most model specification sensitivity
  4      analyses for the Six Cities Study, the Reanalysis Team conducted the majority of its sensitivity
  5      analyses using only the ACS Study dataset that considered 151 cities.  When a range of city-level
  6      (ecologic) variables (e.g., population change, measures of income, maximum temperature,
  7      number of hospital beds, water hardness) were included in the analyses, the results generally did
  8      not change.  The only exception was that associations with fine particles and sulfate were
  9      reduced when city-level measures of population change or SO2 were included in the model.
10           A major product of the Reanalysis Project is the determination that both pollutant variables
11      and mortality appear to be spatially correlated in the ACS  Study dataset. If not identified and
12      modeled correctly, spatial correlation could cause substantial errors  in both the regression
13      coefficients and their standard errors. The Reanalysis Team identified several methods for
14      addressing this, each of which resulted in some reduction in the estimated regression coefficients.
15      The full implications and interpretations of spatial correlations in these analyses have not been
16      resolved, and were noted to be an important subject for future research.
17           When the Reanalysis Team sought to take into account both the underlying variation from
18      city to city (random effects) and variation from  the spatial  correlation between cities, associations
19      were still found between mortality and sulfates or fine particles. In results of various models
20      using alternative methods to address spatial autocorrelation and including different ecologic
21      covariates, fine particle-mortality associations ranged from 1.11 to 1.29 (RR reported by original
22      investigators was 1.17) for a 24.5 /ug/m3 increase in PM2 5. With the exception of sulfur dioxide,
23      consideration of other pollutants in these models did not alter the associations found with
24      sulfates. The authors reported associations that were stronger for SO2 than for sulfate, which
25      may indicate that the sulfate with artifact was "picking up" some of the SO2 association, perhaps
26      because the artifact is in part proportional to the prevailing SO2 concentration (Coutant, 1977).
27      It should be recognized that the Reanalysis Team did not use data adjusted for artifactual sulfate
28      for most alternative analyses. When they did use adjusted  sulfate data, relative risks of mortality
29      from all causes and cardiopulmonary disease increased.  This result suggests that more analyses
30      with adjusted sulfate might result in somewhat higher relative risks associated with sulfate. The
31      Reanalysis Team concluded: "it suggests that uncontrolled spatial autocorrelation accounts for

        March 2001                                6-86        DRAFT-DO NOT QUOTE OR CITE

-------
  1      24% to 64% of the observed relation. Nonetheless, all our models continued to show an
  2      association between elevated risks of mortality and exposure to airborne sulfate." (Krewski
  3      et al., 2000, p. 230)
  4          In summary, the reanalyses generally confirmed the original investigator's findings of
  5      associations between mortality and long-term exposure to PM, while recognizing that increased
  6      mortality may be attributable to more than one ambient air pollution component.  Regarding the
  7      validity of the published Harvard Six-Cities and ACS Studies, the HEI Reanalysis Report
  8      concluded that: "Overall, the reanalyses assured the quality of the original data, replicated the
  9      original results, and tested those results against alternative risk models and analytic approaches
 10      without substantively altering the original findings of an association between indicators of
 11       particulate matter air pollution and mortality."
 12
 13       6.2.3.2.2 AHSMOG Analyses
 14           The Adventist Health Study of Smog (AHSMOG) represents a third major U.S. prospective
 15       cohort study of chronic PM exposure-mortality effects. The study enrolled 6,338 non-smoking
 16       non-Hispanic white  Seventh  Day Adventist residents of California, ages 27 to 95 years, in 1977.
 17       The participants had resided for at least 10 years within 5 miles (8 km) of their then-current
 18       residence locations,  either within the 3 major California air basins (San Diego, Los Angeles, or
 19       San Francisco) or else were part of a random 10% sample of Adventist Health Study participants
 20       in the rest of California.  The study has been extensively described and initial results reported
 21       elsewhere (Hodgkin et al., 1984; Abbey et al., 1991; Mills et al., 3991). In the latest AHSMOG
 22      analyses (Abbey et al., 1999), mortality status of the subjects after ca. 15-years of follow-up
 23      (1977-1992) was determined by various tracing methods, finding 1,628 deaths (989 female,
 24      639 male) in the cohort. This is a 50% percent increase in the follow-up period vs. previous
 25      AHSMOG reports, which increases the power of the latest analyses over past published ones.
 26      Of 1,575 deaths from all natural (non-external) causes, 1,029 were cardiopulmonary, 135 were
 27      non-malignant respiratory (ICD9 codes 460-529), and 30 were lung cancer (ICD9 code 162)
 28      deaths.  Abbey et al.  (1999) also created another death category, contributing respiratory causes
 29      (CRC).  CRC included any mention of nonmalignant respiratory disease as either an underlying
 30      or a "contributing cause" on the death certificate. Numerous analyses were done for the CRC
31      category, due to the large numbers and relative specificity of respiratory causes as a factor in the

        March 2001                               6-87        DRAFT-DO NOT QUOTE OR CITE

-------
 1     deaths. Education was used to index socio-economic status, rather than income. Physical
 2     activity and occupational exposure to dust were also used as covariates.
 3          Cox proportional hazard models adjusted for a variety of covariates, or stratified by sex,
 4     were used. The "time" variable used in most of the models was survival time from date of
 5     enrollment, except that age on study was used for lung cancer effects due to the expected lack of
 6     short-term effects. A large number of covariate adjustments were evaluated, yielding results for
 7     all non-external mortality as shown in Table 6-7 and described by Abbey et al. (1999).
 8     Essentially no statistically significant PM related effects were observed for either males or
 9     females, except RR =  1.08 for males in relation to PM10 > 100 yUg/m3, d/yr.
10
11
           TABLE 6-7. RELATIVE RISK OF MORTALITY FROM ALL NONEXTERNAL
                CAUSES, BY SEX AND AIR POLLUTANT, FOR AN ALTERNATIVE
                         COVARIATE MODEL IN THE ASHMOG STUDY

Pollution Index
PMIO>100, d/yr.
PM|0 mean
SO4 mean
O3>100ppb, h/yr.
SO2 mean

Pollution Incr.
30 days/yr.
50 ,ug/m3
15 Mg/m3
551 h/yr. (IQR)
3.72 (IQR)

RR
0.958
0.879
0.732
0.90
1.00
Females
LCL
0.899
0.713
0.484
0.80
0.91

UCL
1.021
1.085
1.105
1.02
1.10

RR
1.082
1.242
1.279
1.140
1.05
Males
LCL
1.008
0.955
0.774
0.98
0.94

UCL
1.162
1.616
2.116
1.32
1.18
        LCL = Lower 95% confidence limit        UCL = Upper 95% confidence limit
        Source: Abbey et al. (1999).
12          An analogous pattern of results was found for cause-specific mortality analyses of the
13     AHSMOG data. That is, positive and statistically significant effects on cardiopulmonary deaths
14     were found in models that  included both sexes and adjustment for age, pack-years of smoking,
15     and body-mass index (BMI) (RR =1.14, 95% CI 1.03-1.56 for 30 day/yr > 100 /ug/m3 PM10).
16     Subsets of the cohort had elevated risks: (a) former smokers had higher RR's than never-
17     smokers (RR for PM10 exceedances for never-smokers was marginally significant by itself);
18     (b) subjects with low intake of anti-oxidant vitamins A, C, E had significantly elevated risk of
       March 2001                              6-88        DRAFT-DO NOT QUOTE OR CITE

-------
  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
response to PM10, whereas those with adequate intake did not (suggesting that dietary factors or,
possibly, other socio-economic or life style factors for which they are a surrogate may be
important covariates); and (c) there also appeared to be a gradient of PM10 risk with respect to
time spent outdoors, with those who had spent at least 16 h/wk outside at greater risk from PM10
exceedances.  The extent to which time spent outdoors is a surrogate for other variables or is a
modifying factor reflecting temporal variation in exposure to ambient air pollution is not certain.
For example, if the males spent much more time outdoors than females, outdoor exposure time
could be confounded with gender. When the cardiopulmonary analyses are broken down by
gender (Table 6-8), the RR's for female deaths were generally smaller than that for males,
although none of the risks for PM indices or gaseous pollutants were statistically significant.
           TABLE 6-8.  RELATIVE RISK OF MORTALITY FROM CARDIOPULMONARY
                 CAUSES, BY SEX AND AIR POLLUTANT, FOR AN ALTERNATIVE
                                       COVARIATE MODEL
                                                 Females
                                                                      Males
         Pollution Index
                  Pollution Incr.
RR
LCL
UCL
RR
LCL
UCL
PM10>100, d/yr.
PMio mean
SO4 mean
O3>100ppb, h/yr.
O3 mean
SO2 mean
30 days/yr.
50 ^g/m3
IS^g/m3
551 h/yr. (IQR)
lOppb
3.72 (IQR)
0.929
0.841
0.857
0.88
0.975
1.02
0.857
0.639
0.498
0.76
0.865
0.90
1.007
1.107
1.475
1.02
1.099
1.15
1.062
1.219
1.279
1.06
1.066
1.01
0.971
0.862
0.002
0.87
0.920
0.86
1.162
1.616
1.018
1.29
1.236
1.18
         LCL = Lower 95% confidence limit
         Source: Abbey et al. (1999).
                                   UCL = Upper 95% confidence limit
13          The AH SMOG cancer analyses showed a confusing array of results for lung cancer
14     mortality (Table 6-9). For example, RR's for lung cancer deaths were statistically significant for
15     males for PM10 and O3 metrics, but not for females. In contrast, such cancer deaths were
16     significant for mean NO2 only for females (but not for males), but lung cancer metrics for mean
17     SO2 were significant for both males and females. This pattern is not readily interpretable, but is
       March 2001                               6-89        DRAFT-DO NOT QUOTE OR CITE

-------
          TABLE 6-9. RELATIVE RISK OF MORTALITY FROM LUNG CANCER BY AIR
          POLLUTANT AND BY GENDER FOR AN ALTERNATIVE COVARIATE MODEL

Pollution
Index
PM10>100,d/yr.
PM,0mean
NO2 mean
O3>100ppb,h/yr


O3 mean
SO2 mean


Pollution
Incr.
30 days/yr.
50 //g/m3
19.78
(IQR)
551h/yr
(IQR)


lOppb
3.72 (IQR)


Smoking
Category
All1
All
All
AH
never
smoker
past
smoker
All
All
never
smokers

RR
1.055
1.808
2.81
1.39


0.805
3.01
2.99
Females
LCL
0.657
0.343
1.15
0.53


0.436
1.88
1.66

UCL
1.695
9.519
6.89
3.67


1.486
4.84
5.40

RR
1.831
12.38
1.82
4.19
6.94
4.25
1.853
1.99

Males
LCL
1.281
2.552
0.93
1.81
1.12
1.50
0.994
1.24


UCL
2.617
60.107
3.57
9.69
43.08
12.07
3.453
3.20

         'All = both never smokers and past smokers.
         LCL = Lower 95% confidence limit
         Source: Abbey et al. (1999).
UCL = Upper 95% confidence limit
 1     reasonably attributable to the very small numbers of cancer-related deaths (18 for females and
 2     12 for males), resulting in wide RR confidence intervals and very imprecise effects estimates.
 3          The analyses reported by Abbey et al. (1999) attempted to separate PM10 effects from those
 4     of other pollutants by use of two-pollutant models, but no quantitative findings from such models
 5     were reported. Abbey et al. mentioned that the PM10 coefficient for CRC remained stable or
 6     increased when other pollutants were added to the model. Lung cancer mortality models for
 7     males evaluated co-pollutant effects in detail and indicated that NO2 was non-significant in all
 8     two-pollutant models but the other pollutant coefficients were stable. The PMIO and O3 effects
 9     remained stable when SO2 was added, suggesting possible independent effects, but PM10 and O3
10     effects were hard to separate because these pollutants were highly correlated in this study.
11     Again, however, the very small number of lung cancer observations and likely great imprecision
12     of reported effects estimates markedly diminish the credibility of these results.
       March 2001
     6-90
DRAFT-DO NOT QUOTE OR CITE

-------
  1          Other analyses, by Beeson et al. (1998), evaluated essentially the same data as in Abbey
  2     et al. (1999), but focused on lung cancer incidence (1977-1992). There were only 20 female and
  3     16 male lung cancer cases among the 6,338 subjects. Exposure metrics were constructed to be
  4     specifically relevant to cancer, being the annual average of monthly exposure indices from
  5     January,  1973 through the following months, but ending 3 years before date of diagnosis of the
  6     case (i.e., representing a 3-year lag between exposure and diagnosis of lung cancer). The
  7     covariates in the Cox proportional hazards model were pack-years of smoking and education, and
  8     the time variable was attained age. Many additional covariates were evaluated for inclusion, but
  9     only 'current use of alcohol' met criteria for inclusion in the final model.  Pollutants evaluated
 10     were PM10, SO2, NO2, and O3. No interaction terms with the pollutants proved to be significant,
 11     including outdoor exposure times. The RR estimates for male lung cancer cases were:
 12     (a) positive and statistically significant for all PM10 indicators; (b) positive and predominantly
 13     significant for O3 indicators, except for mean O3, number of O3 exceedances > 60 ppb, and in
 14     former smokers; (c) positive and significant for mean SO2, except when restricted  to proximate
 15     monitors; and (d) positive but not significant for mean NO2. A very high non-credible RR found
 16     for mean PM10 for males (31.1, CI = 3.98, 243.9) may be attributable to the small number of
 17     cases (N = 16) and the large standard increment (50 /ug/m3) used.  When analyses are restricted to
 18     use of air quality data within 32 km of the residences of subjects, the RR is reduced to 9.26 over
 19     50 Mg/m3 and the RR over the IQR of 24 /wg/m3 in the full data set is 5.21. The female RR's were
 20     all much smaller than for males, not being statistically significant for any indicator of PM10 or O3,
 21      but being significant for mean SO2.
 22          The AHSMOG investigators also attempted to compare effects of fine vs. coarse particles
 23     (McDonnell et al, 2000). For AHSMOG participants living near an airport (n=3,769), daily
 24     PM2 5 concentrations were estimated from airport visibility using previously-described methods
 25      (Abbey et al, 1995b). Table 6-10 shows the results of this analysis for the male subset near
 26     airports (n=1266). Given the smaller numbers  of subjects in these subset analyses, it is not
 27      necessarily surprising that no pollutants are statistically significant in these regressions.  It is
 28      important, however, to caveat that the PM2 5 exposures were estimated from visibility
29      measurements (increasing exposure measurement error), and a very uneven and clustered
30      distribution of exposures was presented by the authors. Also, the PM10.2 5 values were calculated
31

        March 2001                               6-91         DRAFT-DO NOT QUOTE OR CITE

-------
               TABLE 6-10. COMPARISON OF REPORTED PM10, PM2 s, and PM10.2 5IQR
            RELATIVE RISKS FOR VARIOUS MORTALITY CAUSES IN A MALE SUBSET
                     OF THE AHSMOG STUDY FOR ONE-POLLUTANT MODELS
                              [and for PM,.„ & PM10.r< Simultaneous Models]


PM,5

(Range = 24.3 ^g/m3)
Underlying
Mortality Cause
Rel.
Risk
RR
95% CI
RR
t-Stat.

PM,(1
(Range = 29 5 /^g/m3)
Rel
Risk
RR RR
95% CI t-Stat.

PM.o.25
(Range = 97 ^g/m
Rel.
Risk
RR
95% CI

3)
RR
t-Stat
         AllCause      1.22 [124]  (0,95,1.58)   1 37 [1 24]  1.15
         [Two                  [0.91,1.67]
         Pollutants]

         Non-Malignant   1.64(1.55]  (0.93,290)   127[097]  1.48
         Respiratory              [0.80,3.03]

         LungCancer    2.23 [2.10]  (0.56,8.94)   0.58 [46]   1.84
         [Two                  [0.45,9.90]
         Pollutants]	
       (0.94,141)  1.25   105 [0.99]   (0.92,1.20)  0.70 [-12]
                                [0.84,1.16]
       (093,2.34)  1.33   1.19 [1.06]   (0.88,1.62)  1 00 [ 30]
                                [0.74,1.52]

       (059,567)  065   1 25 [1 07]   (063,2.49)  0.52 [.15]
                                [0.49,231]
         Source: McDonnell et al., 2000
 1      from the differencing of PM10 and PM2 5, likely contributing to additional measurement error for

 2      the coarse particle (PM10_2 5) variable used in the analyses.

 3

 4      6.2.3.2.3  Relationship of AHSMOG to Six Cities and A CS Study Findings

 5          The results of the recent AHSMOG mortality analyses (Abbey et al.,  1999; McDonnell

 6      et al., 2000) are compared here with findings from the earlier Six Cities study (Dockery et al.,

 7      1993), the ACS study (Pope et al., 1995), and the recent HEI reanalyses of the latter two studies.

 8      Table 6-11 compares the estimated RR for total and cardiopulmonary mortality respectively

 9      among the studies. The number of subjects  in these studies varies greatly (8,111 subjects in the

10      Six-Cities Study; 295,223 subjects in the 50 fine particle cities and 552,138 subjects in the

11      151 sulfate cities of the ACS Study; and 6,338 in the California Seventh-day Adventist

12      AHSMOG Study) and may partially account for differences among their results.

13          As shown in Table 6-11, the Six Cities study found significant associations with  all PM

14      indicators. In the Krewski et al. (2000) reanalysis of the ACS study data, stronger associations

15      were found for PM2 5 than PMI5 (RR's of 1.14 and 1.04 for a 20 /ug/m3 change in PM2 5 and PM,5,

16      respectively), though both associations were significant.  Most recently, McDonnell et  al. (2000)

17      reported evidence from the  AHSMOG analyses suggestive of somewhat stronger associations
       March 2001
6-92
DRAFT-DO NOT QUOTE OR CITE

-------


>
H
-
U
O
C/3
PS
U
U
U
Z
u
2
H
ij
Q
Z <*5
"*^ ^D
„ jgr
^0 *^
W Q
H £
^ £
rs (2
V} EM)
U w
O ^
M t/5
« U
^ «?"
KS
^^^
<

b
(••^
0
z
o
1
o
u
rt
SO
pa
H



















s
0.
c
i
u
t.
8.
2
£
^
"ra
i
s













0
13
•5



c
o
•s a
•M g
Z2
•*. =a
o ^
% a
P-g
H U
-J3 U
e x?
OJ o^
U- U-l
i 0s
**•%
2 t5
I £
"3 S
0 (^






|u

13
0 ^
— 1
*3 ^
H S




U
^^
— >rt
C3 0
H
_ o:
2 ^
o >•
H n




















~55-f^f K^S"
^-^ ^~- ^ 2---^
a.5; 0,^ ^-^-oS^
"Ifs'l^rlliXvo'
0~~ Q~~ 0 O00
s^ s; s; c






•^j ~^j *^j "Q
Is'l^l 1 ^ ?
a. § 5- o a-^SS
•5^--§^:--§-§S.;;:-
§^ |~ | !§§
o"^ o""-1 C O — •• ~"~
c s: c s;





ili&ii^g
^•^-^^"'— ^-s-g.

^ ^
-^ ~^


o

- S -, ^ ^
I g | g g g~ g g


^ ~^* -^
S, "5 Q

"^ a s
ffi 1J ^J
1 •* ^ ^
"*>% rN ""^s 233^^

OOOcoCoCn^^
K K *•* L^ Q^J Qj i ^
i3^o5x^XX^
• u
^s?
5e
13
t04
3 u
                                   J -3

                                   o
                                   "1
                                   U
                                   6-2
lm
lati
                                   U
lmonary mortal]
e Risk (95% Cl]
                                   U
                                                        !§
                                                  b b
                                                               s
                                                               cw
                                                                             s
                                                                             so
                                                                             -
                                                                           s- a
                                                                           ||
                                                                           Sf-a
March 2001
   6-93
DRAFT-DO NOT QUOTE OR CITE

-------
 1      with fine particles than coarse particles, though the associations were only reported for males and
 2      none reached statistical significance.
 3          Overall, the results most recently reported for the AHSMOG study (Abbey et al., 1999;
 4      McDonnell et al., 2000) do not find consistent, statistically significant associations between
 5      mortality and long-term PM exposure, though the authors conclude that some evidence was
 6      suggestive of associations with fine particles. However, the lack of consistent findings in the
 7      AHSMOG study does not cast doubt on the findings of the Six Cities and ACS studies, which
 8      both had larger study populations (especially the ACS study), were based on measured PM data
 9      (in contrast with AHSMOG PM estimates based on TSP or visibility measurements), and have
10      been validated through an exhaustive reanalysis. When considering the results of these three
11      studies, along with the results of the reanalysis of the Six Cities and ACS studies, it can be
12      concluded that there is evidence for an association between long-term exposure to PM (especially
13      fine particles) and mortality.
14          There is no obvious statistically significant relationship between PM effect sizes, gender,
15      and smoking status across these studies. The AHSMOG analyses show no significant
16      relationships between PM10 and total mortality or cardiovascular mortality for either sex, and
17      only for male lung cancer incidence and lung cancer deaths in a predominantly non-smoking
18      sample.  The ACS results, in contrast, show similar and significant associations with total
19      mortality for both "never smokers" and "ever smokers", although the ACS cohort may include a
20      substantial number of long-term former smokers with much lower risk than current smokers.
21      The Six Cities study cohort shows the strongest evidence of a higher PM effect in current
22      smokers than in non-smokers, with female former smokers having a higher risk than male former
23      smokers. This study suggests that smoking status may be viewed as an "effect modifier" for
24      ambient PM, just as smoking may be a health effect modifier for ambient O3 (Cassino et al.,
25      1999).
26          When the ACS study results are compared with the AHSMOG study results for SO4=
27      (PM10 was not considered in the ACS study), the total mortality effect sizes per 15 ^g/m3 SO4=
28      for the males in the AHSMOG population are seen to fall between the Six-Cities and the ACS
29      effect estimates for males: RR= 1.28 for AHSMOG male participants; RR=1.61 for Six-Cities
30      Study male non-smokers; and RR=1.10 for never smoker males in the ACS study. The
31      AHSMOG study 95% confidence intervals encompass both of those other studies' sulfate RR's.

        March 2001                               6-94        DRAFT-DO NOT QUOTE OR CITE

-------
  1      6.2.3.3  Studies by Particulate Matter Size-Fraction and Composition
  2      6,2.3.3.1 Six Cites, ACS, and AH SMOG Study Results
  3          Ambient PM consists of a mixture that may vary in composition over time and from place
  4      to place. This should logically affect the relative toxicity of PM indexed by mass at different
  5      times or locations. Some semi-individual chronic exposure studies have investigated relative
  6      roles of various PM components in contributing to observed air pollution associations with
  7      mortality. Unfortunately, only a limited number of the chronic exposure studies have included
  8      direct measurements of chemical-specific constituents of the PM mixes indexed by mass
  9      measurements used in their analyses.  As shown in Table 6-12, the Harvard Six-Cities study
10      (Dockery et al., 1993) results indicated that the PM2 5 and SO4= RR associations (as indicated by
11      their respective 95% CI's and t-statistics) were more consistent than those for the coarser mass
12      components. However, the effects of sulfate and non-sulfate PM2 5 are indicated to be quite
13      similar. Acid aerosol (H+) exposure was also considered by Dockery et al. (1993), but only less
14      than one year of measurements collected near the end of the follow-up period were available in
15      most cities; so, the Six-Cities results were much less conclusive  for the acidic component of PM
16      than for the other PM metrics measured over many years during  the study. The Six-Cities study
17      also yielded total mortality RR estimates for the reported range across those cities of PM2 5 and
18      SO4= concentrations that, although not statistically different, were roughly double the analogous
19      RR's for the TSP-PM15 and PM15J 5 mass components.
20
21
               TABLE 6-12. COMPARISON OF ESTIMATED RELATIVE RISKS FOR
                ALL-CAUSE MORTALITY IN SIX U.S. CITIES ASSOCIATED WITH
                 THE REPORTED INTER-CITY RANGE OF CONCENTRATIONS
                        OF VARIOUS PARTICULATE MATTER METRICS
PM Species
S04=
PM25-S04=
PM25
PM15.25
TSP-PM,,
Concentration Range
(//g/m3)
8.5
8.4
18.6
9.7
27.5
Relative Risk
Estimate
1.29
1.24
1.27
1.19
1.12
RR
95% CI
(1.06-1.56)
(1.16-1.32)
(1.06-1.51)
(0.91-1.55)
(0.88-1.43)
Relative Risk
t-Statistic
3.67
8.79
3.73
1.81
1.31
        Source:  Dockery et al. (1993); U.S. Environmental Protection Agency (1996).

       March 2001                              6-95        DRAFT-DO NOT QUOTE OR CITE

-------
1
2
3
4
5
6
Table 6-13 presents comparative PM25 and SO4 results from the ACS study, which
indicate that both had substantial, statistically significant impacts in all-cause and
cardiopulmonary mortality. On the other hand, RR for lung cancer was notably larger (and
substantially significant)


for SO4 than PM2 5 (not significant).











TABLE 6-13. COMPARISON OF REPORTED SO4= AND PM25 RELATIVE
RISKS FOR VARIOUS MORTALITY CAUSES IN THE AMERICAN
CANCER SOCIETY STUDY

Mortality Cause
SO4=
(Range = 19.9Mg/m3)
PM25
(Range = 24.5 yug/m3)
Relative RR RR Relative
Risk 95% CI t-Statistic Risk



All Cause
Cardiopulmonary
Lung Cancer
1.15 (1.09-1.22) 4.85
1.26 (1.15-1.37) 5.18
1.35 (1.11-1.66) 2.92
1.17
1.31
1.03
RR RR
95% Cl t-Statistic
(1.09-1.26) 4.
(1.17-1.46) 4
(0.80-1.33) 0,
.24
.79
.38
         Source: Pope et al. (1995).
 7           The most recent AHSMOG study analysis reported by Abbey et al. (1999) employed PM10
 8      as its PM mass index, finding some significant associations with total and by-cause mortality,
 9      even after controlling for potentially confounding factors (including other pollutants). This
10      analysis also considered SO4= as a PM index for all health outcomes studied except lung cancer,
11      but SO4= was not as strongly associated as PM,0 with mortality and was not statistically
12      significant for any mortality category.
13           Overall, the semi-individual long-term PM exposure studies conducted to-date collectively
14      confirm earlier cross-sectional study indications that the fine mass component of PM10 (and
15      usually especially its  sulfate constituent) are more strongly correlated with mortality than is the
16      coarse PM,0_2 5 component. However, the greater precision of PM2 5 population exposure
17      measurement (both analytical and spatial) relative to PM10_2 5 makes conclusions regarding their
18      relative contributions to observed PM10-related associations less certain than if the effect of their
19      relative errors of measurement could be addressed.
20
        March 2001                                6-96        DRAFT-DO NOT QUOTE OR CITE

-------
1
2
3
4
5
6
7
Single-pollutant results about PM components are informative, as shown in Table 6-14 for
total mortality and in Table 6-15 for cardiopulmonary causes. The t-statistics are compared for
studies where appropriate: mean PM10, PM10_2 5, PM2 5, and sulfate for the Six
et al., 1993); mean PM2 5 and sulfate for ACS (Pope et al., 1995); mean PM10
PM10 exceedances of 100 yug/m3 for AHSMOG (Abbey et al., 1999).


Cities (Dockery
and sulfate, and



TABLE 6-14. COMPARISON OF TOTAL MORTALITY RELATIVE RISK
















ESTIMATES AND T-STATISTICS FOR PARTICULATE MATTER
IN THREE PROSPECTIVE COHORT STUDIES
PM Index Study Subgroup Relative Risk
PMIO (50 ^g/rn3) Six Cities All 1.504a; 1.530b
Male Nonsmoker 1.280a
AHSMOG Male Nonsmoker 1.242
PM2 5 (25 ywg/m3) Six Cities All 1 .364a; 1 .379b
Male Nonsmoker 1 .207a
ACS (50 cities) All 1.174
Male Nonsmoker 1 .245
SO4=(15,ug/m3) Six Cities All 1.504a; 1.567"
Male Nonsmoker 1.359
ACS (151 cities) All 1.111
Male Nonsmoker 1 . 1 04
AHSMOG Male Nonsmoker 1.279
Days/yr. with AHSMOG Male Nonsmoker 1.082
PM10>100^g/m3
(30 days)
PMlo.25(25/ug/m3) Six Cities All 1.8143; 1.560b
Male Nonsmoker 1 .434a
COMPONENTS

t Statistic
2.94a; 3.27"
0.8 la
1.616
2.94a; 3.73"
0.81a
4.35
1.960
2.94a; 3.67b
0.81"
5.107
1.586
0.960
2.183
2.94" 1.816"
0.81a
 aMethod 1 compares Portage vs. Steubenville (Table 3, Dockery et al., 1993).
 bMethod 2 is based on ecologic regression models (Table 12-18, U.S. Environmental Protection Agency, 1996).
 'Method 1 not recommended for PM10.25 analysis, due to high concentration in Topeka.
March 2001
6-97
DRAFT-DO NOT QUOTE OR CITE

-------
        TABLE 6-15. COMPARISON OF CARDIOPULMONARY MORTALITY RELATIVE
              RISK ESTIMATES AND T-STATISTICS FOR PARTICIPATE MATTER
                  COMPONENTS IN THREE PROSPECTIVE COHORT STUDIES
PM Index
PM.oCSO^ug/m3)


PM25(25/ug/m3)



S04=(15Mg/m3)





Days/yr. with
PM!0>100(30days)

PM10.25(25Mg/m3)
Study
Six Cities
AHSMOG

Six Cities
ACS (50 cities)


Six Cities
ACS (151 cities)


AHSMOG

AHSMOG

Six Cities
Subgroup
All
Male Nonsmoker
Male Non. - CRCC
All
All
Male
Male Nonsmoker
All
All
Male
Male Nonsmoker
Male Nonsmoker
Male Non. - CRCC
Male Nonsmoker
Male Non. - CRCC
All
Relative Risk
1.744a
1.219
1.537
1.527a
1.317
1.245
1.245
1.743a
1.190
1.147
1.205
1.279
1.219
1.082
1.188
2.251a
t Statistic
2.94a
1.120
2.369
2.94a
4.699
3.061
1.466
2.94a
5.470
3.412
2.233
0.072
0.357
1.310
2.370
2.94ab
       "Method 1 compares Portage vs. Steubenville (Table 3, Dockery et al., 1993).
       bMethod 1 not recommended for PMI0.25 analysis due to high concentration in Topeka.
       cMale non. - CRC = AHSMOG subjects who died of any contributing non-malignant respiratory cause.
1          Estimates for Six Cities parameters were calculated in two ways: (1) mortality RR for the

2     most versus least polluted city in Table 3 of Dockery et al. (1993) adjusted to standard

3     increments; and (2) ecological regression fits in Table 12-18 of U.S. Environmental Protection

4     Agency (1996). The Six Cities study of eastern and mid-western U.S. cities suggests a strong
5     and highly significant relationship for fine particles and sulfates, a slightly weaker but still highly

6     significant relationship to PM,0, and a marginal relationship to PMIO_2 5.  The ACS study looked at
      March 2001
6-98
DRAFT-DO NOT QUOTE OR CITE

-------
 1      a broader spatial representation of cities, and found a stronger statistically significant relationship
 2      to PM2 5 than to sulfate (no other pollutants were examined). The AHSMOG study at California
 3      sites (where sulfate levels are typically low) found significant effects in males for PM10
 4      100 Aig/m3 exceedances, and a marginal effect of mean PM10, but no PM effects for females or
 5      with sulfates.  On balance, the overall results shown in Tables 6-14 and 6-15 suggest statistically
 6      significant relationships between long-term exposures to PM10, PM2 5, and/or sulfates and excess
 7      total and cause-specific cardiopulmonary mortality.
 8
 9      6.2.3.3.2 The Washington University-EPRI Veterans' Cohort Mortality Study
10           Lipfert et al. (2000b) recently reported preliminary results from new large-scale mortality
11      analyses using a prospective cohort of up to 70,000 men assembled by the U.S. Veterans
12      Administration (VA) in the mid 1970s. Like the ACS cohort, this cohort was not originally
13      designed to study air pollution, but was linked to air pollution data collected separately,  much of
14      it subsequent to the start of the study.  The study cohort was male, middle-aged (51 + 12 years)
15      and included a larger proportion of African-Americans (35%) than the U.S.  population as a
16      whole and a large percentage of current or former smokers (81%). The  cohort was selected at the
17      time of recruitment as being mildly to moderately hypertensive, with screening diastolic blood
18      pressure (DBP) in the range 90 to 114 mm Hg (mean 96, about 7 mm more than the U.S.
19      population average) and average systolic blood pressure (SBP) of 148 mm Hg. Also, although
20      the subjects had all been healthy enough to be in the U.S. armed forces at one time, their
21      pre-existing health status at time of study recruitment should be noted:  12% had a pulmonary
22      abnormality on physical examination, 9% were diabetic, 19% had a history of heart disease, 7%
23      had a history of stroke, and 56% had a positive cardio-renal family history.
24           The subjects received all or most of their medical care at a VA facility, perhaps suggesting
25      relatively low socioeconomic status.   The  medical care was presumably similar for the entire
26      cohort since it was standardized in the Hypertension, Screening, and Treatment Program (HSTP)
27      clinics. Long records of patient treatment allowed comparison of the health of the mortality
28      cohort with the U.S. population at large, showing about the same proportion of heart attacks, but
29      slightly higher percentages of lung cancer and stroke.  Apart from treatment records and
30      determination of vital status, there appeared to be no further follow-up on other individual risk
31      factors; for example, it is likely that the veterans enrolled in the HTSP received medication or

        March 2001                                6-99        DRAFT-DO NOT QUOTE  OR CITE

-------
 1      other interventions during subsequent visits to VA clinics if they were found to have elevated
 2      blood pressure, whereas no such regular diagnosis or treatment can be assumed for comparable
 3      individuals outside this VA cohort. Pollutant levels of the county of residence at the time of
 4      entry into the study were used for analyses versus levels at the VA hospital area.  Contextual
 5      socioeconomic variables were also assembled at the ZIP-code and county levels. The ZIP-code
 6      level variables were average education, income, and racial mix. County-level variables included
 7      altitude, average annual heating-degree days, percentage Hispanic, and socioeconomic indices.
 8      Census tract variables included poverty rate and racial mix. County-wide air pollution variables
 9      included TSP, PMIO, PM2 5, PM15, PM15.2 5, SO4, O3, CO, and NO2 levels at each of the 32 VA
10      clinics where veterans were enrolled. The study that led to the development of this clinical
11      cohort (Veterans Administration Cooperative Study Group on Antihypertensive Agents, 1970;
12      1967) was a "landmark" VA cooperative study demonstrating that anti-hypertensive treatment
13      markedly decreased morbidity and mortality. The clinical cohort itself involves actual clinical
14      rather than research settings.  Thus, there was no single protocol, no matched control or placebo
15      groups, and no extensive data collection forms to provide systemic information on such things as
16      the presence of other risk factors (for hypertension) (Perry et al., 1982).
17           Three sequential mortality follow-up periods (1976-81, 1982-88, 1989-96) were considered
18      separately in statistical analyses, which evaluated relationships of mortality in each of those
19      periods to air pollution in different preceding,  concurrent, or subsequent periods (i.e., up to 1975,
20      1975-81, 1982-88, and 1989-86, for TSP in the first three periods, PM10 for the last, and NO2,
21      95 percentile O3, and 95 percentile CO for all four periods). Mortality in the above-noted periods
22      was also evaluated in relation to SO4 in each of the same four periods noted for NO2, O3, and CO,
23      and to PM25, PM15, and PM15.25 in 1979-81 and 1982-84. The preliminary screening models used
24      proportional hazards regression models to identify age, SBP, DBP, body mass index (BMI,
25      nonlinear), age and race interaction terms, and present or former smoking as baseline predictors,
26      with one or two pollution variables added.  In the final model using 233 terms (of which 162
27      were interactions of categorized SBP, DBP, and BMI variables with age), the most significant
28      non-pollution variables were SBP, DBP, BMI, and their interactions with age, smoking status,
29      average ZIP education, race, poverty, height, and a clinic-specific effect.  Many of the particle
30      effects were not statistically significant, or were significantly negative. The most consistently
31      positive effects were found for O3 and NO2 exposures in the immediately preceding years.

        March 2001                               6-100       DRAFT-DO NOT QUOTE OR CITE

-------
  1           It is difficult to assess the methodological soundness of this study or to interpret its
  2      preliminary results.  These findings may reflect one or more unintentional forms of confounding
  3      (e.g., use of SBP and DBF as mortality predictors). Elevated SBP and DBP are generally
  4      accepted as risk factors in mortality.  However, elevated blood pressure may also be an important
  5      intermediate step in a causal pathway from PM exposure to cardiopulmonary disease states
  6      and/or related mortality, although the mechanisms involved are not fully understood. Possible
  7      mechanisms may include inducing irregularities in heart rate or rhythms, decreased heart rate
  8      variability, increasing blood viscosity, and so on (as discussed elsewhere), with elevated blood
  9      pressure possibly serving as a marker for such pathophysiological changes. Rothman and
 10      Greenland (1998, p. 255) warn that, "In order to avoid bias due to the inappropriate control of
 11      variables, the following criterion is usually added to the list: 3. It [the variable] must not be
 12      affected by exposure or disease (although it may affect exposure). ... The third criterion excludes
 13      variables that are intermediate on the causal pathway from exposure to disease. This exclusion
 14      can be relaxed under certain conditions, but in doing so special analytic techniques must be
 15      applied . . ." Such techniques apparently have not been applied in this case. Data reanalysis,
 16      omitting SBP and DBP as predictors  of mortality, might clarify the situation.
 17           Also complicating interpretation of the study is the choice of the study population. The
 18      restricted range of DBP at the time of recruitment limits the comparability of these findings to
 19      more general populations, including both veterans and civilians with lower DBP or with more
20      highly elevated DBP (not necessarily medically treated). Even if the veterans were
21      disproportionately of low SES, their enrollment in the HTSP by reason of national service likely
22      allowed them more regular treatment (e.g., by medication) of their identified hypertension than
23      the general public.  Such medication would presumably considerably narrow the range of blood
24      pressure values downward in subsequent years of the study compared to the SBP and DBP values
25      at time of recruitment into the study, thus making it more difficult to detect PM effects on
26      mortality related to blood pressure or other associated cardiovascular pathophysiology.
27
28      6.2.3.4 Population-Based Mortality Studies in Children
29           Older cross-sectional mortality studies suggest that the very young may represent an
30      especially susceptible sub-population for PM-related mortality. For example, Lave and Seskin
31      (1977) found mortality among those 0-14 years of age to be significantly associated with TSP.

        March 2001                              6-101       DRAFT-DO NOT QUOTE OR CITE

-------
 1      More recently, Bobak and Leon (1992) studied neonatal (ages less than one month) and post-
 2      neonatal mortality (ages 1-12 months) in the Czech Republic and reported significant and robust
 3      associations between post-neonatal mortality and PM10, even after considering other pollutants.
 4      Post-neonatal respiratory mortality showed highly significant associations for all pollutants
 5      considered, but only PM10 remained significant in simultaneous regressions. The exposure
 6      duration was longer than a few days, but shorter than in the adult prospective cohort studies.
 7      Thus, the limited available studies reviewed in the 1996 PM AQCD were highly suggestive of an
 8      association between ambient PM concentrations and infant mortality, especially among post-
 9      neonatal infants.
10          More recent studies since the 1996 PM AQCD have focused specifically on ambient PM
11      relationships to (a) intrauterine mortality (Pereira et al., 1998) and morbidity (Dejmek et al.,
12      1999) and (b) early post neonatal mortality (Woodruff et al., 1997; Bobak and Leon, 1999;
13      Loomis et al., 1999; Lipfert et al., 2000c).  In the case of the Pereira study of intrauterine (pre-
14      natal) mortality during one year (1991-1992) in Brazil, PM10 was not found to be a significant
15      predictor, but CO's involvement was suggested by the association between increased
16      carboxyhemoglobin (COHb) in fetal blood and ambient CO levels on the day of delivery
17      measured in a separate study.  Another study (Dejmek et al., 1999) evaluated possible impacts of
18      ambient PM10 and PM2 5 exposure (monitored by EPA-developed VAPS methods) during
19      pregnancy on intrauterine growth retardation (IUGR) risk in the highly polluted Teplice District
20      of Northern Bohemia in the Czech Republic during three years (1993-1996). Mean levels of
21      pollutants (PM, NO2, SO2) were calculated for each month of gestation and three concentration
22      intervals (low, medium, high)  derived for each pollutant. Preliminary analyses of data found no
23      significant associations of IUGR with NO2, but SO2 and PM,0 early in pregnancy were
24      significantly associated with IUGR. Odds ratios for IUGR for PM,0 and PM2 5 levels were
25      determined by logistic regressions for each month during gestation, after adjusting for potential
26      confounding  factors (e.g., smoking, alcohol consumption during pregnancy, etc.). Definition of
27      an IUGR birth was any one for which the birth weight fell below the 10th percentile by gender
28      and age for live births in the Czech Republic (1992-93). The OR's for IUGR were significantly
29      related to PM,0 during the first month of gestation:  that is, as compared to low PM10, the medium
30      level PM10 OR= 1.47  (CI 0.99-2.16), and the high level PMIO OR = 1.85 (CI 1.29-2.66). PM25
31      levels were highly correlated with PM,0 (r = 0.98) and manifested similar patterns (OR = 1.16, CI

        March 2001                              6-102        DRAFT-DO NOT QUOTE OR CITE

-------
  1     0.08-0.69 for medium PM2 5 level; OR = 1.68, CI 1.18-2.40 for high PM2 5 level). These results
  2     suggest effects of PM exposures (probably including fine particles such as sulfates, acid aerosols,
  3     and PAHs in the Teplice ambient mix) early in pregnancy (circa embryo implantation) on
  4     subsequent fetal growth and development.
  5           More consistent results indicating likely early post-natal PM exposure effects on neonatal
  6     infant mortality have emerged from other new studies.  Woodruff et al. (1997), for example, used
  7     cross-sectional methods to evaluate possible association of post-neonatal mortality with ambient
  8     PM10 pollution.  This study involved an analysis of a cohort of circa 4 million infants born during
  9     1989 - 1991 in 86 U.S. metropolitan statistical areas (MSAs). Data from the National Center for
 10     Health Statistics-linked birth/infant death records were combined at the MSA level with PM10
 11     data from EPA's Aerometric database.  Infants were categorized as having high, medium, or low
 12     exposures based on tertiles of PM10 averaged over the first 2 postnatal months. Relationships
 13     between this early neonatal PM10 exposure  and total and cause-specific post-neonatal mortality
 14     rates (from 1 mo. to 1 yr of age) were examined using logistic regression analyses, adjusting for
 15     demographic and environmental factors. Overall post-neonatal mortality rates per 1,000 live
 16     births were 3.1 among infants in areas with low PM10 exposures, 3.5 among infants with medium
 17     PM10 exposures, and 3.7 among highly PM exposed infants. After adjustment for other
 18     covariates, the odds ratio (OR) and 95% confidence intervals for total post-neonatal mortality for
 19     the high versus the low exposure group was 1.10 (CI= 1.04-1.16). In normal birth weight infants,
 20     high PM10 exposure was associated with mortality for respiratory causes (OR = 1.40, CI=1.05-
 21      1.85) and sudden infant death syndrome (OR  = 1.26, CI=1.14-1.39). Among low birth weight
 22     babies, high PM10 exposure was positively (but not significantly) associated with mortality from
 23      respiratory causes (OR =1.18, CI=0.86-1.61). However, other pollutants (e.g., CO) were not
 24      considered as possible confounders.  This study provides results consistent with some earlier
 25      reports indicating that outdoor PM air pollution may be associated with increased risk of post-
 26      neonatal mortality (e.g., Bobak and Leon, 1992), but lack of consideration of other air pollutants
 27      as potential confounders in this new study reduces the certainty that PM is the specific causal
28      outdoor air pollutant in this case.
29           Recently, Lipfert et al. (2000c) reported replicating the basic findings of Woodruff et  al.
30      (1997) using a  similar modeling approach but annual average PMIO air quality data for one year
31      (1990) instead  of PM10 averaged over the first two post natal months during 1989-1991. The

        March 2001                                6-103        DRAFT-DO NOT QUOTE  OR CITE

-------
  1      quantitative relationship between the individual risk of infant mortality did not differ among
  2      infant categories (by age, by birthweight, or by cause), but PM10 risks for SIDs deaths were
  3      higher for babies of smoking mothers.  SO4= was a strong negative predictor of SIDs mortality for
  4      all age and birth weight categories. The authors (a) noted difficulties in ascribing the reported
  5      PM10 and SO4= associations to effects of the PM pollutants per se versus the results possibly
  6      reflecting interrelationships between  the air pollution indices, a strong well-established
  7      East-West gradient in U.S. SIDS cases, and/or underlying sociodemographic factors (e.g., the
  8      socioeconomic or education level of parents) and (b) hypothesized that a parallel gradient in use
  9      of wood burning in fireplaces or woodstoves and consequent indoor wood smoke exposure might
10      explain the observed cross-sectional study results.
11           The basic findings from Woodruff et al. (1997) also appear to be bolstered by a more recent
12      follow-up study by Bobak and Leon (1999), who conducted a matched population-based
13      case-control study covering all births registered in the Czech Republic from 1989 to 1991 that
14      were linked to death records. They used conditional logistic regression to estimate the effects of
15      suspended particles and nitrogen oxides on risk of death in the neonatal and early post-neonatal
16      period, controlling for maternal socioeconomic status and birth weight, birth length, and
17      gestational age. The effects of all pollutants were strongest in the post-neonatal period and
18      specific for respiratory causes. Only  PM showed a consistent association when all pollutants
19      were entered in one model.  Thus, in  this study, it appears that long-term exposure to PM is the
20      air pollutant metric most strongly associated with excess post-neonatal deaths.
21           The study by Loomis et al. (1999) of infant mortality in Mexico City during 1993-1995
22      adds additional interesting information pointing towards likely fine particle impacts on infant
23      mortality. That is, in Mexico City (where mean 24-h PM2 5 = 27.4 ptg/m3), infant mortality was
24      found to be associated with PM2 5, NO2, and O3 in single pollutant GAM Poisson models, but
25      much less consistently with NO2, and O3 than PM2 5 in multipollutant models. The estimated
26      excess risk for PM2 5-related infant mortality lagged 3-5 days was  18.2% (95% CI 6.4, 30.7) per
27      25 Mg/m3 PM2 5. It is not clear, however, the extent to which such a notable increased risk for
28      infant mortality might be extrapolated to U.S. situations,  due to (a) possible differences in
29      prenatal maternal or early post natal infant nutritional status and/or (b) possible enhancement of
30      PM-related risks associated with exposure to PM under higher altitude conditions in Mexico City
31      versus most U.S. cities.

        March 2001                               6-104        DRAFT-DO NOT QUOTE OR CITE

-------
  1      6.2.3.5  Shortening-of-Life Associated With Long-Term Ambient Particulate Matter
  2              Exposure
  3           The public health burden of mortality associated with exposure to ambient PM depends not
  4      only on the increased risk of death, but also on the length of life shortening that is attributable to
  5      those deaths. However, the 1996 PM AQCD concluded that confident quantitive determination
  6      of years of life lost to ambient PM exposure is not yet possible; life shortening may range from
  7      days to years (U.S. Environmental Protection Agency, 1996). A newly published analysis has
  8      now attempted to estimate life-shortening associated with chronic PM exposures.
  9
 10      6.2.3.5.1 Life-Shortening Estimates Based on Semi-Individual Cohort Study Results
 11           Brunekreef (1997) reviewed the available evidence of the mortality effects of long-term
 12      exposure to PM air pollution and, using life table methods,  derived an estimate of the reduction
 13      in life expectancy implied by those effect estimates.  Based on the results of Pope et al. (1995)
 14      and Dockery et al. (1993), a relative risk of 1.1  per 10 /ug/m3 exposure over 15 years was
 15      assumed for the effect of PM air pollution on men 25-75 years of age. A 1992 life table for men
 16      in the Netherlands was developed for 10 successive five-year categories that make up the
 17      25-75 year old age  range.  Life expectancy of a 25 year old  was then calculated for this base case
 18      and compared with the calculated life expectancy for the PM exposed case where the death rates
 19      were increased in each age group by a factor of 1.1. A difference of 1.11 years was found
 20      between the "exposed" and "clean air" cohorts' overall life  expectancy at age 25. Looked at
 21      another way, this implies that the expectation of the lifespan of persons who actually died from
 22      air pollution was reduced by more than 10 years, since they represent a small percentage of the
 23      entire cohort population. A similar calculation by the authors for the 1969-71 life table for U.S.
 24      white males yielded an even larger reduction of 1.31 years for the entire population's life
 25      expectancy at age 25.  Thus, these calculations imply that relatively small differences in long-
 26      term exposure to ambient PM can have substantial effects on life expectancy.
 27
28      6.2.3.5.2  Potential Effects of Infant Mortality on Life-Shortening Estimates
29           Deaths among children can logically have the greatest influence on a population's overall
30      life expectancy, but the Brunekreef (1997) life table calculations did not consider any possible
31      long-term air pollution exposure effects on the population aged <25 years. As discussed above,

        March 2001                               6-105        DRAFT-DO NOT QUOTE OR CITE

-------
  1      some of the older cross-sectional studies and the more recent studies by Bobak and Leon (1992),
  2      Woodruff et al. (1997), Bobak and Leon (1999), and Loomis et al. (1999) suggest that infants
  3      may be among sub-populations notably affected by long-term PM exposure. Thus, although it is
  4      difficult to quantify, any premature PM associated mortality that does occur among children due
  5      to long-term PM exposure, as suggested by these studies, would significantly increase the overall
  6      population life shortening over and above that estimated by Brunekreef (1997) for long-term PM
  7      exposure of adults aged 25 years and older.
  8
  9      6.2.3.6 Salient Points Derived from Analyses of Chronic Particulate Matter Exposure
10             Mortality Effects
11           A review of the studies summarized in the previous PM AQCD (U.S. Environmental
12      Protection Agency, 1996) indicates that past epidemiologic studies of chronic PM exposures
13      collectively indicate increases in mortality to be associated with long-term exposure to airborne
14      particles of ambient origins.  The PM effect size estimates  for total mortality from these studies
15      also indicate that a substantial portion of these deaths reflected cumulative PM impacts above
16      and beyond those exerted by acute exposure events.
17           The recent HEI-sponsored reanalyses of the ACS and Harvard Six-Cities studies (Krewski
18      et al., 2000) "replicated the original results, and tested those results against alternative risk
19      models and analytic approaches without substantively altering the original findings of an
20      association between indicators of particulate matter air pollution and mortality."  Several
21      questions, including the questions (1-4) posed at the outset of this Section (6.2.3) were
22      investigated by the Krewski et al. (2000) sensitivity analyses for the Six City and ACS studies
23      data sets.  Key results emerging from the HEI reanalyses and other new chronic PM mortality
24      studies are as follow:
25           (1) A much larger number of confounding variables and effects modifiers were considered
26      in the Reanalysis Study than in the original Six  City and ACS studies.  The only significant air
27      pollutant other than PM2 5 and SO4 in the ACS study was SO2, which greatly decreased the PM2 5
28      and sulfate effects  when included as a co-pollutant (Krewski et al., 2000, Part II, Tables 34-38).
29      A similar reduction in particle effects occurred in any multi-pollutant model with SO2. The most
30      important new effects modifier was education.  The AHSMOG study suggested that other metrics
31      for air pollution, and other personal covariates such as time spent outdoors and consumption of

        March 2001                               6-106        DRAFT-DO NOT QUOTE OR CITE

-------
  1     anti- oxidant vitamins, might be useful. Both individual- level covariates and ecological-level
  2     covariates shown in (Krewski et al., 2000, Part II, Table 33) were evaluated.
  3          (2) Specific attribution of excess long-term mortality to any specific particle component or
  4     gaseous pollutant was refined in the reanalysis of the ACS study. Both PM2 5and sulfate were
  5     significantly associated with excess total mortality and cardiopulmonary mortality and to about
  6     the same extent whether the air pollution data were mean or median long-term concentrations or
  7     whether based on Original Investigator or Reanalysis Team data. The association of mortality
  8     with PM15 was much smaller, though still significant, and the associations with the coarse
  9     fraction (PM15.2 5) or TSP were even smaller and not significant.  The lung cancer effect was
 10     significant only for sulfate with the original investigator data, or for new investigators with
 11     regional sulfate artifact adjustment for the 1980-1981 data (Krewski et al., 2000, Part II,
 12     Table 31). Associations of mortality with long-term mean concentrations of criteria gaseous
 13     co-pollutants were generally non-significant except for SO2 (Krewski et al., 2000, Part II, Tables
 14     32, 34-38) which was highly significant, and for cardiopulmonary disease with warm-season
 15     ozone. However, the regional association of SO2 with SO4 and SO2 with PM2 5 was very high,
 16     and the effects of the separate pollutants could not be distinguished. Krewski et al. (2000,
 17     p. 234) concluded that, "Collectively, our reanalyses suggest that mortality may be associated
 18     with more than one component of the complex mix of ambient air pollutants in urban areas of the
 19     United States."
 20          (3) The extensive temporal data on air pollution concentrations over time in the Six City
 21     Study allowed the Reanalysis Team to evaluate time scales for mortality for long-term exposure
 22     to a much greater extent than reported in Dockery et al. (1993). The first approach was to
 23     estimate the  log- hazard ratio as a function of follow up time using a flexible spline-function
 24     model (Krewski et al., 2000, Part II, Figures 2 and 3). The results for both SO4 and PM2 5 suggest
 25     very similar relationships, with larger risk after initial exposure decreasing to 0 after about 4 or
 26     5 years, and a large increase in risk at about 10 years follow-up time.
 27          The analyses of the ACS Study proceeded somewhat differently, with less temporal data
28     but many more cities. Flexible spline regression models for  PM2 5 and sulfate as function of
29     estimated cumulative exposure (not defined) were very nonlinear and showed quite different
30     relationships (Krewski et al., 2000, Part II, Figures 10 and 11). The PM2 5 relationship shows the
31      mortality log-hazard ratio increasing up to about 15 Aig/m3 and relatively flat above about

        March 2001                               6-107       DRAFT-DO NOT QUOTE OR CITE

-------
 1      22 //g/m3, then increasing again. The sulfate relationship is almost piecewise linear, with a low
 2      near- zero slope below about 11 /ug/m3 and a steep increase above that concentration.
 3           A third approach evaluated several time-dependent PM2 5 exposure indicators in the Six
 4      City study.  They are: (a) constant (at the mean) over the entire follow-up period; (b) annual
 5      mean within each of the 13 years of the study; (c) city-specific mean concentration for the earliest
 6      years of the study, i.e., very long-term effect; (d) exposure estimate in 2 years preceding death;
 7      (e) exposure estimate in 3 to 5 years preceding death; (f) exposure estimate > 5 years preceding
 8      death. The time-dependent estimates (a-e) for mortality risk are generally similar and statistically
 9      significant (Krewski et al., 2000, Part II, Table 53), with RR of 1.14 to 1.19 per 24.5 //g/m3 being
10      much lower than the risk of 1.31 estimated for exposure at the constant mean for the period.
11      Thus, it is highly likely the duration and time patterns of long-term exposure  affect the risk of
12      mortality, and further study of this question (along with that of mortality displacement from
13      short-term exposures) would improve estimates of life-years lost from PM exposure.
14           (4)  The Reanalysis Study also advanced our understanding of the shape of the relationship
15      between mortality and PM. Again using flexible spline modeling, Krewski et al. (2000, Part II,
16      Figure 6) found a visually near linear relationship between all cause and cardiopulmonary
17      mortality residuals and mean sulfate concentrations, near linear between cardiopulmonary
18      mortality and mean PM25, but a somewhat nonlinear relationship between all cause mortality
19      residuals  and mean PM2 5 concentrations that flattens above about 20 yUg/m3.  The relationship
20      with lung cancer is much weaker, as noted above. The confidence bands around the fitted curves
21      are very wide, however, neither requiring a linear relationship nor precluding a nonlinear
22      relationship if suggested by reanalyses. An investigation of the mortality relationship for other
23      indicators may be useful in identifying a threshold, if one exists, for chronic PM exposures.
24           (5)  With regard to the role of various PM constituents in the PM-mortality association, past
25      cross-sectional studies have generally found that the fine particle component, as indicated either
26      by PM2 5 or sulfates, was the PM constituent most consistently associated with mortality. While
27      the relative measurement errors of the various PM constituents must be further evaluated as a
28      possible source of bias in these estimate comparisons, the Six-Cities and AHSMOG prospective
29      semi-individual studies both indicate that the fine mass components of PM are  more strongly
30      associated with the mortality effects of chronic PM  exposure than are coarse PM components.


        March 2001                               6-108        DRAFT-DO NOT  QUOTE  OR CITE

-------
  1           (6)  Recent investigations of the public health implications of such chronic PM exposure-
  2      mortality effect estimates were also reviewed. Life table calculations by Brunekreef (1997)
  3      found that relatively small differences in long-term exposure to airborne PM of ambient origin
  4      can have substantial effects on life expectancy. For example, a calculation for the 1969-71 life
  5      table for U.S. white males indicated that a chronic exposure increase of 10 yUg/m3 PM was
  6      associated with a reduction of 1.31 years for the entire population's life expectancy at age 25.
  7      Also, new evidence of associations  of PM exposure with infant mortality (Bobak and Leon,
  8      1992, 1999; Woodruff et al., 1997; Loomis et al., 1999) and/or intrauterine growth retardation
  9      (Dejmek et al., 1999) and consequent increase risk for many serious health conditions associated
 10      with low birth weight, if further substantiated, would imply that life shortening in the entire
 11      population from long-term PM exposure could well be significantly larger than that estimated by
 12      Brunekreef (1997).
 13
 14
 15      6.3 MORBIDITY EFFECTS OF PARTICIPATE MATTER EXPOSURE
 16           This morbidity discussion is presented below in several subsections, dealing with:  (a) acute
 17      cardiovascular morbidity effects of ambient PM exposure; (b) effects of short-term PM exposure
 18      on the incidence of respiratory and other medical visits and hospital admissions; and (c) short-
 19      and long-term PM exposure effects  on lung function and respiratory symptoms in asthmatics and
20      non-asthmatics.
21
22      6.3.1  Cardiovascular Effects Associated with Acute Ambient Particulate
23            Matter Exposure
24      6.3.1.1  Introduction
25          Very little information specifically addressing acute cardiovascular morbidity effects of PM
26      existed at the time of the 1996 PM AQCD. While the literature still remains relatively sparse, an
27      important new body of data now exists that both extends the available quantitative information
28      on the ecologic relationship between ambient pollution and hospital admissions and which, more
29      intriguingly, illuminates some of the physiological changes that may occur on the mechanistic
30      pathway leading from PM exposure to adverse cardiac outcomes.

        March 2001                               6-109       DRAFT-DO NOT QUOTE OR CITE

-------
 1           This section begins with a brief summary of the conclusions that were reached in the 1996
 2      PM AQCD regarding acute cardiovascular impacts of PM. Next, new studies are reviewed
 3      which fall into two general classes: ecologic time series studies of daily hospitalizations in
 4      relation to ambient PM and other pollutants; and individual-level studies of temporal changes in
 5      physiological measures of cardiac function as they relate to ambient pollution.  This review is
 6      followed by discussion of several issues that are important in interpreting the available data,
 7      including the identification of potentially susceptible sub-populations, the roles of environmental
 8      co-factors such as weather and other air pollutants, temporal lags in the  relationship between
 9      exposure and outcome, and the relative importance of various size-classified PM components
10      (e.g.,PM25,PM,0,PM10.25).
11
12      6.3.1.2  Summary of Key Findings on Cardiovascular Morbidity from  the 1996 Particulate
13              Matter Air quality Criteria Document
14           Just two studies were available for review in the 1996 PM AQCD that provided data on
15      acute cardiovascular morbidity outcomes (Schwartz and Morris, 1995; Burnett et al., 1995).
16      Both studies were of ecologic time series design, using standard statistical methods. Analyzing
17      four years of data on the > 65 year old Medicare population in Detroit, MI,  Schwartz and Morris
18      (1995) reported significant associations between ischemic heart disease admissions and PM10,
19      controlling for environmental covariates.  Based on an analysis of admissions data from
20      168 hospitals throughout Ontario, Canada, Burnett and colleagues (1995) reported significant
21      associations between fine particle sulfate concentrations, as well as other air pollutants, and daily
22      cardiovascular admissions. The relative risk due to sulfate particles was slightly larger for
23      respiratory than for cardiovascular hospital admissions.  The 1996 PM AQCD concluded on the
24      basis of these studies that, "There is a suggestion of a relationship to heart disease, but the results
25      are based on only two studies and the estimated effects are smaller than those for other
26      endpoints" (U.S. Environmental Protection Agency, 1996 p. 12-100). The  PM AQCD went on
27      to state that acute impacts on CVD admissions had been demonstrated for elderly populations
28      (i.e., > 65), but that insufficient data existed to assess relative impacts on younger populations.
29           Also relevant to an evaluation of the acute impacts of particles on cardiovascular endpoints
30      are insights gained from time series studies of daily mortality, which, aside from the outcome
31      studied,  are nearly identical in design and analysis to time series studies of hospitalizations. It is

        March 2001                               6-110       DRAFT-DO NOT QUOTE OR CITE

-------
  1     also probable that acute effects of air pollution on cardiovascular hospitalizations and mortality
  2     follow qualitatively similar etiologic mechanisms.
  3           Several acute mortality studies reviewed in the 1996 PM AQCD analyzed cause-specific
  4     deaths (usually total cardiovascular and total respiratory) in relation to ambient particle
  5     concentrations. The PM AQCD noted that, in general, cause-specific analyses "reported higher
  6     estimated relative risks for respiratory and cardiovascular categories than for total or other
  7     categories" (U.S. Environmental Protection Agency, 1996 p. 12-349). It was noted that these
  8     findings were consistent with analyses of case reports from historic air pollution episodes, like
  9     the December, 1952 London episode, in which the mortality impacts were greatest among the
 10     elderly and those with pre-existing respiratory and/or cardiovascular disease. A comparative
 11     analysis of age- and cause-specific mortality effects  of particles in modern-day Philadelphia with
 12     those observed in the 1952 London episode concluded that the patterns of mortality were largely
 13     consistent, once the order of magnitude difference in exposure levels was taken into account
 14     (Schwartz,  1994a).
 15          Viewed as a group, the acute morbidity and mortality studies reviewed in the 1996 PM
 16     AQCD were thus consistent with the notion that acute health risks of PM are larger for
 17     cardiovascular and respiratory causes than for other  causes.  Given the tendency for end-stage
 18     disease states to include both respiratory and cardiovascular impairment, and the associated
 19     diagnostic overlap that often exists, it was not possible on the basis of these studies alone to
20     determine which of the two organ systems, if either, was more critically impacted.
21
22     6.3.1.3  New Particulate Matter-Cardiovascular Morbidity Studies
23      6.3.1,3.1 Acute Hospital Admission Studies
24           Numerous new studies have reported associations between daily measures of ambient PM
25      and daily hospital admissions for cardiovascular disease (see Table 6-16). Again, of particular
26      interest are results from multi-city studies, as discussed most extensively below, which likely
27      yield more precise effect estimates than those derived from smaller independent studies of
28      individual cities with fewer overall study observations. Results from several new multi-city
29      studies (Schwartz, 1999; Samet et al., 2000a,b; Zanobetti et al., 2000b), that provide combined
30      estimates of PM-CVD effects across numerous U.S.  cities and regions, provide evidence
31      substantiating significant PM effects on cardiovascular-related hospital admissions and visits.

        March 2001                               6-111        DRAFT-DO NOT QUOTE OR CITE

-------
65
3
to
o
o
             TABLE 6-16.  ACUTE PARTICULATE MATTER EXPOSURE AND CARDIOVASCULAR HOSPITAL ADMISSIONS
Reference citation.
Location, Duration
PM Index, Mean or
Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covanates, concentration-response
                                                                                                Results and Comments. Design Issues,
                                                                                                Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
o
z;
3
o
g
H
ra
O
          United States

          Samet et al. (2000a,b)
          14 US cities
          1985-1994, but range of years
          varied by city

          PMm (/ug/m3) mean, median, IQR:
          Birmingham, AL: 34.8, 30.6, 26.3
          Boulder, CO: 24.4, 22.0, 14.0
          Canton, OH: 28.4,25.6, 15.3
          Chicago, II:  36.4, 32.6, 22 4
          Colorado Springs, CO: 26.9, 22.9,
          11.9
          Detroit, MI:  36.8, 32.0, 28.2
          Mmneapolis/St. Paul, MM:  27.4,
          24.1, 17.9
          Nashville, TN: 31.6,29.2, 17.9
          New Haven, CT: 29.3, 26.0, 20.2
          Pittsburgh, PA: 36.0, 30.5, 27.4
          Provo/Orem, UT: 38.9, 30.3, 22.8
          Seattle, WA: 31.0,26.7,20.0
          Spokane, WA: 45.3, 36.2, 33.5
          Youngstown, OH:  33.1,29.4,
          18.6
                                  Daily medicare hospital admissions for total
                                  cardiovascular disease, CVD (ICD9 codes 390-429), in
                                  persons 65 or greater. Mean CVD counts ranged from
                                  3 to 102/day in the 14 cities. Covanates: SO2, NO2, O3,
                                  CO, temperature, relative humidity, barometric pressure.
                                  Stats:  In first stage, performed city-specific, single-
                                  pollutant, generalized additive robust Poisson regression
                                  with seasonal, weather, and day of week controls.
                                  Repeated analysis for days with PM1U less than 50 Mg/m3
                                  to test for threshold. Lags of 0-5 considered, as well as
                                  the quadratic function of lags 0-5.  Individual cities
                                  analyzed first. The 14 risk estimates were then analyzed
                                  in several second stage analyses: combining risks across
                                  cities using inverse variance weights, and regressing risk
                                  estimates on potential effect-modifiers and pollutant
                                  confounders.
                                                     City-specific risk estimates for a 10 /ug/m3
                                                     increase in PMH, ranged from -1.2% in
                                                     Canton to 2.2% in Colorado Springs.
                                                     Across-city weighted mean risk estimate
                                                     was largest at lag 0, diminishing rapidly at
                                                     other lags. Only the mean of lags 0 and
                                                     1 was significantly associated with CVD.
                                                     There was no evidence of statistical
                                                     heterogeniety in risk estimates across cities
                                                     for CVD,  City-specific risk estimates were
                                                     not associated with the percent of the
                                                     population that was non-white, living in
                                                     poverty, college educated, nor unemployed.
                                                     No evidence was observed that PM,0 effects
                                                     were  modified by weather.  No multi-
                                                     pollutant modeling results presented.
                                                     However, no association was observed
                                                     between the  city-specific PM,0 risk
                                                     estimates and the city-specific correlation
                                                     between PMlo and co-pollutants.  These
                                                     results suggest a weak association between
                                                     PM|0 and total cardiovascular hospital
                                                     admissions among the elderly.
Percent Excess CVD Risk (95%
CI), combined over cities per
50 /ug/m3 change in PM,,,.

PMU): Odlag.
 5.5% (4.7, 6.2)
PMIO: 0- Id lag.
 6.0% (5. 1,6.8)
                 0-1 d lag.
 7.6% (6.0, 9.1)
 n

-------
                         TABLE 6-16 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND CARDIOVASCULAR
                        	                      HOSPITAL ADMISSIONS
O
O
          Reference citation.
          Location, Duration
          PM Index, Mean or
          Median, IQR
                                           Study Description: Health outcomes or codes,
                                           Mean outcome rate, sample or population size,
                                           ages.  Concentration measures or estimates.
                                           Modeling methods: lags, smoothing, co-pollutants,
                                           covanates, concentration-response
                                                     Results and Comments Design Issues,
                                                     Uncertainties, Quantitative Outcomes
                                         PM Index, Lag, Excess
                                         Risk % (95% LCL, UCL),
                                         Co-Pollutants
ON

OJ




O
 H
 6
 o
 2
 o
 H
O
 c
 o
 H
 W
 O
 ^
 o
 HH
 H
 W
          United States (cont'd)

          Schwartz (1999)
          8 US metropolitan counties
          1988-1990
          median, IQR for PM,0 (,ug/m3):
          Chicago, IL: 35,23
          Colorado Springs, CO: 23, 14
          Minneapolis, MN: 28,15
          New Haven, CT:  37, 25
          St. Paul, MN: 34,23
          Seattle, WA:  29,20
          Spokane, WA: 37,33
          Tacoma, WA: 37,27
          Zanobetti et al. (2000b)
          10 US cities
          1986-1994

          PM,,, (,ug/m3) median, IQR:
          Canton, OH: 26, 15
          Birmingham, AL: 31, 26
          Chicago, II: 33,23
          Colorado Springs, CO: 23, 13
          Detroit, MI: 32, 28
          Minneapolis/St. Paul, MN: 24,18
          New Haven, CT: 26,21
          Pittsburgh, PA: 30, 28
          Seattle, WA: 27,21
          Spokane, WA: 36, 34
                                           Daily hospital admissions for total cardiovascular
                                           diseases (1CD9 codes 390-429) among persons over
                                           65 years. Median daily hospitalizations: 110,3,  14,
                                           18, 9, 22, 6, 7, alphabetically by city. Covanates: CO,
                                           temperature, dewpoint temp. Stats'  robust Poisson
                                           regression after removing admission outliers; generalized
                                           additive models with LOESS smooths for control of
                                           trends,  seasons, and weather. Day of week dummy
                                           variables Lag 0 used for all covanates.
Derived from the Samet et al. (2000a,b) study, but for a
subset of  10 cities. Daily hospital admissions for total
cardiovascular disease, CVD (ICD9 codes 390-429), in
persons 65 or greater. Median CVD counts ranged from
3 to 103/day in the 10 cities. Covariates: SO2, O3, CO,
temperature, relative humidity, barommetnc pressure.
Stats: In first stage, performed single-pollutant
generalized additive robust Poisson regression with
seasonal, weather, and day of week controls.  Repeated
analysis for days with PMH) less than 50 Aig/m3 to test
for threshold. Lags of 0-5 considered, as well as the
quadratic function of lags 0-5. Individual cities analyzed
first. The 10 risk estimates were then analyzed in several
second stage analyses: combining nsks across cities
using inverse variance weights, and regressing nsk
estimates on potential effect-modifiers and pollutant
confounders.
                                                     In single-pollutant models, similar PM,,,
                                                     effect sizes obtained for each county. Five
                                                     of eight county-specific effects were
                                                     statistically significant, as was the PM1()
                                                     effect pooled across locations. CO effects
                                                     significant in six of eight counties. The
                                                     PM|(1 and CO effects were both significant
                                                     in a two pollutant model that was run for
                                                     five counties where the PMH/CO
                                                     correlation was less than 0.5.  Results
                                                     reinforce those of Schwartz, 1997.
Same basic pattern of results as in Samet
et al. (2000a,b). For distnbuted lag
analysis, lag 0 had largest effect, lags 1 and
2 smaller effects, and none at larger lags.
City-specific slopes were independent of
percent poverty and percent non-white.
Effect size increase when data were
restricted to days with PMH) less than 50
^g/m3. No multi-pollutant models
reported; however, no evidence of effect
modification by co-pollutants in second
stage analysis.  Suggests a weak association
between PM10 and total cardiovascular
hospital admissions among the elderly.
Percent Excess Risk (95% CI).
Effects computed for 50 //g/m3
change in PM10.

PM10. Od.
Individual counties:
Chicago: 4.7 (2.6, 6.8)
COSpng-  5.6 (-6.8, 19.0)
Minneap:  4 1 (-3.6, 12.5)
NewHav: 5.8(2.1,97)
St. Paul:  8 6 (2.9, 14.5)
Seattle. 3.6 (~0.1, 7.4)
Spokane:  67(0.9,  12.8)
Tacoma:  5.3(3.1,7.6)

Pooled: 5.0 (3.7, 6.4)
3.8(2.0, 5.5) w, CO

Percent Excess Risk (SE)
combined over cities:
Effects computed for 50 A'g/m3
change in  PM10.

PM,,,:  Od.
 5.6 (4.7,  6.4)
PM10:  0-1 d.
 6.2 (5.4,  7.0)
PMUI < 50/^g/m3:  0-1 d.
 7.8 (6.2, 9.4)

-------
                         TABLE 6-16 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND CARDIOVASCULAR
                                                                      HOSPITAL ADMISSIONS
o
o
Reference citation.
Location, Duration
PM Index, Mean or
Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages.  Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
                                                                                                Results and Comments. Design Issues,
                                                                                                Uncertainties, Quantitative Outcomes
                                         PM Index, Lag, Excess
                                         Risk % (95% LCL, UCL),
                                         Co-Pollutants
Ox
a
Tl
H
6
o
z;
o
H
O
c
o
H
 O
 t— H
 H
 W
          United States (cont'd)

          Linn et al. (2000)
          Los Angeles
          1992-1995

          Mean PMI(1 n, (,ug/m3):  18
Morris and Naumova (1998)
Chicago, IL
1986-1989
mean, median, IQR, 75th
percentile:
PM10 (//g/m3): 41,38,23,51
Schwartz (1997)
Tucson, AZ
1988-1990
mean, median, IQR:
PMIO (//g/m3): 42, 39, 23
Hospital admissions for total cardiovascular diseases
(CVD), congestive heart failure (CHF), myocardial
infarction (MI), cardiac arrhythmia (CA) among all
persons 30 years and older, and by sex, age, race, and
season. ICD9 codes not given.  Mean hospital
admissions for CVD: 428.  Covariates: CO, NO2, O3,
temperature, rainfall. Daily gravimetric PMIO estimated
by regression of every sixth day PM|0 on daily real-time
PM|0 data collected by TEOM.  Poisson regression with
controls for seasons and day of week. Reported results
for lag 0 only. Results reported as Poisson regression
coefficients and their standard errors. The number of
daily CVD admissions associated with the mean PM,0
concentration can be computed by multiplying the PMU1
coefficient by the PMH) mean  and then exponentiating.
Percent effects are calculated  by dividing this result by
the mean daily admission count for CVD.

Daily hospital admissions for congestive heart failure,
CHF (ICD9 428), among persons over 65 years. Mean
hospitalizations: 34/day. Covariates: O3, NO2, SO2,
CO, temperature, relative humidity.  Gases measured at
up to eight sites, daily PMH, measured at one site.  Stats:
GLM for time series data. Controlled for trends and
cycles using dummy variables for day of week, month,
and year. Residuals were modeled as negative binomial
distribution. Lags of 0-3 days examined.

Daily hospital admissions for total cardiovascular
diseases (ICD9 codes 390-429) among persons over
65 years.  Mean hospitalizations:  13.4/day. Covariates:
03, NO2, CO, SO2, temperature, dewpoint temperature.
Gases measured at multiple sites; daily PMH) at one site.
Stats: robust Poisson regression; generalized additive
model with LOESS smooth for controlling trends and
seasons, and regression splines to control weather.
Lags of 0-2 days examined.
                                                                                      In year-round, single-pollutant models,
                                                                                      significant effects of CO, N02, and PMI()
                                                                                      on CVD were reported.  PMH) effects
                                                                                      appeared larger in winter and fall than
                                                                                      in spring and summer. No consistent
                                                                                      differences in PM,0 effects across sex,
                                                                                      age, and race. No multi-pollutant results
                                                                                      presented.
CO was only pollutant statistically
significant in both single- and
multi-pollutant models. Exposure
misclassification may have been larger
for PM10 due to single site. Results suggest
effects of both CO and PM,0 on congestive
heart failure hospitalizations among
elderly, but CO effects appear more robust.


Both PM10 (lag 0) and CO significantly and
independently associated with admissions,
whereas other gases were  not. Sensitivity
analyses reinforced these basic results.
Results suggest independent effects of both
PMIO and CO for total cardiovascular
hospitalizations among the elderly.
                                         % increase with PM,0 change of
                                         50 ,ug/m3:

                                         PMlfflot:  Od.
                                         CVD ages 30+
                                          3.25% (2.04, 4.47)

                                         MI ages 30+
                                          3.04% (0.06, 6.12)

                                         CHF ages 30+
                                          2.02% (-0.94, 5.06)

                                         CA ages 30+
                                          1.01% (-1.93, 4.02)
Percent Excess Risk (95% CI)
per 50 ptg/m3 change in PM,,,.

PM,,,: Od.
 3.92% (1.02, 6.90)
 1.96% (-1.4, 5.4) with
   4 gaseous pollutants
Percent Excess Risk (95% CI)
per 50 pig/m3 change in PMU).

PMI(): Od.
 6.07% (1.12, 1.27)-
 5.22% (0.17, 10.54) w. CO

-------
3
to
o
o
TABLE 6-16 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND CARDIOVASCULAR
                                             HOSPITAL ADMISSIONS
          Reference citation.
          Location, Duration
          PM Index, Mean or
          Median, IQR
                  Study Description: Health outcomes or codes,
                  Mean outcome rate, sample or population size,
                  ages.  Concentration measures or estimates.
                  Modeling methods: lags, smoothing, co-pollutants,
                  covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
O
o
2
o
H
O
o
H
W
O
*>
O
HH
51
          United States (cont'd)

          Gwynn et al (2000)
          Buffalo, NY
          Lippmann et al. (2000)
          Detroit, MI
          1992-1994
                  Air pollution health effects associations with total,
                  respiratory, and CVD hospital admissions (HA's)
                  examined using Poisson model controlling for weather,
                  seasonably, long-wave effects, day of week, holidays.
                  Vanous cardiovascular (CVD)-related, and respiratory
                  (COPD, Pneumonia) hospital admissions (HA's) for
                  persons 65+ yr. analyzed, using GAM Poisson models,
                  adjusting for season, day of week, temperature, and
                  relative humidity. The air pollution variables analyzed
                  were: PM10, PM25, PMUI.25, sulfate, H+, O3, SO2, NO2,
                  and CO. However, this study site/period had very low
                  acidic aerosol levels. As noted by the authors  85% of H+
                  data was below detection limit (8 nmol/m3).
Positive, but non-significant assoc. found
between all PM indices and circulatory
hospital admissions. Addition of gaseous
pollutants to the model had minimal effects
on the PM RR estimates.
For heart failure, all PM metrics yielded
significant associations. Associations for
IHD, dysrhythmia, and stroke were positive
but generally non-sig. with all PM indices.
Adding gaseous pollutants had negligible
effects on various PM metric RR estimates.
Most consistent effect of adding co-
pollutants was to widen confidence bands
on the PM metric RR estimates; a common
statistical artifact of correlated predictors.
Despite usually low levels, H+ had strong
association with respiratory admissions on
the few days it  was present. The general
similarity of the PM25 and PM,(K, 5 effects
per //g/m3 in this study suggest  similarity in
human toxicity of these two inhalable mass
components in study locales/penods where
PM2 5 acidity not usually present.
Percent excess CVD HA nsks
(95% CI) per PMIO = 50 Mg/m3;
SO4 = 15 ^g/m1;
H+ = 75 nmoles/m3; COH = 05
units/1,000 ft:
PM10(lag3) = 5.7%(-3.3, 15.5)
SO4(Iagl) = 0.1%(-0.1,04)
H" (lag 0)= 1.9% (-0.3, 4.2)
COH (lag 1) = 2.2% (-1.9, 6.3)

Percent excess CVD H A nsks
(95% CI) per 50 ,ug/m3 PMIO,
25 /ug/m3 PM25 and PM10.25:

IHD:
  PM2 5 (lag 2) 4.3 (-1.4, 10.4)
  PMIO (lag 2) 8.9 (0.5, 18.0)
  PM1025(lag2)10.5(2.7, 18.9)
Dysrhythmia:
  PM25 (lag 1)3.2 (-6.5, 14.0)
  PM10 (lag 1)2.9 (-6.8,  13.7)
  PM10.25 (lag 0)0.2 (-12.2,  14.4)
Heart Failure:
  PM, 5 (lag 1)9.1 (2.4,16.2)
  PM10 (lag 0)9.7 (0.2, 20.1)
  PM10.2 5 (lag 0)5.2 (-3.3, 14.5)
Stroke:
  PM2 5 (lag 0)1.8 (-5.3, 9.4)
  PMIO (lag 1)4.8 (-5.5, 16.2)
  PM1M5 (lag 1)4.9 (-4.7, 15.5)

-------
I
O
o
               TABLE 6-16 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND CARDIOVASCULAR
                                                           HOSPITAL ADMISSIONS
Reference citation.
Location, Duration
PM Index, Mean or
Median, 1QR
Study Descnption:  Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covanates, concentration-response
                                                                                             Results and Comments. Design Issues,
                                                                                             Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
 H
 a
 o
 g
 H
/O
 c
 o
 H
 W
 O
 pa
 n
 !•—<
 H
          United States (cont'd)

          Moolgavkar (2000b)
          Three urban counties: Cook, IL;
          Los Angeles, CA; Mancopa, AZ.
          1987-1995

          Pollutant median, IQR:
          Cook: PM,(): 35, 22
          LA: PMU): 44, 26
             PM25:22, 16
          Mancopa: PM,,,:41, 19
                                 Analysis of daily hospital admissions for total
                                 cardiovascular diseases, CVD, (ICD9 codes 390-429)
                                 and cerebrovascular diseases, CRD, (ICD9 430-448)
                                 among persons aged 65 and over.  For Los Angeles,
                                 a second age group, 20-64, was also analyzed. Median
                                 daily CVD admissions were 110, 172, and 33 in Cook,
                                 LA, and Mancopa counties, respectively. PM,0 available
                                 only every sixth day in LA and Maricopa counties.
                                 In LA, every-sixth-day PM2 5 also was available.
                                 Covanates: CO, N02, O3, SO2, temperature, relative
                                 humidity.  Stats: generalized additive Poisson regression,
                                 with controls for day of week and smooth temporal
                                 variability. Single-pollutant models estimated for
                                 individual lags from 0 to 5. Two-pollutant models also
                                 estimated, with both pollutants at same lag.
                                                   In single-pollutant models in Cook and LA
                                                   counties, PM was significantly associated
                                                   with CVD admissions at lags 0, 1, and 2,
                                                   with diminishing effects over lags. PM2 5
                                                   was also significant in LA for lags 0 and 1.
                                                   For the 20-64 year old age group in LA,
                                                   nsk estimates were similar to those for 65+.
                                                   In Mancopa county, no positive PMU)
                                                   associations were observed at any lag.
                                                   In two-pollutant models in Cook and LA
                                                   counties, the PM^PM^ risk estimates
                                                   diminished and/or were rendered non-
                                                   significant.  Little evidence observed for
                                                   associations between CRD admissions and
                                                   PM. These results suggest that PM is not
                                                   independently associated with CVD or
                                                   CRD hospital admissions.
Percent Excess CVD Risk (95%
CD
Effects computed for 50 Mg/rn3
change in PMlo and 25 //g/m3
change in PM25.
Cook 65+:
PMI(), 0 d.
 4.2 (3.0, 5.5)
PMM), Od w/NO2.
 1.8(0.4,3.2)
LA 65+:
PMlo, 0 d.
 3.2(1.2,5.3)
PM10, 0 d. w/CO
 -1.8 (-4.4, 0.9)

PM25)Od.
 4.3(2.5,6.1)
PM25, Od. w/CO
 0.8 (-1.3, 2.9)

LA 20-64 years old:
PMIO> 0 d.
 4.4 (2.2, 6.7)
PM,0, 0 d. w/CO
 1.4 (-1.3, 4.2)
PM2S, Od.
 3.5(1.8,5.3)
PM2S, Od., w/CO)
 2.3 (-0.2, 4.8)
Maricopa:
PM10, 0 d.
 -2.4 (-6.9, 2.3)

-------
I
O
O
               TABLE 6-16 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND CARDIOVASCULAR
                                                             HOSPITAL ADMISSIONS
Reference citation.
Location, Duration
PM Index, Mean or
Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
                                                                                                Results and Comments. Design Issues,
                                                                                                Uncertainties, Quantitative Outcomes
                                        PM Index, Lag, Excess
                                        Risk % (95% LCL, UCL),
                                        Co-Pollutants
ON

-J
 O
 o
 2
 3
o
 G
 O
 H
 W
 O
 7*
 O
 H
 W
United States (cont'd)

Tolbert et al. (2000a)
Atlanta
Period): 1/1/93-7/31/98
Mean, median, SD:
PM|0 0/g/m3):  30.1,28.0, 12.4

Period 2: 8/1/98-8/31/99
Mean, median, SD:
PM10(^g/m3):  29.1,27.6, 12.0
PM25(,ug/m3): 19.4, 17.5,9.35
CP (Mg/m3):  9.39, 8.95, 4.52 10-
100 nm PM counts
(count/cm3):  15,200, 10,900,
26,600
10-100 nm PM surface area
(um2/cm3): 62.5,43.4,116
PM,5 soluble metals (,ug/m3):
0.0327, 0.0226, 0.0306
PM25 Sulfates (Mg/m3):  5.59, 4.67,
3.6
PM2s Acidity Cug/m3): 0.0181,
0.0112,0.0219
PM25 organic PM (^g/m3). 6.30,
5.90,3.16
PM2 5 elemental carbon (/ug/m3):
2.25,1.88, 1.74
Preliminary analysis of daily emergency department (ED)
visits for dysrhythmias, DYS, (ICD 9 code 427) and all
cardiovascular diseases, CVD, (codes 402, 410-414, 427,
428, 433-437, 440, 444, 451-453) for persons aged 16
and older in the period before (Period 1) and during
(Period 2) the Atlanta superstation study. ED data
analyzed here from just 18 of 33 participating hospitals;
numbers of participating hospitals increased during
period 1.  Mean daily ED visits for dysrhythmias and all
CVD in period 1 were 6.5 and 28.4, respectively.  Mean
daily ED visits for dysrhythmias and all CVD in period
2 were 11.2 and 45.1, respectively.  Covariates: NO2,
O3, S02, CO temperature, dewpoint, and, in period 2
only, VOCs. PM measured by both TEOM and Federal
Reference Method; unclear which used in analyses.
For epidemiologic analyses, the two time periods were
analyzed separately.  Poisson regression analyses were
conducted with cubic splines for time, temperature and
dewpoint. Day of week and hospital entry/exit indicators
also included. Pollutants were treated a-priori as three-
day moving averages of lags 0, 1, and 2. Only single-
pollutant  results reported.
In period 1, significant negative association
(p=0.02) observed between CVD and 3-day
average PMI0. There was ca. 2% drop in
CVD per 10 //g/m3 increase in PM10. CVD
was positively associated with NO2
(p=0.11) and negatively associated with
SO2 (p=0.10). No association observed
between dysrhythmias and PM,0 in period
1. However, dysrhythmias were positively
associated with NO2 (p=0.06).  In period 2,
i.e., the first year of operation of the
superstation, no associations seen with
PMH, or PM2 5. However, significant
positive associations observed between
CVD and elemental carbon (p=0.005) and
organic matter (p=0.02), as well as with CO
(p=0.001).  For dysrhythmias, significant
positive associations observed with
elemental carbon (p=0.004), CP (p=0.04),
and CO (p=0.005).  These preliminary
results should be interpreted with caution
given the incomplete and variable nature of
the databases analyzed.  Of most concern is
the extent to which the data included in the
reported  analyses from only about half the
participating hospitals are representative of
the entire set across the study area.
Percent Excess Risk (p-value):
Effects computed for 50 ,ug/m3
change in PMH); 25 Mg/m3 for CP
and PM25; 25,000 counts/cm3 for
10-100 nm counts.

Period 1:
PM,,,, 0-2 d. avg.
CVD:  -8.2(0.02)
DYS: 4.6 (0.58)

Period 2:
0-2 d. avg. in all cases
CVD % effect; DYS % effect:
PMIO: 5.1 (-7.9, 19.9); 13.1
(-14.1,50.0)
PM25: 6.1 (-3.1,16.2); 6.1
(-12.6,28.9)
CP: 17.6 (-4.6, 45.0); 53.2 (2.1,
129.6)
10-100 nm counts: -11.0
(0.17); 3.0 (0.87)

-------
O
O
O
                        TABLE 6-16 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND CARDIOVASCULAR
                                                                      HOSPITAL ADMISSIONS
         Reference citation.
         Location, Duration
         PM Index, Mean or
         Median, IQR
                                  Study Description:  Health outcomes or codes,
                                  Mean outcome rate, sample or population size,
                                  ages. Concentration measures or estimates.
                                  Modeling methods' lags, smoothing, co-pollutants,
                                  covariates, concentration-response
                                                    Results and Comments. Design Issues,
                                                    Uncertainties, Quantitative Outcomes
                                         PM Index, Lag, Excess
                                         Risk % (95% LCL, UCL),
                                         Co-Pollutants
         Canada

         Burnett et al. (1995)
         Ontario, Canada
         1983-1988

         Sulfate
         Mean: 4.37 /ug/m3
         Median:  3.07 AJg/m3
         95th percentile: 13 pi
                                  168 Ontario hospitals.  Hospitalizations for coronary
                                  artery disease, CAD (ICD9 codes 410,413), cardiac
                                  dysrhythmias, DYS (code 427), heart failure, HF (code
                                  428), and all three categories combined (total CVD).
                                  Mean total CVD rate: 14.4/day.  1986 population of
                                  study area: 8.7 million. All ages, <65, >=65. Both sexes,
                                  males,  females.  Daily sulfates from nine monitoring
                                  stations.  Ozone from 22 stations. Log hospitalizations
                                  filtered with 19-day moving average prior to GEE
                                  analysis. Day of week effects removed. 0-3 day lags
                                  examined. Covariates: ozone, ozone2, temperature,
                                  temperature2. Linear and quadratic sulfate terms
                                  included in model.
                                                     Sulfate lagged one day significantly assoc.
                                                     with total CVD admissions with and
                                                     without ozone in the model. Larger
                                                     associations observed for coronary artery
                                                     disease and heart failure than for cardiac
                                                     dysrhythmias.  Suggestion of larger
                                                     associations for males and the sub-
                                                     population 65 years old and greater. Little
                                                     evidence for seasonal differences in sulfate
                                                     effects after controlling for covariates.
                                         Effects computed for 95th
                                         percentile change in SO4

                                         SO4, 1 d, no covariates:

                                         Total CVD: 2.8(1.8,3.8)
                                         CAD:  2.3(0.7,3.8)
                                         DYS:  1.3 (-2.0, 4.6)
                                         HF: 3.0(0.6,5.3)

                                         Males: 3.4(1.8,5.0)
                                         Females:  2.0(0.2,3.7)

                                         <65:  2.5 (0.5, 4.5)
                                         >=65:  3.5(1.9,5.0)
a
8
Z
3
o
H
m
O
&
o
HH
H
W
Burnett etal. (1997a)
Canada's 10 largest cities
1981-1994

COH daily maximum
Mean: 0.7 103 In feet
Median: 0.6 103 In feet
95th percentile:  1.5  103 In feet
Daily hospitalizations for congestive heart failure (ICD9
code 427) for patients over 65 years at 134 hospitals.
Average hospitaiizations:  39/day. 1986 population of
study area:  12.6 million. Regressions on air quality
using generalized estimating equations, controlling for
long-term trends, seasonality, day of week, and inter-
hospital differences. Models fit monthly and pooled over
months. Log hospitalizations filtered with 19-day
moving average prior to GEE analysis. 0-3 day lags
examined.  Covariates: CO, SO2, NO2, O3, temperature,
dewpoint temperature.
COH significant in single-pollutant models
with and without weather covariates. Only
InCO and In NO2 significant in multi-
pollutant models. COH highly colinear
with CO and NO2.  Suggests no particle
effect independent of gases. However, no
gravimetric PM data were included.
SO4, Id, w. temp and O3:

Total CVD: 3.3 (1.7,4.8)

Effects computed for 95% change
in COH:

0 d lag:
  5.5% (2.5, 8.6)
0 d lag w/weather:
  4.7% (1.3, 8.2)
0 d lag w/CO, NO2, SO2, 03:
  -2.26 (-6.5,  2.2)

-------
3.
K)
O
O
TABLE 6-16 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND CARDIOVASCULAR
                                            HOSPITAL ADMISSIONS                            	
         Reference citation.
         Location, Duration
         PM Index, Mean or
         Median, IQR
                  Study Description:  Health outcomes or codes,
                  Mean outcome rate, sample or population size,
                  ages. Concentration measures or estimates.
                  Modeling methods: lags, smoothing, co-pollutants,
                  covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
H
6
o
2
3
         Canada (cont'd)

         Burnett etal. (1997b)
         Metro-Toronto, Canada
         1992-1994

         Pollutant: mean, median, IQR:
         COH(103lnft): 0.8,0.8,0.6
         H+(nmol/m3): 5, 1,6
         SO4 (nmol/m3): 57,33,57
         TP(Mg/m3):  28,25,22
         FPO/g/m3):  17, 14,  15
         CP(,ug/m3):  12,10,7
                  Daily unscheduled cardiovascular hospitalizations (ICD9
                  codes 410-414,427, 428) for all ages. Average hospital
                  admissions: 42.6/day.  Six cities of metro-Toronto
                  included Toronto, North York, East York, Etobicoke,
                  Scarborough, and York, with combined 1991 population
                  of 2.36 million. Used same stat model as in Burnett
                  et al., 1997c.  0- 4 day lags examined, as well as multi-
                  day averages.  Covariates: O3, NO2, SO2, CO,
                  temperature, dewpomt temperature.
Relative risks > 1 for all pollutants in
univanate regressions including weather
variables; all but H+ and FP statistically
significant.  In multivanate models, the
gaseous pollutant effects were generally
more robust than were paniculate effects.
However, in contrast to Burnett et al.
(1997c), COH remained significant in
multivariate models. Of the remaining
particle metrics, CP was the most robust to
the inclusion of gaseous covanates. Results
do not support independent effects of FP,
SO4, or H+ when gases are controlled.
Percent excess risk (95% CI) per
50 Mg/m3 PM,,,, 25 ^tg/rn3 PM25
and PM(10_25), and IQR for other
indicators.
COH: 0-4 d.
 6.2 (4.0, 8.4)
 5.9(2.8,9.1) w. gases
H+:2-4d.
 2.4 (0.4, 4.5)
 0.5 (-1.6, 2.7) w. gases
S04: 2-4 d.
 1.7 (-0.4, 3.9)
 -1.6 (-4.4, 1.3) w gases
TP: l-4d.
 7.7(0.9,14.8)
 -0.9 (-8.3, 7.1) w. gases
FP:2-4d.
 5.9(1.8,10.2)
 -1.1 (-7.8, 6.0) w gases
CP:0-4d.
  13.5(5.5,22.0)
 8.1 (-1.3, 18.3) w. gases
 O
 H
 trt
 O
 &
 O
 HH
 H

-------
O
O
                        TABLE 6-16 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND CARDIOVASCULAR
                                                                     HOSPITAL ADMISSIONS
Reference citation.
Location, Duration
PM Index,  Mean or
Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
                                                                                               Results and Comments.  Design Issues,
                                                                                               Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
 Tl
 H
 O
 H
O
          Canada (cont'd)

          Burnett etal. (1999)
          Metro-Toronto, Canada
          1980-1994

          Pollutant: mean, median, IQR:
          FPN Gug/m3):  18,16,10
          CPCS, G/g/tn3):  12,10,8
          PMlow(Atg/m5): 30,27,15
                                 Daily hospitalizations for dysrhythmias, DYS (ICD9
                                 code 427; mean 5/day); heart failure, HF (428; 9/d);
                                 ischemic heart disease, IHD (410-414; 24/d); cerebral
                                 vascular disease, CVD (430-438; 10/d); and diseases of
                                 the penpheral circulation, DPC (440-459; 5/d) analyzed
                                 separately in relation to environmental covariates. Same
                                 geographic area as m Burnett et al., 1997b. Three size-
                                 classified PM metrics were estimated, not measured,
                                 based on a regression on TSP, S04, and COH in a subset
                                 of every 6th-day data. Generalized additive models used
                                 and non-parametric LOESS prefilter applied to both
                                 pollution and hospitalization data. Day of week controls.
                                 Tested  1-3 day averages of air pollution ending on lags
                                 0-2. Covariates: O3, NO2, SO2, CO, temperature,
                                 dewpoint temperature, relative humidity.
                                                    In univanate regressions, all three PM
                                                    metrics were associated with increases in
                                                    cardiac outcome (DVS, HF, IHD). No
                                                    associations with vascular outcomes, except
                                                    for CPest with DPC. In multi-pollutant
                                                    models, PM effects estimates reduced by
                                                    variable amounts (often >50%) for specific
                                                    endpoints and no statistically significant (at
                                                    p<0.05) PM associations seen with any
                                                    cardiac or circulatory outcome (results not
                                                    shown). Use of estimated PM metrics limits
                                                    interpretation of pollutant-specific results.
                                                    However, results suggest that  linear
                                                    combination of TSP, SO4, and COH does
                                                    not have a strong independent association
                                                    with cardiovascular admissions when full
                                                    range of gaseous pollutants also modeled.
Single pollutant models:
Percent excess risk (95% CI) per
50 //g/m3 PM10; 25 ^g/m3 PM2S;
and25/ug/m3PM(lo.25).

All cardiac HA (lags 2-5 d):
PM2 5 1-poll = 8.1(2.45, 14.1)
PM25 w/4 gases = -1.6 (-10.4,
8.2); w/CO = 4.60 (-3.39, 13.26)
PM,0l-poll = 12.07 (1.43, 23.81)
w/4 gases =-1.40 (-12.53,
11.16)
w/CO= 10.93 (0.11,22.92)
PM^ 1-poll = 20.46 (8.24,
34.06)
w/4 gases =12.14 (-1.89, 28.2);
w/CO=19.85(7.19, 34.0)
DYS:
FPal(Od): 6.1(1.9, 10.4)
CP^Od):  5.2 (-0.21,1 08)
PM1()est: (Od): 8.41 (2.89,14.2)

HF:
FPra,(0-2d):  6.59(2.50,10.8)
CPH, (0-2d):  7.9(2.28,13)
PM10a,(0-2d): 9.7(4.2,15.5)
IHD:
FPB,(0-2d):  8.1(5.4, 10.8)
CP^Od):  3.7(1.3,6.3)
                                                                                                                                      PM,,
                                                                                                                                   ,(0-1 d): 8.4(5.3, 11.5)
 O
 t— (
 H
 w

-------
o
tf
K)
O
O
TABLE 6-16 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND CARDIOVASCULAR
                                              HOSPITAL ADMISSIONS
          Reference citation.
          Location, Duration
          PM Index, Mean or
          Median, IQR
                   Study Description:  Health outcomes or codes,
                   Mean outcome rate, sample or population size,
                   ages. Concentration measures or estimates.
                   Modeling methods: lags, smoothing, co-pollutants,
                   covariates, concentration-response
Results and Comments.  Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
 "rt
 H
 O
 H
O
 G
 O
 H
 w
 O
 &
 O
 HH
 H
 W
          Canada (cont'd)

          Stieb et al. (2000)
          Saint John, Canada
          7/1/92-3/31/96
          mean and S.D.:
                     ):  14.0,9.0
                     ):  8.5,5.9
          H+(nmol/m3):  25.7,36.8
          Sulfate (nmol/m3): 31.1,29.7
          COM mean (103 In ft): 0.2,0.2
          COHmax(103lnft):  0.6,0.5
                   Study of daily emergency department (ED) visits for
                   angina/myocardial infarction (mean 1.8/day), congestive
                   heart failure (1.0/day), dysrhythmia/conduction
                   disturbance (0.8/day), and all cardiac conditions
                   (3.5/day) for persons of all ages.  Covariates included
                   CO, H2S, NO2, O3, SO2, total reduced sulfur (TRS), a
                   large number of weather variables, and 12 molds and
                   pollens. Stats: generalized additive models with LOESS
                   prefiltenng of both ED and pollutant variables, with
                   variable window lengths.  Also controlled for day of
                   week and LOESS-smoothed functions of weather.
                   Single-day, and five day average, pollution lags tested
                   out to lag 10. The strongest lag, either positive or
                   negative, was chosen for final models. Both single and
                   multi-pollutant models reported.  Full-year and May-Sep
                   models reported.
In single-pollutant models, significant
positive associations observed between all
cardiac ED visits and PMI(), PM25, H,S, O3,
and SO2. Significant negative associations
observed with H+, sulfate, and COH max.
PM results were similar when data were
restricted to May-Sep.  In multi-pollutant
models, no PM metrics were significantly
associated with all cardiac ED visits in full
year analyses, whereas both O3 and SO2
were.  In the May-Sep subset, significant
negative association found for sulfate.  No
quantitative results presented for non-
significant variables in these multi-
pollutant regressions. In cause-specific,
single-pollutant models, PM tended to be
positively associated with
dysrhythmia/conductive disturbances but
negatively associated with congestive heart
failure (no quantitative results presented).
The objective decision rule used for
selecting lags reduced the risk of data
mining; however, the biological plausibility
of lag effects beyond 3-5 days is open to
question.  Some reported associations
likely to be spurious.
Percent Excess Risk (p-value)
computed for 50 ,ug/m3 PMH), 25
//g/m3 PM2 5 and  mean levels of
sulfate and COH.

Full year results for all cardiac
conditions, single pollutant
models:

PM10: 3d.
 32.5(10.2,59.3)
PM25:  3d.
 15.1 (-0.3,32.8)
H+:4-9d. avg.
 -1.8(0.010)
Sulfate: 4d.
 -6.0(0.001)
COH max: 7d.
 -5.4(0.027)

Full year results for all cardiac
conditions, multi-pollutant
models:

No significant PM associations.

-------
to
                        TABLE 6-16 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND CARDIOVASCULAR
                                                                    HOSPITAL ADMISSIONS
         Reference citation.
         Location, Duration
         PM Index, Mean or
         Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments.  Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
N)
         Europe

         Atkinson etal. (1999a)
         Greater London, UK
         1992-1994

         Pollutant: mean, median, 90-10
         percentile range:
         PM10 Cug/m3): 28.5, 24.8, 30.7
         Black Smoke G/g/m3): 12.7,10.!
         16.1
Daily emergency hospital admissions for total
cardiovascular diseases, CVD (1CD9 codes 390-459),
and ischemic heart disease, IHD (ICD9 410-414), for all
ages, for persons less than 65, and for persons 65 and
older.  Mean daily admissions for CVD: 172.5 all ages,
54.5 <65, 117.8 >65; for IHD: 24.5 <65, 37.6 >65.
Covanates: NO2, O3, SO2, CO, temperature, relative
humidity. Poisson regression using APHEA
methodology; sine and cosine functions for seasonal
control; day of week dummy variables. Lags of 0-3, as
well as corresponding multi-day averages ending on lag
0, were considered.
In single-pollutant models, both PM
metrics showed positive associations with
both CVD and IHD admissions across age
groups. Two-pollutant models were
mentioned, but quantitative results were
not given.
Effects computed for 50 /^g/
PM|(1 and 25 ^g/m3 BS
PM,0 0 d.
All ages:
CVD: 3.2(0.9,5.5)
0-64yr:
CVD: 5.6(2.0,9.4)
IHD:  6.8(1.3,12.7)
65+ yr:
CVD: 2. 5 (-0.2, 5.3)
IHD:  5.0(0.8,9.3)

Black Smoke 0 d.
All ages:
CVD: 2.95(1.00,4.94)
0-64yr:
CVD: 3.12(0.05,6.29)
IHD:  2.78 (-1.88, 7.63)
65+ yr:
CVD: 4.24(1.89,6.64)
IHD (lag 3): 4.57 (0.86, 8.42)
3
O
c
O
H
O
h-H
H
W

-------
CO
3
tr
K>
O
o
               TABLE 6-16 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND CARDIOVASCULAR
                                                            HOSPITAL ADMISSIONS
          Reference citation.
          Location, Duration
          PM Index, Mean or
          Median, IQR
                                 Study Description: Health outcomes or codes,
                                 Mean outcome rate, sample or population size,
                                 ages.  Concentration measures or estimates.
                                 Modeling methods- lags, smoothing, co-pollutants,
                                 covariates, concentration-response
                                                    Results and Comments.  Design Issues,
                                                    Uncertainties, Quantitative Outcomes
                                        PM Index, Lag, Excess
                                        Risk % (95% LCL, UCL),
                                        Co-Pollutants
K>
 H
 O
 O
 2
 o
 H
O
 c
 o
 H
 w
 o
 &
 n
 H
 tn
          Europe (cont'd)

          Prescott et al. (1998)
          Edinburgh, Scotland
          1981-1995(85 and SOZ)
          1992-1995 (PM,,,, NO2, O3, CO)
          Means for long and short series:
          BS: 12.3,8.7
          PM,0:  NA, 20.7
Wordleyetal. (1997)
Birmingham, UK
4/1/92-3/31/94
mean, min, max:
PMu,G/g/m3):  26,3, 131
Diaz etal. (1999)
Madrid, Spain
1994-1996

TSP by beta attenuation
Summary statistics not given.
                                 Daily emergency hospital admissions for cardiovascular
                                 disease (ICD9 codes 410-414, 426-429, 434-440) for
                                 persons less than 65 years and for persons 65 or older.
                                 Separate analyses presented for long (1981-1995) and
                                 short (1992-1995) series. Mean hospital admissions
                                 for long and short series: <65, 3.5, 3.4; 65+, 8.0, 8.7.
                                 Covariates: SO2, NO2, O3, CO, wind speed, temperature,
                                 rainfall. PM ,„ measured by TEOM. Stats: Poisson
                                 log-linear regression; trend and seasons controlled by
                                 monthly dummy variables over entire series; day of week
                                 dummy variables; min daily temperature modeled using
                                 octile dummies. Pollutants expressed as cumulative lag
                                 1-3 day moving avg.
Daily hospital admissions for acute ischemic heart
disease (ICD9 codes 410-429) for all ages.  Mean
hospitalizations: 25.6/day.  Covariates: temperature
and relative humidity.  Stats:  Linear regression with
day of week and monthly dummy variables, linear trend
term. Lags of 0-3 considered, as well as the mean of
lags 0-2.

Daily emergency hospital admissions for all
cardiovascular causes (ICD9 codes 390-459) for
the Gregorio Maranon University Teaching Hospital.
Mean admissions: 9.8/day. Covariates: SO2, NO2, O3,
temperature, pressure, relative humidity, excess sunlight.
Stats: Box-Jenkins time-series methods used to remove
autocorrelations, followed by cross-correlation analysis;
sine and cosine terms for seasonahty; details unclear.
                                                    In long series, neither BS nor NO2 were
                                                    associated with CVD admissions in either
                                                    age group. In the short series, only 3-day
                                                    moving average PM,0 was positively and
                                                    significantly associated with CVD
                                                    admissions in single-pollutant models,
                                                    and only for persons 65 or older. PM,0
                                                    effect remained largely unchanged when
                                                    all other pollutants were added to the
                                                    model (quantitative results not given).
No statistically significant effects observed
for PM10 on ischemic heart disease
admissions for any lag. Note that PM,0
was associated with respiratory admissions
and with cardiovascular mortality in the
same study (results not shown here).
No significant effects of TSP on CVD
reported.
Percent Excess Risk (95% CI):
Effects computed for 50 Mg/m3
change in PM1() and 25 /^g/m3
change in BS.

Long series:
BS, l-3d. avg.
<65: -0.5 (-5.4, 4.6)
65+: -0.5 (-3.8, 2.9)

Short series:
BS, l-3d. avg.
<65: -9.5 (-24.6, 8.0)
65+: 5.8 (-4.9, 17.8)

PM,0,l-3d. avg.
<65: 2.0 (-12.5, 19.0)
65+: 12.4(4.6,20.9)

% change (95% CI) per
50 Atg/m3 change PM,(I
IHD admissions:
PM,0 0-dlag:
  1.4% (-4.4, 7.2)
PM10 1-dlag:
  -1.3% (-7.1,4.4)

No quantitative results presented
for PM.

-------
I
g.
o
o
               TABLE 6-16 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND CARDIOVASCULAR
                                                             HOSPITAL ADMISSIONS
Reference citation.
Location, Duration
PM Index, Mean or
Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
                                                                                                Results and Comments. Design Issues,
                                                                                                Uncertainties, Quantitative Outcomes
                                        PM Index, Lag, Excess
                                        Risk % (95% LCL, UCL),
                                        Co-Pollutants
H
6
o
 H
O

 1
 O
 n
 h-H
 H
 W
          Australia
          Morgan etal. (1998)
          Sydney, Australia
          1990-1994

          mean, median, IQR, 90-10
          percenti le range:
          Daily avg. bscat/104m: 0.32, 0.26,
          0.23, 0.48
          Daily max l-hrbscat/104m:  0.76,
          0.57,60,1.23
Asia
Wong etal. (1999)
Hong Kong
1994-1995
median, IQR for PMU
45.0,34.8
                                  Daily hospital admissions for heart disease (ICD9 codes
                                  410, 413, 427,428) for all ages, and separately for
                                  persons less than 65 and persons 65 or greater. Mean
                                  daily admissions: all ages, 47.2; <65, 15.4; 65+, 31.8.
                                  PM measured by nephelometry (i.e., light scattering),
                                  which is closely associated with PM2 5.  Authors give
                                  conversion for Sydney as PM2 5 =30 x bscat.  Covariates:
                                  O3, NO2, temperature, dewpoint temperature. Stats:
                                  Poisson regression; trend and seasons controlled with
                                  linear time trend and monthly dummies; temperature
                                  and dewpoint controlled with dummies for eight  levels
                                  of each variable; day of week and holiday dummies.
                                  Single and cumulative lags from 0-2 considered.
                                  Both single and multi-pollutant models were examined.
Daily emergency hospital admissions for cardiovascular
diseases, CVD (ICD9 codes 410-417, 420-438,
440-444), heart failure, HF (ICD9 428), and ischemic
heart disease, IHD (ICD9 410-414) among all ages and
in the age categories 5-64, and 65+.  Median daily CVD
admissions for all ages: 101. Covariates: NO2, O3, SO2,
temperature, relative humidity. PM,,, measured by
TEOM. Stats: Poisson regression using the APHEA
protocol; linear and quadratic control of trends; sine and
cosine control for seasonality; holiday and day of week
dummies; autoregressive terms. Single and cumulative
lags from 0-5 days considered.
                                                    In single-pollutant models, NO2 was
                                                    strongly associated with heart disease
                                                    admissions in all age groups.  PM was
                                                    more weakly, but still significantly
                                                    associated with admissions for all ages
                                                    and for persons 65+. The NO2 association
                                                    in the 65+ age group was unchanged in
                                                    the multi-pollutant model, whereas the PM
                                                    effect disappeared when NO2 and O3 were
                                                    added to the model.. These results suggest
                                                    that PfA is not robustly associated with
                                                    heart disease admissions when NO2 is
                                                    included, similar to the sensitivity of PM
                                                    to CO in other studies.
In single-pollutant models, PM|0, NO2,
SO2, and O3 all significantly associated
with CVD admissions for all ages and
for those 65+. No multi-pollutant risk
coefficients were presented; however,
the PM,0 effect was larger when O3 was
elevated (i.e., above median).  A much
larger PM,0 effect was observed for HF
than for CVD or IHD. These results
confirm the presence of PM,0 associations
with cardiovascular admissions in single-
pollutant models, but do not address the
independent role of PM1(1.
                                        Percent Excess Risk (95% CI):
                                        Effects computed for 25 jug/m3
                                        PM2 5 (converted from bscat).

                                        24-hr avg. PM25Od.
                                          <65:  1.8 (-2.9, 6.7)
                                          65+: 4.9(1.6,8.4)
                                          All:   3.9(1.1,6.8)

                                        24-hr PM25, 0 d w. NO2 and O3.
                                          65+: 0.12 (-1.3, 1.6)

                                        l-hrPM25, Od.
                                          <65: 0.19 (-1.6, 2.0)
                                          65+:  1.8(0.5,3.2)
                                          All:   1.3(0.3,2.3)
Percent Excess Risk (95% CI):
Effects computed for 50 //g/m3
change in PMI0.

PM10, 0-2d. avg.

CVD:
  5-64:  2.5 (-1.5, 6.7)
  65+:  4.1(1.3,69)
  All:   3.0 (0.8, 5.4)

HF (PM,,,, 0-3 d ave.):
  All: 26.4(17.1,36.4)

IHD (PM,0, 0-3 d ave.):
  All: 3.5  (-0.5, 7.7)

-------
  1          For example, Schwartz (1999) extended the analytical approach he had used in Tucson
  2     (described below) to eight more U.S. metropolitan areas, limiting analyses to a single county in
  3     each location to enhance representativeness of the air pollution data. The locations analyzed
  4     were: Chicago, IL; Colorado Springs, CO; New Haven, CT; Minneapolis, MN; St. Paul, MN;
  5     Seattle, WA; Spokane, WA; and Tacoma, WA.  Again, the analyses focused on total
  6     cardiovascular (CVD) hospital admissions among persons >65 years old.  In univariate
  7     regressions, remarkably consistent PM10 associations with CVD admissions were found across
  8     the eight locations, with a 50 /ug/m3 increase in PM10 associated with 3.6 to 8.6% increases in
  9     admissions. The univariate eight-county pooled PM10 effect was 5.0% (CI 3.7-6.4), similar to the
 10     6.1 % effect per 50 ^g/m3 observed in the previous Tucson analysis.  In a bivariate model that
 11     included CO, the pooled PM10 effect size diminished somewhat to 3.8% (CI 2.0-5.5) and the CO
 12     association with CVD admissions was generally robust to inclusion of PM10 in the model.
 13          Additional new results have also recently been published for analyses of daily CVD
 14     hospital admissions in persons 65 and older in relation to PM10 for a subset of 14  cities from
 15     among the 90 cities evaluated in the NMMAPS  multi-city study (Samet et al., 2000a,b). Cities
 16     included Birmingham, AL; Boulder, CO; Canton, OH; Chicago, IL; Colorado Springs, CO;
 17     Detroit, MI; Minneapolis/ St. Paul, MN; Nashville, TN; New Haven, CT; Pittsburgh, PA;
 18     Provo/Orem, UT; Seattle, WA; Spokane, WA; and Youngstown, OH. The range  of years studied
 19     encompassed 1985-1994, although this varied by city.  Covariates included SO2, NO2, O3, and
 20     CO; however these were not analyzed directly as regression covariates.  Individual cities were
 21      analyzed first by Poisson regression methods on PM10 for lags  from 0 to 5 days. An overall risk
 22     estimate was then computed by taking the inverse-variance weighted mean of the  city-specific
 23      risk estimates. The city-specific risk estimates for PM10 were also examined for correlations with
 24      omitted covariates, including other pollutants. No relationship was observed between city-
 25      specific risk estimates and measures of socio-economic status,  including percent living in
 26      poverty, percent non-white, and percent with college educations.  The overall weighted mean risk
 27      estimate for PM>0 was greatest for lag 0 and for the mean of lags 0-1. For example, the mean risk
28      estimate for the mean of lags 0-1 was a 6.0% increase in CVD  admissions per 50 jUg/m3 PM10
29      (95% CI: 5.1- 6.8). The mean risk was larger in a subgroup of data where PM10 was less than
30      50 ywg/m3, suggesting the lack of a threshold. While no multi-pollutant results were presented,


        March 2001                              6-125       DRAFT-DO NOT QUOTE OR CITE

-------
 1     the authors argued that confounding was not present because the city-specific risk estimates did
 2     not correlate with city-mean co-pollutant levels.
 3           Zanobetti et al. (2000b), in an analysis of a subset of 10 cities from among the 14 evaluated
 4     by Samet et al. (2000a,b), did include other gaseous co-pollutants in their analyses of CVD
 5     hospital admissions for the elderly (>65 yr) during 1986-1994.  After analyzing single cities first,
 6     the 10 risk estimates were further analyzed in several second-stage analyses, combining risks
 7     across cities and regressing risks on potential risk modifiers and co-pollutant confounders.  The
 8     same basic pattern of results obtained by Samet et al. (2000a,b) were found, with strongest PM10
 9     associations on lag 0 day, smaller effects on lag 1 and 2, and none at longer lags. The cross-city
10     weighted mean estimate at 0 day lag was excess risk = 5.6% (95% CI 4.7, 6.4) per 50 jUg/m3
11     PMIO increment. The 0-1 day lag average excess CVD risk = 6.2% (95% CI 5.4, 7.0) per
12     50 yUg/m3 PM10 increment. Effect size estimates increased when data were restricted to days with
13     PM10 < 50 yWg/m3.  No evidence of gaseous (CO, O3, SO2) co-pollutant modification of PM
14     effects was seen in the second stage analyses.
15           Turning to some examples of independent single-city analyses, PM10 associations with
16     CVD hospitalizations were also examined in a study by Schwartz (1997), which analyzed three
17     years of daily data for Tucson,  AZ linking total CVD hospital admissions for persons >  65  years
18     old with PM10, CO, O3, and NO2. As was the above case in Chicago, only one site monitored
19     daily PM10 while multiple sites did so for gaseous pollutants. Both PM,0 and CO were
20     independently (i.e., robustly) associated with CVD-related admissions, whereas O3 and NO2 were
21     not. The percent effect of a 50 /ug/m3 increase in PMIO changed only slightly from 6.07 (CI 1.12 -
22      11.27) to 5.22 (CI  0.17 - 10.54) when CO was included in the model along with PM10.
23           Morris and Naumova (1998) also found associations with PM10 in their analysis of four
24     years of congestive heart failure data among people > 65 years old in Chicago, IL.  While the
25     analysis was directed primarily at evaluating modification by temperature of CO effects on
26     congestive heart failure admissions (building on previous results of Morris et al., 1995), results
27     were also presented for PM)0, as well  as for O3, NO2, and SO2. As many as eight monitoring sites
28     were available for calculating daily gaseous pollutant concentrations; however, only one site in
29     Chicago monitored daily PM10. Only same-day results were presented, based on an initial
30     exploratory analysis showing strongest effects for same-day pollution exposure (i.e., lag 0).
31     Strong and robust  associations were seen  between congestive heart failure admissions and CO.

       March 2001                               6-126        DRAFT-DO NOT QUOTE OR CITE

-------
  1     Associations between hospitalizations and PM10 were also observed in univariate regressions
  2     3.-9% (1.0, 6.9) per 50 >ug/m3 PMJO increase), but these diminished somewhat in a multi-pollutant
  3     model (2.0%, CI -1.4, 5.4). Although these results may suggest a more robust association with
  4     CO than with PM10, the observed differences might be explained by differential exposure
  5     misclassification for PM,0 (monitored at one site) as compared with CO (eight sites).
  6          More recently, Lippmann et al. (2000) reported findings from analyses of relationships
  7     between PM10, PM2 5, or PM10.2 5 and various categories of CVD hospital admissions among the
  8     elderly (65+ yr) in Detroit during 1992-1994.  The most striking findings were notable percent
  9     excess risk for: (a) ischemic heart disease (IHD) in relation to PM indices, i.e. 8.9% (0.5, 18.0)
 10     per 50 ^g PM10; 10.5% (2.8, 18.9)  per 25 ^g/m3 PM10_25; and 4.3% (-1.4, 10.4) per 25 //g/m3
 11     PM25 (all at lag 2d); and (b) heart failure, i.e. 9.7% (0.2, 20.1) per 50 Mg/m3 PM10; 5.2% (-3.3,
 12     14.5) per 25 .ug/m3 PM]0.2 5; and 9.1% (2.4, 6.2) per 25 Aig/m3 PM2 5 (the first two at lag 0 d and
 13     the latter at lag 1 d). As discussed earlier with regard to Lippmann et al. (2000) mortality
 14     findings, it is difficult to discern whether the observed associations with coarse fraction particles
 15     (PMio-2 s) are independently due to  such particles or may possibly be attributed to the moderately
 16     correlated fine particle (PM25) fraction in Detroit.
 17          Tolbert et al. (2000a) also recently reported preliminary results of analyses  of daily hospital
 18     emergency department visits for dysrhythmias (DYS) and all CVD categories for persons aged
 19     16 yrs or older, based on analyses of data from 18 of 33 participating hospitals in Atlanta.
 20     During Period 1 of the study (1993-1998), PM10 from the EPA AIRS database was reported to be
 21      negatively associated with CVD visits. During the second period (Aug. 1998 - Aug. 1999) based
 22     on use of Atlanta supersite data, however, the effect estimate for CVD excess risk of 5.1% per
 23      50 //g/m3 PMIO (although not statistically significant) does comport well with those noted above
 24      from the Samet and Schwartz studies. Also, although Period 2 CVD estimates (excess risk =
 25      6.1 %; CI - 3.1, 16.2) per 50 /wg/m3 PM2 5 were not significant at p < 0.05, positive associations
 26      with certain fine particle components, i.e., elemental carbon (p < 0.005) and organic carbon
 27      (p < 0.02), were significant along with CO (p < 0.005). Significant positive associations were
 28      also found for DYS visits versus elemental carbon (p = 0.004), coarse particles (p < 0.04), and
29      CO (p < 0.005). However, much caution applies to acceptance of the Tolbert et al.  (2000a)
30      findings until more complete analyses from all participating hospitals are carried  out and
31      reported.

        March 2001                               6-127       DRAFT-DO NOT QUOTE OR CITE

-------
  1           As for other U.S. studies of cardiovascular morbidity in single cities or urban counties, in
  2      an analysis of 1992-1995 Los Angeles data, Linn et al. (2000) also found that PM10, CO, and NO2
  3      were all significantly associated with increased cardiovascular admission in single-pollutant
  4      models among persons aged 30 yr and older, but no multi-pollutant modeling results were
  5      presented. Lastly, Moolgavkar (2000b) analyzed PM10, CO, NO2, O3, and SO2 in relation to daily
  6      total cardiovascular (CVD) and total cerebrovascular (CrD) admissions for persons aged
  7      >65 from three urban counties (Cook, IL; Los Angeles, CA; Maricopa, AZ). In univariate
  8      regressions, PM)0 (and PM2 5 in LA) were associated at some lags with CVD admissions in Cook
  9      and LA counties, but not in Maricopa county.  In two-pollutant models in Cook and LA counties,
10      the PM risk estimates diminished and/or were rendered non-significant.
11           The above analyses of daily PM10 and CO in U.S. cities, overall, indicate that elevated
12      concentrations of both PMIO and CO may enhance risk of CVD-related morbidity leading to acute
13      hospitalizations.  The Lippmann results appear to implicate PM2 5 and/or PM10_2 5 in increased
14      hospital admissions for some categories of CVD among the elderly; and the Tolbert results very
15      preliminarily also hint at both fine and coarse particle components contributing to CVD-related
16      emergency department visits.
17           Four separate analyses of hospitalization data in Canada have been reported by Burnett and
18      coworkers since 1995 (Burnett et al., 1995, 1997b,c, 1999). A variety of locations, outcomes,
19      PM exposure metrics, and analytical approaches were used in these studies, which hinders
20      somewhat the ability to draw broad, conclusions across the full group.  The first (Burnett et al.,
21      1995), reviewed briefly in the 1996 PM AQCD, analyzed six years of data from 168 hospitals in
22      Ontario, CN. Cardiovascular and respiratory hospital admissions were analyzed in relation to
23      sulfate and ozone concentrations.  Sulfate lagged one day was associated with CVD admissions,
24      with a percent effect of 2.8 (CI1.8-3.8) per 13  jug/m3 without O3 in the model and 3.3
25      (CI 1.7-4.8) with O3  included.  When CVD admissions were split out into sub-categories, larger
26      associations were seen between sulfates and coronary artery disease and heart failure than for
27      cardiac dysrhythmias. Sulfate associations with total admissions  were larger for the elderly
28      sub-population > 65  (3.5% per 13 /wg/ni3) than for those <65 years old (2.5% per 13 /wg/m3).
29      There was little evidence for seasonal differences in sulfate associations.
30           Burnett et al. (1997a) analyzed daily congestive heart failure hospitalizations in relation to
31      carbon monoxide and other air pollutants (O3, NO2, SO2, COH) in ten large Canadian cities as a

        March 2001                              6-128       DRAFT-DO NOT QUOTE OR CITE

-------
  1     replication of an earlier U.S. study by Morris et al. (1995).  The Burnett Canadian study
  2     expanded upon the previous work both by its size (11 years of data for each of 10 large cities)
  3     and also by including a measure of PM air pollution (coefficient of haze, COH), whereas no PM
  4     data were included in the earlier Morris et al. study. The Burnett study was restricted to the
  5     population > 65 years old. The authors noted that all pollutants except O3 were correlated,
  6     making it difficult to separate them statistically.  COH, CO, and NO2 measured on the same day
  7     as admission (i.e., lag 0) were all strongly associated with congestive heart failure admissions in
  8     univariate models. In multi-pollutant models, CO remained a strong predictor, whereas COH did
  9     not (gravimetric PM measures were not evaluated).
 10          The roles played by size-selected gravimetric and chemically speciated particle metrics as
 11     predictors of CVD hospitalizations were explored in analysis of data from metropolitan Toronto
 12     for the summers of 1992-1994 (Burnett etal., 1997b). The analysis used dichotomous sampler
 13     (PM2 5, PM10, and PM10_2 5), hydrogen ion, and sulfate data collected at a central site as well as O3,
 14     NO2, SO2, CO, and COH data collected at multiple sites in Toronto. Hospital admissions
 15     categories included total cardiovascular (i.e., the sum of ischemic heart disease, cardiac
 16     dysrhythmias, and heart failure) and total respiratory.  Model specification with respect to
 17     pollution lags was completely data-driven, with all lags and averaging times out to 4 days prior to
 18     admission evaluated in exploratory analyses and "best" metrics chosen on the basis of maximal
 19     t-statistics. The relative risks of CVD admissions were positive and generally statistically
 20     significant for all pollutants analyzed in univariate regressions, but especially so for O3, NO2,
 21      COH, and PM10_2 s (i.e., regression t-statistics > 3).  Associations for gaseous pollutants were
 22     generally robust to inclusion of PM covariates, whereas the PM  indices (aside from COH) were
 23      not robust to inclusion of multiple gaseous pollutants.  In particular, PM25 was not a robust
 24      predictor of CVD admissions in multi-pollutant models: whereas an 25 jwg/m3 increase in PM2 5
 25      was associated with a 5.9% increase (t=l  .8) in CVD admissions in a univariate model, the
 26      percent effect was reduced to -1.1 (t=0.3) in a model that included O3,  NO2, and SO2.  COH, like
 27      CO and NO2, is generally thought of as a  measure of primary motor-vehicle emissions during the
 28      non-heating season. The authors concluded that "particle mass and chemistry could not be
29      identified as an independent risk factor for exacerbation of cardiorespiratory diseases in this
30      study beyond that attributable to climate and gaseous air pollution."


        March 2001                               6-129        DRAFT-DO NOT QUOTE OR CITE

-------
 1           Burnett et al. (1999) later reported results of a more extensive attempt to explore cause-
 2      specific hospitalizations for persons of all ages in relation to a large suite of gaseous and PM air
 3      pollutant measures, using 15 years of Toronto data. Cardiovascular admissions were split out
 4      into separate categories for analysis:  dysrhythmias, heart failure, and ischemic heart disease.
 5      The analyses also examined several respiratory causes, as well as cerebrovascular and diseases of
 6      the peripheral circulation (the latter categories being included because they should show PM
 7      associations if one mechanism of PM action is related to increased plasma viscosity, as suggested
 8      by Peters et al., 1997a). The PM metrics analyzed were PM2 5, PM10, and PM10.2 5 estimated from
 9      daily TSP and TSP sulfate data, based on a regression analysis on dichotomous sampling data
10      that were available every sixth day during an eight-year subset of the full study period. This use
11      of estimated rather than measured PM components limits the interpretation of the PM results
12      reported here. In general, use of estimated PM exposure metrics will tend  to increase exposure
13      measurement error and thereby tend to decrease effects estimates. Model specification for lags
14      was again data-driven based on maximal t-statistics. Although some statistically significant
15      associations with one or another PM metric were found in univariate models, there were no
16      significant PM associations with any of the three CVD hospitalization outcomes in multi-
17      pollutant models. For example, whereas an 25 jug/m3 increase in estimated PM2 5 was associated
18      with a 8.05% increase (t-statistic = 6.08) in ischemic heart disease admissions in a univariate
19      analysis, the PM2 5 association was reduced to 2.25% (n.s.) when NO2 and  SO2 were included in
20      the model.  The gaseous pollutants dominated most regressions. There also were no associations
21      between PM and cerebral or peripheral vascular disease admissions.
22           The Burnett et al. studies provide some of the most extensive results  for PM in conjunction
23      with multiple gaseous pollutants, but the inconsistent use of alternative PM metrics in the various
24      analyses confuses the picture somewhat. A general finding appears to be lack of robustness of
25      associations between cardiovascular outcomes and PM in multi-pollutant analyses.  This was
26      seen for COH in the analysis of 10 Canadian cities (Burnett et al., 1997a),  for PM2 5 and PM10 in
27      the analysis of summer data in Toronto (Burnett et al., 1997b), and for linear combinations of
28      TSP and sulfates (i.e., estimated PM25, PM,0, and PM10_25) in the analysis of 15 years of data in
29      Toronto (Burnett et al., 1999). One exception was the association reported between CVD
30      admissions to 168 Ontario hospitals and sulfate concentrations (Burnett et  al., 1995), where the
31      sulfate association was robust to the inclusion of O3. Also, although gravimetric PM variables

        March 2001                               6-130        DRAFT-DO NOT QUOTE OR CITE

-------
  1      were not robust predictors in the Toronto summer analysis, COH was (Burnett et al., 1997b),
  2      perhaps reflecting the impact of primary motor vehicle emissions.  This contrasts, however, with
  3      COH's lack of robustness in the 10-city analysis (Burnett et al., 1997a). It is difficult to
  4      determine how much weight should be ascribed to the finding of lack of robustness of PM versus
  5      gaseous pollutants effects, given the expected large measurement error likely associated with the
  6      estimated PM metrics.
  7           Several pertinent new European studies, mainly in the U.K., have also been published since
  8      the 1996 PM AQCD. For example, Atkinson et al. (1999b) reported significant associations of
  9      both ambient PM10 and black smoke (BS) with daily admissions for total cardiovascular disease
 10      and ischemic heart disease for  1992-1994 in London, UK, using standard time series regression
 11      methods. Associations were observed for persons aged < 65 yr and for persons aged > 65 yr.
 12      While the authors mention analysis of co-pollutants, no quantitative results were presented for
 13      multi-pollutant models. Also, associations with PM10, but not with BS were reported for
 14      analyses of daily emergency hospital admissions for cardiovascular diseases from 1992-1995 for
 15      Edinburgh, UK (Prescott et al., 1998).  Associations were present only in persons 65 and older.
 16      The authors reported that the PM10 associations were unaffected by inclusion of other pollutants;
 17      however, results were not shown.  Standard time series regression methods were used.  PM10was
 18      measured by TEOM.  On the other hand, no associations between PM10 and daily ischemic heart
 19      disease admissions were observed by Wordley and colleagues (1997) in an analysis of two years
20      of daily data from Birmingham, UK. However, PM10 was associated with respiratory admissions
21      and cardiovascular mortality during the  same study period.  The inconsistency of results across
22      causes and outcomes is difficult to interpret, but may relate in part to the relatively short time
23      series analyzed. The authors stated that gaseous pollutants did not  have significant associations
24      with health outcomes independent of PM, but no results were presented for models involving
25      gaseous pollutants.
26           Also relevant to the present review of associations between acute cardiovascular morbidity
27      and PM are nine recent studies  of acute cardiovascular (CVD) mortality (Borja-Aburto et al.,
28      1997, 1998; Michelozzi et  al., 1998; Morgan et al., 1998; Ponka et  al., 1998; Schwartz et al.,
29      1996a; Simpson et al., 1997; Wordley et al., 1997; Zmirou et al., 1998) reviewed earlier in
30      Section 6.2.2. Acute mortality can be viewed as a more severe manifestation of the same
31      pathophysiologic mechanism responsible for acute hospital admissions following PM exposure.

        March 2001                              6-131       DRAFT-DO NOT QUOTE OR CITE

-------
 1      All nine studies reported significant associations between acute CVD mortality and measures of
 2      ambient PM, though the PM metrics utilized and the relative risk estimates varied across studies.
 3      PM measurement methods included gravimetrically analyzed filter samples (TSP, PM10, PM2 5,
 4      PM10_2 5), beta gauge (particle attenuation of beta radiation), nephelometry (light scattering), and
 5      black smoke (filter reflectance). Where tested, PM associations with acute CVD mortality
 6      appeared to be generally more robust to inclusion of gaseous covariates than was the case for
 7      acute hospitalization studies (Borja-Aburto et al., 1997, 1998; Morgan et al., 1998; Wordley
 8      et al., 1997; Zmirou et al., 1998). One study which examined multiple alternative PM metrics
 9      reported strongest associations with PM2 5 and no associations for PM10_2 5 and hydrogen ion.
10      These results for acute cardiovascular mortality are qualitatively consistent with those reviewed
11      above for hospital admissions.
12           One additional recent study of acute PM exposure impacts on CVD mortality is of interest
13      here. Checkoway et al. (2000) studied the possible association between occurrence of out-of-
14      hospital sudden cardiac arrest (SCA) and daily PM levels in the Seattle metropolitan area as
15      measured both by nephelometry and PM10 from three monitoring sites.  A case-crossover study
16      evaluated 362 SCA cases identified from October 1988 through June 1994. The cases had no
17      prior history of clinically recognized heart disease or other life-threatening conditions. Lag
18      periods for index days of 0 to 5 days were studied.  There was no evidence of confounding by
19      ambient daily exposure to CO or SO2.  Relative risk estimates for SCA showed no evidence of an
20      association of 24-h PM levels with increased risks at any lag time studied. The notable strength
21      of this study was the availability of personal risk factor information.  The authors provided
22      interpretation of the null results, posing the possibilities that: (1) this case group had a low
23      prevalence of previously detected compromised cardiovascular health (inclusion of persons with
24      prior history of cardiovascular disease should be studied); (2) PM exposures in the Seattle area
25      may be too low to cause an effect; (3) potential for exposure misclassification exists, and (4) the
26      sample size was not large enough to either find or rule out a relative risk less than 1.5. One
27      important comment by the HEI Health Review Committee attached as part of the report was that
28      the induction time for the occurrence of primary arrhythmia may be much shorter than the 24-h
29      time window studied and, as such, other hazard periods (i.e., 1, 2, or  4-h concentrations) might
30      be of more interest for evaluating PM effects on sudden cardiac arrest.


        March 2001                               6-132       DRAFT-DO NOT QUOTE OR CITE

-------
  1           Figure 6-6 illustrates excess risk estimates derived from United States studies of PMIO
  2     exposure and cardiovascular hospitalizations, standardized to a 50 //g/m3 exposure to PM10.
  3     Results from available studies show both pooled outcomes for total CVD hospital admissions
  4     and studies presenting single U.S. cities.  The Samet et al. (2000b) pooled cross-city results for
  5     14 U.S. Cities and Schwartz (1999) results for 8 U.S. cities likely provide the most precise
  6     estimates for relationships of U.S. ambient PM10 exposure to increased risk for cardiovascular
  7     disease hospitalization.  Those estimates, and those derived from most other studies depicted in
  8     Figure 6-6, generally appear to confirm likely excess risk of CVD-related hospital admissions for
  9     U.S. cities  in the range of 3-10% per 50 /ug/m3 PM10, especially among the elderly (>65 yr).
 10     Also, other individual-city results from Detroit are indicative of excess risk for ischemic heart
 11     disease and heart failure in the range of ca. 4.0 to 10.0% per 25 yag/m3 of PM2 5 or PM10.2 5, as are
 12     preliminary individual-city findings from Atlanta suggestive of ca. 4.3% and 10.5% excess risk
 13     per 25 yug/m3 of PM2 5 and PM10.2 5, respectively.
 14
 15     6.3.1.3.2 Individual-Level Studies of Cardiovascular Physiology
 16          New  studies carried out by various groups  have evaluated longitudinal associations
 17     between ambient PM and physiologic measures of cardiovascular function. In contrast to the
 18     ecologic time-series studies discussed above, these studies measure outcomes and most
 19     covariates at the individual level, making it possible to draw conclusions regarding individual
 20     risks, as well as to explore mechanistic hypotheses. Heterogeneity of responses across
 21      individuals, and across subgroups defined on the basis of age, sex, pre-existing health status, etc.,
 22     can be assessed.  While exposure assessment remains largely ecologic (i.e., the entire population
 23      is usually assigned the same exposure value on a given day), exposure is generally well
 24      characterized in the small, spatially-clustered study populations.  The recent studies fall  into two
 25      broad classes:  those addressing cardiac rhythm, and those addressing blood characteristics.
26      While significant uncertainty still exists regarding the interpretation of results from these new
27      studies, the varied responses that have been reported to be associated with ambient PM and
28      co-pollutants are of much interest in regard to mechanistic hypotheses  concerning
29      pathophysiologic processes potentially underlying CVD-related mortality/morbidity effects
30      discussed in preceding sections.


        March 2001                               6-133        DRAFT-DO NOT QUOTE OR CITE

-------
             Samet et al. (2000) -
                14US Cities

               Schwartz (1999) -
                 8 US Cities

             Moolgavkar (2000b) -
               Maricopa, AZ

             Moolgavkar (2000b) -
                 LA.CA

             Moolgavkar (2000c) -
               Cook County

               Linn et al. (2000) -
                  LA.CA

               Schwartz (1997) -
                 Tucson.AZ

            Tolbert et al. (2000a) -
                 Atlanta
       Morris and Naumova (1998) -
              Chicago

           Lippmann et al (2000) -
Total CVD
       Period 1 (AIRS Data)
      	•	1
                         CHF
                                 i   »    i
                                Period 2 (Supersite Data)
                          HFU
                          IHD!"-
                                -15       -10       -50         5         10
                                       Reconstructed Excess Risk Percentage
                                                50 ug/m3 Increase in
      Figure 6-6.  Acute cardiovascular hospitalizations and participate matter exposure excess
                  risk estimates derived from selected U.S. PM,0 studies.  CVD = cardiovascular
                  disease. CHF = congestive heart failure.
1      Heart Rate Rhythm and Variability

2           Alterations in heart rate and/or rhythm have been hypothesized as possible mechanisms by

3      which ambient PM exposures may exert acute effects on human health. Decreased heart rate

4      variability, in particular, has been identified as a predictor of increased cardiovascular morbidity

5      and mortality (see Appendix 6B). Several independent studies have recently reported temporal

6      associations between PM exposures and various measures of heart beat rhythm in panels of
       March 2001
               6-134
DRAFT-DO NOT QUOTE OR CITE

-------
  1      elderly subjects (Liao et al., 1999; Pope et al., 1999a,b,c; Dockery et al., 1999; Peters et al.,
  2      1999a, 2000a; Gold etal., 1998; 2000).
  3           Liao and colleagues (1999) studied 26 elderly subjects (age 65-89 years; 73% female) over
  4      three consecutive weeks at a retirement center in metropolitan Baltimore, 18 of whom were
  5      classified as "compromised" based on previous cardiovascular conditions (e.g., hypertension).
  6      Daily six-minute resting electrocardiogram (ECG) data were collected, and time intervals
  7      between sequential R-R intervals recorded.  A Fourier transform was applied to the R-R interval
  8      data to separate its variance into two major components: low frequency (LF, 0.04-0.15 Hz) and
  9      high frequency (HF, 0.15-0.40 Hz).  The standard deviation of all normal-to-normal (N-N; also
 10      designated R-R) heartbeat intervals (SDNN) was computed for use as a time-domain outcome
 11      variable.  PM2 5  was monitored indoors by TEOM and outdoors by dichotomous sampler.
 12      Outdoor PM2 5 levels ranged from 8.0 to 32.2 yag/m3 (mean = 16.1 /wg/m3). Regression analyses
 13      controlled for inter-subject differences in average variability, allowing each subject to serve as
 14      his/her own control. Consistent associations were seen between decreases in all three outcome
 15      variables (LF, HF, SDNN) and increases in PM2 5 concentrations (both indoors and outdoors),
 16      with associations being stronger for the 18 "compromised"subjects. No analyses of heart rate
 17      were reported.
 18           Pope and colleagues (1999c) reported similar findings in a panel of six elderly subjects
 19      (69-89 years, 5/6 male) with histories of cardiopulmonary disease, and one 23-year old male
 20      subject suffering from Crohn's disease and arrhythmias. Subjects carried Holter monitors for up
 21      to 48 hours during different weeks that varied in ambient PM10 concentrations. N-N heartbeat
 22      intervals were recorded and used to calculate several measures of heart rate variability in the time
 23      domain: the standard deviation of N-N intervals (SDNN), which is a broad measure of both high
 24      and low frequency variations; the standard deviation of the averages of N-N intervals in all five
 25      minute segments (SDANN), which is a measure of ultra-low frequency variations; and the root
26      mean squared differences between adjacent N-N intervals (r-MSSD), which is a measure of high
27      frequency variations. Daily gravimetric PM10 data obtained from three sites in the study area
28      ranged from circa 10 /ug/m3 to 130 ,ug/m3 during the study. A simple step function in
29      concentration was observed with high levels occurring only during the first half of the 1.5 month
30      study period. Regression analysis with subject-specific intercepts was performed, with and
31      without control for daily barometric pressure and mean heart rate.  Same-day, previous-day, and

        March 2001                               6-135       DRAFT-DO NOT QUOTE OR CITE

-------
  1      the two-day mean of PM10 were considered. SDNN and SDANN were negatively associated with
  2      both same-day and previous-day ambient PM10, and results were unaffected by inclusion of
  3      covariates. Heart rate, as well as r-MSSD, were both positively, but less strongly, associated
  4      with PM10. No co-pollutants were studied.
  5           The Pope et al. (1999c) study discussed above was nested within a larger cohort of
  6      90 subjects who participated in a study of heart rate and oxygen saturation in the Utah Valley
  7      (Dockery et al., 1999; Pope et al., 1999b). The investigators hypothesized that decreases in
  8      oxygen saturation might occur as a result of PM exposure, and that this could be a risk factor for
  9      adverse cardiac outcomes. The study was carried out in winter months (mid November through
10      mid-March), when frequent inversions lead to fine particle episodes.  PM10 levels at the three
11      nearest sites averaged from 35 to 43 /ug/m3 during the study, with daily 24-h levels ranging from
12      5 to 147 |Wg/m3. Two populations were studied: 52 retired Brigham Young University
13      faculty/staff and their spouses, and 38 retirement home residents. Oxygen saturation (SpO2) and
14      heart rate (HR) were measured once or twice daily by an optical sensor applied to a finger.
15      In regression analyses that controlled for inter-individual differences in mean levels, SpO2 was
16      not associated with PM10, but was highly associated with barometric pressure. In contrast, HR
17      significantly increased in association with PM10 and significantly decreased in association with
18      barometric pressure in joint regressions. Including CO in the regressions did not change these
19      basic findings.  This was the first study of this type to examine the interrelationships among
20      physiologic measures (i.e., SpO2 and HR), barometric pressure, and PM10. The profound
21      physiological effects of barometric pressure noted here highlight the importance of carefully
22      controlling for barometric pressure effects in studies of cardiac physiology.
23           Gold and colleagues (1998, 2000) obtained somewhat different results in a study of heart
24      rate variability among 21 active elderly subjects, aged 53-87 yr, in a Boston residential
25      community.  Resting, standing, exercising, and recovering ECG measurements were performed
26      weekly using a standardized protocol on each subject, which involved 25 min/week of
27      continuous Holter ECG monitoring. Two time-domain measures were extracted:  SDNN  and
28      r-MSSD (see above for definitions). Heart rate also was analyzed as an outcome. Continuous
29      PM10 and PM2 5 monitoring was conducted by TEOM at a site 6 km from the study site, with PM
30      data corrected for loss of semivolatile mass.  Data on CO, O3, NO2, SO2, temperature and relative
31      humidity were available from nearby sites. Outcomes were regressed on PM2 5 levels in the

        March 2001                               6-136       DRAFT-DO NOT QUOTE OR CITE

-------
  1      0-24 hour period prior to ECG testing, with and without control for HR and temperature. As for
  2      the other studies discussed above, declines in SDNN were associated with PM2 5 levels, in this
  3      case averaged over 4 hours. These associations reached statistical significance at the 0.05 level
  4      only when all testing periods (i.e., resting, standing, exercise) were combined. In contrast to the
  5      above studies, both HR and r-MSSD here were negatively associated with PM2 5 levels (i.e.,
  6      lower HR and r-MSSD) when PM2 5 was elevated. These associations were statistically
  7      significant overall, as well as for several of the individual testing periods, and were unaffected by
  8      covariate control.
  9           Peters and colleagues (1999a) reported HR results from a retrospective analysis of data
 10      collected as part of the MONICA study (monitoring of trends and determinants in cardiovascular
 11      disease) carried out in Augsburg,  Germany. Analyses focused on 2,681 men and women aged
 12      25-64 years who had valid ECG measurements taken in winter 1984-1985 and again in winter
 13      1987-1988. Ambient pollution variables included TSP, SO2, and CO. The earlier winter included
 14      a 10-day episode with unusually high levels of SO2 and TSP, but not of CO. Pollution effects
 15     were analyzed in two ways: dichotomously comparing the episode and non-episode periods, and
 16     continuously using regression analysis. However, it is unclear from the report to what extent the
 17     analyses reflect between-subject vs. within-subject effects. A statistically significant increase in
 18     mean heart rate was observed during the episode period versus other periods, controlling for
 19     cardiovascular risk factors and meteorology.  Larger effects were observed in women.  Insingle-
 20     pollutant regression analyses, all three pollutants were associated with increased HR.
 21           In another retrospective study, Peters and colleagues (2000a) examined incidence of cardiac
 22     arrhythmias among 100 patients (mean age 62.2 yr.; 79%  male) with implanted cardioverter
 23      defillibrators followed over a three year period.  PM2 5 and PM,0 were measured in South Boston
 24     by the TEOM method, along with black carbon,  O3, CO, temperature and relative humidity; SO2
 25      and NO2 data were obtained from  another site. The 5th percentile, mean, and 95th percentiles of
 26      PM10 concentrations were 7.8, 19.3, and 37.0 Atg/m3, respectively.  The corresponding values for
 27      PM2 5 were 4.6, 12.7, and 26.6 /ug/m3.  Logistic regression was used to analyze arrhythmia events
 28      in relation to pollution variables, controlling for between-person differences, seasons, day-of-
29      week, and meteorology in two subgroups: 33 subjects with at least one arrhythmia event; and
30      6 subjects with 10  or more arrhythmia events. In the larger subgroup, only NO2 on the previous
31      day, and the mean  NO2 over five days, were significantly associated with arrhythmia incidence.

        March 2001                               6-137       DRAFT-DO NOT QUOTE OR CITE

-------
 1      In patients with 10 or more events, the NO2 associations were stronger. Also, some of the PM2 5
 2      and CO lags became significant in this subgroup.  These results should be interpreted cautiously
 3      given the large number of statistical tests performed.
 4           The above studies present a range of intriguing findings suggesting possible effects of PM
 5      on cardiac rhythm.  Three independent studies reported decreases in HR variability associated
 6      with PM  in elderly cohorts, although r-MSSD (a measure of high-frequency HR variability)
 7      showed elevations with PM in one study (Pope et al., 1999a). Also, all of the studies which
 8      examined HR found an association with PM; most reported positive associations, whereas one
 9      (Gold et al., 2000) reported a negative relationship.  However, variations in methods and results
10      across the studies argue for caution in drawing strong conclusions regarding PM effects from
11      them, especially in light of the complex intercorrelations which exist among measures of cardiac
12      physiology, meteorology, and air pollution (Dockery et al., 1999).
13
14      Viscosity and Other Blood Characteristics
15           Peters et al. (1997a) state that plasma viscosity is determined by fibrinogen and other large
16      asymmetrical plasma proteins such as immunoglobulin M and <*2-macroglobulin. They note that
17      in a cohort study of elderly men and women, fibrinogen concentrations were strongly related to
18      inflammatory markers such as neutrophil count and acute-phase proteins, (C-reactive protein and
19      ^-antichymotrypsin) and to self-reported infections. Fibrinogen  contributes to plasma viscosity,
20      which is a risk factor for ischemic heart disease.
21           Support for a mechanistic hypothesis, relating to enhanced blood viscosity, is suggested in
22      a recent analysis of plasma viscosity data collected in a population of 3256 German adults in the
23      MONICA study (Peters et al., 1997a).  Each subject provided one blood sample during October
24      1984 to June 1985.  An episode of unusually high air pollution concentrations occurred during a
25      13 day period while these measurements were being collected.  The authors reported that, among
26      the 324 persons who provided blood during the episode, there was a statistically significant
27      elevation in plasma viscosity as compared with the 2932 persons  studied at other times. The
28      odds ratio for plasma viscosity exceeding the 95th percentile was  3.6 (CI 1.6-8.1) among men
29      and 2.3 (CI 1.0-5.3) among women. Analysis of the distribution  of blood viscosity data
30      suggested that these findings were driven by changes in the upper tail of the distribution rather


        March 2001                              6-138         DRAFT-DO NOT QUOTE OR CITE

-------
  1     than by a general shift in mean viscosity, consistent with the likelihood of a susceptible
  2     sub-population of individuals.
  3           Peters et al. (2000b) reported on a prospective cohort study of a subset of male participants
  4     from the above-described Augsburg, Germany MONICA study. Based on a survey conducted in
  5     1984/85, a sample of 631 randomly selected men/aged 45-64 yr), free of cardiovascular disease at
  6     entry, were evaluated in a 3-yr follow-up that examined relationships of air pollution to serum
  7     C-reactive protein concentrations. C-reactive protein is a sensitive marker of inflammation,
  8     tissue damage, and infections, with acute and chronic infections being related to coronary events,
  9     as well as inflammation being related to systemic hypercoagulability and the onset of acute
 10     ischemic syndromes. During the 1985 air pollution episode affecting Augsburg and other parts
 11     of Germany, the odds of abnormal increases in serum C-reactive protein (i.e., >90th percentile of
 12     pre-episode levels = 5.7 mg/L) tripled and associated increases in TSP levels of 26 //g/m3 (5-day
 13     averages) were associated with an odds ratio of 1.37 (95% CI1.08-1.73) for C-reactive protein
 14     levels exceeding the 90th percentile levels in two pollutant models also including SO2 levels.
 15     The estimated odds ratio for a 30 ,ag/m3 increase in the 5-day mean for SO2 was 1.12 (95% CI
 16     0.92 to 1.47; non-significant).
 17           Two other recent studies also examined blood indices in relation to PM pollution (Seaton
 18     et al., 1999; Prescott et al.,  1999). Seaton and colleagues collected sequential blood samples (up
 19     to 12) over an 18 month period in 112 subjects (all over age 60) in Belfast and Edinburgh, UK.
 20     Blood samples were analy/ed for hemoglobin, packed cell volumes, blood counts, fibrinogen,
 21      factor VII, interleuken 6, C-reactive protein. In a subset of 60 subjects, plasma albumin also was
 22      measured.  PM10 data monitored by TEOM were collected from ambient sites in each city.
 23      Personal exposure estimates for the three days preceding each blood draw were derived from
 24      ambient data adjusted by time-activity patterns and I/O penetration factors. No co-pollutants
 25      were analyzed. Data were analyzed by analysis of covariance, controlling for city, seasons,
 26      temperature, and between-subject differences.  Significant changes in several of the blood indices
27      were observed in association with either ambient or estimated personal PM10 levels. All changes
28      were negative, except for C reactive protein in relation to ambient PM10, which  was positive.
29           Prescott et al. (1999) also investigated factors  that might increase susceptibility to adverse
30      cardiovascular events resulting from PM exposure.  Using data from a cohort of 1592 subjects
31      aged 55-74 in Edinburgh, UK, baseline measurements of blood fibrinogen and blood and plasma

        March 2001                              6-139        DRAFT-DO NOT QUOTE OR CITE

-------
 1      viscosity were examined as modifiers of the effects of PM (indexed by BS) on the incidence of
 2      fatal and non-fatal myocardial infarction or stroke.  All three blood indices were strong predictors
 3      of increased cardiac event risk.  However, there was no clear evidence of either a main effect of
 4      BS, nor interactions between BS and blood indices.
 5           The above findings add support for some intriguing hypotheses regarding possible
 6      mechanisms by which PM exposure may be linked with adverse cardiac outcomes.  They are
 7      especially interesting in terms of implicating both increased blood viscosity and C-reactive
 8      protein, a biological marker of inflammatory responses thought to be predictive of increased risk
 9      for serious cardiac events.
10
11      6.3.1.4  Issues in the Interpretation of Acute Cardiovascular Effects Studies
12      • Susceptible subpopulations. Because they lack data on individual  subject characteristics,
13       ecologic time series studies provide only limited information on susceptibility factors based on
14       stratified analyses.  The relative impact of PM on cardiovascular (and respiratory) admissions
15       reported in ecologic time series studies are generally somewhat higher than those reported for
16       total admissions. This provides some limited support for hypothesizing that acute effects of
17       PM operate via cardiopulmonary pathways or that persons with pre-existing cardiopulmonary
18       disease have greater susceptibility to PM, or both. Although there is some data from the
19       ecologic time series studies showing larger relative impacts of PM on cardiovascular
20       admissions in adults aged >65 yr as compared with younger populations, the differences are
21       neither striking nor consistent.  However, the individual-level studies of cardiophysiologic
22       function assessed above generally do suggest that elderly persons with pre-existing
23       cardiopulmonary disease are susceptible to subtle changes in heart rate variability in association
24       with PM exposures. Because younger and healthier populations have not yet been assessed, it
25       is not yet possible to say whether the elderly clearly have especially increased susceptibility,
26       but this does represent a reasonable working hypothesis.
27      • Role of other environmental factors. The ecologic time series studies published since 1996 all
28       have controlled adequately for weather influences. Thus, it is deemed unlikely that residual
29       confounding by weather accounts for the PM associations observed. With one possible
30       exception (Pope  et al., 1999a), the roles of meteorological factors have not been analyzed
31       extensively as yet in the individual-level studies of cardiac function. Thus, the possibility of

        March 2001                               6-140       DRAFT-DO NOT QUOTE OR CITE

-------
  1       confounding in such studies cannot yet be readily discounted. Co-pollutants have been
  2       analyzed rather extensively in many of the recent time-series studies of hospital admissions and
  3       PM.  In some studies, PM clearly carries an independent association after controlling for
  4       gaseous  co-pollutants. In others, the "PM effects" are markedly reduced once co-pollutants are
  5       added to the model; but this may in part be due to both'PM and co-pollutants such as CO and
  6       NO2 being emitted from a common source (motor vehicles) and consequent colinearity between
  7       them and/or the gaseous pollutants such as CO having independent effects on cardiovascular
  8       function.
  9     • Temporal patterns of responses following PM exposure.  The evidence from recent time
 10       series studies of CVD admissions suggests rather strongly that PM effects tend to be maximal
 11       at lag 0,  with some carryover to lag 1, with little evidence for important effects beyond lag 1.
 12     • Relation of CVD effects to PM size and chemical composition attributes. Insufficient data
 13       exist  from the time series CVD admissions literature or from the emerging individual-level
 14       studies to provide clear guidance as to which ambient PM components, defined either on the
 15       basis  of size or composition, determine ambient PM CVD effect potency. The epidemiologic
 16       studies published to date have been constrained by the limited availability of multiple PM
 17       metrics.  Where multiple metrics exist, they often are of differential quality due to differences
 18       in numbers of monitoring sites and in monitoring frequency.
 19     • PM effects on blood characteristics related to CVD events. Interesting, though limited, new
 20       evidence has also been derived which is highly suggestive of associations between ambient PM
 21        and increased blood viscosity and increased serum C-reactive protein (both related to increased
 22       risks of serious cardiac events).
 23
 24      6.3.2  Effects of Short-Term Participate Matter Exposure  on the Incidence of
 25             Respiratory Hospital Admissions and Medical Visits
 26      6.3.2.1   Introduction
 27           Among the most severe morbidity measures evaluated with regard to PM exposure are
 28      hospital admissions.  Hospital emergency department (ED) visits represent a somewhat less
 29      severe, but related, outcome that has also been studied in relation to air pollution.  Also doctors'
30      visits represent a related health measure that, although less studied, is relevant to those who also
31      suffer severe health effects, but captures a different population than ED visits: i.e., those who

        March 2001                              6-141       DRAFT-DO  NOT QUOTE OR CITE

-------
 1      choose to visit a private doctor, rather than attend hospital ED. This latter category of pollution-
 2      affected persons can represent a large population, yet one largely unevaluated due to the usual
 3      lack of centralized data regarding doctors' visits.
 4           This section evaluates present knowledge regarding the epidemiologic associations of
 5      hospital admissions and medical visits with ambient PM exposure.  It intercompares various
 6      studies examining each of the size-related PM mass exposure measures (e.g., for PM,0) and study
 7      results for various PM chemical components vis-a-vis their relative associations with health
 8      effects, and their respective extents of coherence with PM associations exhibited across related
 9      health effects measures.  In the following discussion, the main focus for quantitative
10      intercomparisons is on studies and results considering PM metrics that quantitatively measure
11      mass or a specific mass constituent, i.e.,: PM10, PM2 5, sulfates (SO4=), or acidic aerosols (H+).
12      Study results for other related PM metrics (e.g., Black Smoke; BS) are also considered, but only
13      qualitatively, primarily with respect to their coherence or lack of coherence with studies using
14      mass or composition metrics measured in North America. In order to consider potentially
15      confounding effects of other co-existing pollutants, study results for various PM metrics are
16      presented both for: (1) when the PM metric is the only pollutant in the model; and, (2)  the case
17      where a second pollutant (e.g., ozone) is also included.  Results from models with more than two
18      pollutants  included simultaneously are not used for quantitative estimates of coefficient size or
19      statistical strength, due to increased likelihood of bias and variance inflation due to multi-
20      collinearity of various pollutants (e.g., see Harris, 1975).  The approach taken in this section is
21      first to summarize briefly results and implications of the 1996 PM AQCD document regarding
22      this topic,  then to summarize and comment in tabular form on relevant studies newly published
23      since that document, followed by text discussion of key findings most pertinent for present
24      purposes.
25
26      6.3.2.2  Summary of Key Respiratory Hospital Admissions Findings from the 1996
27              Participate Matter Air Quality Criteria Document
28           In the 1996 PM AQCD, it was found that both COPD and pneumonia hospitalization
29      studies showed moderate, but statistically significant, relative risks in the range of 1.06 to
30      1.25 (or 6  to 25% excess risk increment) per 50 ,ug/m3 PM10 increase or its equivalent.  While a
31      substantial number of hospitalizations for respiratory illnesses occur in those >65 years of age,

        March 2001                               6-142       DRAFT-DO NOT QUOTE OR CITE

-------
  1     there are also numerous hospitalizations for those under 65 years of age. Several of the
  2     hospitalization studies restricted their analysis by age of the individuals, but did not explicitly
  3     examine younger age groups. One exception noted was Pope (1991), who reported an increase in
  4     hospitalization for Utah Valley children (aged 0 to 5) for monthly numbers of admissions in
  5     relation to PM10 monthly averages, as opposed to daily admissions in relation to daily PM levels
  6     used in other studies. Studies examining acute associations between indicators of components of
  7     fine particles (e.g., BS; sulfates, SO4=; and acidic aerosols, H+) and hospital admissions were also
  8     reported as finding significant relationships. While sulfates were especially predictive of
  9     respiratory health effects, it was not clear whether the sulfate-related effects were attributable to
 10     their acidity, to the broader effects of associated combustion-related fine particles, or to other
 11     factors.
 12
 13     6.3.2.3  New Respiratory-Related Hospital Admissions Studies
 14          Many recent studies have confirmed PM associations with respiratory hospital admissions.
 15     These studies have examined various admissions categories, including: total respiratory
 16     admissions for all ages and by age; asthma for all ages and by age; chronic obstructive pulmonary
 17     disease (COPD) admissions (usually for patients > 64 yrs.), and pneumonia admissions (for
 18     patients > 64 yrs.). Table 6-17 summarizes important details regarding the study area, study
 19     period, study population, PM indices considered and their concentrations,  the methods employed,
 20     study results and comments, and the "bottom-line" PM index percent excess risks per standard
 21     PM increment  (e.g., 50 ^g/m3 for PM10) from studies published since the 1996 PM AQCD.
 22          The percent excess risk (ER) estimates presented in Table 6-17 are based upon the relative
 23     risks (RR's) provided by the authors, but converted into percent increments per standardized
 24     increments used by the U.S. EPA to facilitate direct intercomparisons of results across studies.
 25     The ER's shown in the table are for the most positively significant pollutant coefficient, which
26     likely overestimates the true health impact of that particular lag (since it may "pick up" some of
27     the effects of intercorrelated adjacent days not included in the model). However, since the ER's
28     usually apply to only a single day of morbidity (e.g., same day, but not following days), the single
29     day results in the table likely underestimate the entire distributed lag effects of a day's pollution
30     on the same and subsequent days (e.g., see Schwartz, 2000b).  Thus, in the absence of a


        March 2001                               6-143       DRAFT-DO NOT QUOTE OR CITE

-------
I
O
O
            TABLE 6-17.  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                   AND HOSPITAL ADMISSIONS STUDIES

Reference/Citation
Location, Duration
PM Index/Concentrations             Study Description:
                                                                                      Results and Comments
                                        PM Index, Lag, Excess Risk %,
                                        (95% CI = LCI, UCL) Co-Pollutants
ON
T1
H
6
o
2!
3
O
c
o
H
W
O
United States

Samet et al, (2000a,b)
Study Period.: 84-95
14 U.S. Cities:  Birmingham,
Boulder, Canton, Chicago, Col.
Springs, Detroit, Minn./St. Paul,
Nashville, New Haven, Pittsburgh,
Provo/Orem, Seattle, Spokane,
Youngstown. Mean pop. aged
65+yr per city =143,000
PM I,, mean = 32.9 /
PM10IQR = NR
          Zanobetti et al. (2000b)
          10 U.S. Cities
                                             Hospital admissions for adults 65+ yrs. for
                                             CVD (mean=22. I/day/city), COPD
                                             (mean=2.0/day/city), and Pneumonia
                                             (mean=5.6/day/city) related to PMU), SO2,
                                             O3, NO2, and CO. City-specific Poisson
                                             models used with adjustment for season,
                                             mean temperature (T) and relative humidity
                                             (RH) (but not their interaction), as well as
                                             barometnc pressure (BP) using LOESS
                                             smoothers (span usually 0.5).  Indicators for
                                             day-of-week and autoregressive terms also
                                             included.
                                   Derived from the Samet et al. (2000a,b)
                                   study, but for a subset of 10 cities. Daily
                                   hospital admissions for total cardiovascular
                                   and respiratory disease in persons aged >65
                                   yr.  Covariates- SO2, O3, CO, temperature,
                                   relative humidity, barometric pressure. In
                                   first stage, performed single-pollutant
                                   generalized additive robust Poisson
                                   regression with seasonal, weather, and day
                                   of week controls. Repeated analysis for
                                   days with PM10 less than 50 ptg/m3 to test
                                   for threshold.  Lags of 0-5 d considered, as
                                   well as the quadratic function of lags 0-5.
                                   Individual cities analyzed first.  The  10 risk
                                   estimates were then analyzed in several
                                   second stage analyses: combining risks
                                   across cities using inverse variance weights,
                                   and regressing risk estimates on potential
                                   effect-modifiers and pollutant confounders.
PM10 positively associated with all three
hospital admission categories, but city
specific results ranged widely, with less
variation for outcomes with higher daily
counts. PM,0 effect estimates not found to
vary with co-pollutant correlation,
indicating that results appear quite stable
when controlling for confounding by
gaseous pollutants.  Analyses found little
evidence that key socioeconomic factors
such as poverty or race are modifiers, but
it is noted that baseline risks may differ,
yielding differing impacts for a given RR.

Same basic pattern of results as in Samet
et al. (2000a,b). For distributed  lag
analysis, lag 0 had largest effect, lags 1
and 2 smaller effects, and none at larger
lags. City-specific slopes were
independent of percent poverty and
percent non-white. Effect size increase
when data were restricted to days with
PMHI less than 50,ug/m3. No multi-
pollutant models reported; however, no
evidence of effect modification by co-
pollutants in second stage analysis.
Suggests association between PM,0 and
total respiratory hospital admissions
among the elderly.
COPD HA's for Adults 65+ vrs.
Lag 0 ER = 7.4% (CI: 5.1, 9.8)
Lagl ER = 7.5%(CI. 5.3, 9.8)
2 day mean (lagO.lagl) ER = 10.3%
         (CI: 7.7, 13)
Pneumonia HA's for Adults 65+ yrs.
Lag 0 ER =8.1% (CI: 6.5, 9.7)
Lag 1 ER = 6.7% (CI: 5.3, 8.2)
2 day mean (lagO, lagl) = 10.3%
         (CI: 8.5, 12.1)
Percent excess respiratory risk (95% CI) per
50 A'g/m3 PMj0 increase:
COPD (0-1 d lag) = 10.6 (7.9, 13.4)
COPD (unconstrained dist. lag) = 13.4 (9.4,
17.4)
Pneumonia (0-1 d lag) = 8.1 (6.5, 9.7)
Pneumonia (unconstrained dist. lag) = 10.1
(7.7, 12.6)
 n

-------
03
g.
NJ
O
o
      TABLE 6-17 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                  AND HOSPITAL ADMISSIONS STUDIES

Reference/Citation
Location, Duration
PM Index/Concentrations            Study Description:
                                                                                    Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% CI = LCI, UCL) Co-Pollutants
 H
 b
 o
 2
 o
 H
O
 c
 o
 H
          United States (cont'd)

          Chen et al. (2000)
          Reno-Sparks, NV (90 - 94)
          Population = 307,000
          B-Gauge PM,0 mean=36.5 ,ug/m3
          PMIO IQR =18.3-44.9 ^g/m3
          PMH) maximum = 201.3 ,ug/m3
          Choudhury et al. (1997)
          Anchorage, Alaska (90 - 92)
          Population = 240,000
          PM,,, mean = 41.5 Mg/m3
          PM10(SD) = 40.87
          PM,0 maximum=565 Mg/m3
Gwynn et al. (2000)
Buffalo, NY (5/88-10/90)
PMU) mn./max. = 24.1/90 8 ^g/
PM,0 IQR = 14.8-29.
SO4" mn./max. = 2.4/3.
SO4- IQR = 23.5 - 7.5 ,ug/m3
H+ mn/max = 36.4/382 nmol/m3
H+IQR =15.7-42.2 nmol/m3
CoH mn/max = 0.2/0.9 10~3 ft.
CoH IQR = 0.1-0.3
                                  Log of COPD (mean=l .72/day) and
                                  gastroenteritis (control) admissions from 3
                                  hospitals analyzed using GAM regression,
                                  adjusting for effects of day-of-week,
                                  seasons, Weather effects (T, WS), and long-
                                  wave effects. No co-pollutants considered.
Using insurance claims data for state
employees and dependents living in
Anchorage, Alaska, number of daily
medical visits determined for asthma (mean
= 2.42/day), bronchitis, and upper
respiratory infections. Used linear
regression, including a time-trend variable,
crude season indicator variables (i.e.,
spring, summer, fall, winter), and a variable
for the month following a volcanic eruption
in 1992.

Air pollutant-health effect associations
with total, respiratory, and circulatory
hospital admissions and mortality examined
using Poisson methods controlling for
weather,  seasonality, long-wave effects, day
of week, holidays,
PM,0 positively associated with COPD
admissions, but no association with
gastroenteritis (GE) diseases, indicating
biologically plausible specificity of the
PM,0-health effects association.
Association remained even after excluding
days with PM,0 above  150 Mg/m3.

Positive association observed between
asthma visits and PM|0. Strongest
association with concurrent-day PM,0
levels.  No co-pollutants considered.
Temperature  and RH did not predict visits,
but did interact with the PM,0 association.
Morbidity relative risk higher with respect
to PMH1 pollution during warmer days.
Strongest associations found between
SO4" and respiratory hospital admissions,
while secondary aerosol H+ and SO4"
demonstrated the most coherent
associations across both respiratory
hospital admissions and mortality.
Addition of gaseous pollutants to the
model had minimal effects on the PM RR
estimates.  CoH weakness in associations
may reflect higher toxicity by acidic sulfur
containing secondary particles versus
carbonaceous primary particles.
COPD All age Admissions
50 ,ug/m3 IQR PM,0 (single pollutant):
ER = 9.4% (CI: 2.2, 17.1)
                                                                                                                 Asthma Medical Visits (all ages):
                                                                                                                 For mean = 50 ,ug/m3 PM,0 (single poll.)
                                                                                                                 Lag = 0 days
                                                                                                                 ER = 20.9%(CI: 11.8,30.8)
Respiratory Hospital Admissions(all ages) PM
Index (using standardized cone, increment)
-Single Pollutant Models
For PM,0 = 50Mg/m3; SO4 = IS^g/m3;
H+ = 75nmoles/m3;COH = 0.5 units/1000ft
 PM10(lag 0) ER =11% (CI: 4.0, 18)
SO4-(lag 0) ER = 8.2% (CI: 4.1, 12.4)
H+(lag 0) ER = 6% (CI: 2.8, 9.3)
CoH(lagO) ER = 3% (CI: - 1.2, 7.4)
o
HH
H

-------
O
O
       TABLE 6-17 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
=^	                           AND HOSPITAL ADMISSIONS STUDIES        	

 Reference/Citation
 Location, Duration
 PM Index/Concentrations           Study Description:
                                                                                  Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% CI = LCI, UCL) Co-Pollutants
cr\
O
O
z:
s
o
c
o
H
trt
O
&
O
h-H
5!
          United States (cont'd)

          Jacobs etal. (1997)
          Butte County, CA (83 - 92)
          Population = 182,000
          PM,<| mean = 34.3 ,ug/m3
          PM,n min/max = 6.6 / 636 /ug/m3
          CoH mean = 2.36 per 1000 lin. ft.
          CoH min/max = 0 /16.5

          Jamasonetal. (1997)
          New York City, NY (82 - 92)
          Population = NR
          PMU, mean = 38.
          Linn etal. (2000)
          Los Angeles, CA (92 - 95)
          Population = NR
          PM|0 mean = 45.5 ,ug/m3
          PMH1 Mm/Max = 5/132 Mg/
                                  Association between daily asthma HA's
                                  (mean = 0.65/day) and rice burning using
                                  Poisson model with a linear term for
                                  temperature, and indicator variables for
                                  season and yearly population.
                                  Co-pollutants were O3 and CO.  PMH,
                                  estimated for 5 of every 6 days from CoH.

                                  Weather/asthma relationships examined
                                  using a synoptic clitnatological multivariate
                                  methodology.  Procedure relates
                                  homogenous air masses to daily  counts of
                                  overnight asthma hospital admission.
                                  Pulmonary hospital admissions (HA's)
                                  (mean=74/day) related to CO, NO2, PM,(),
                                  and O3 in Los Angeles using Poisson model
                                  with long-wave, day of week, holidays, and
                                  weather controls.
Increases in rice straw burn acreage found
to correlate with asthma HA's over time.
All air quality parameters gave small
positive elevations in RR.  PM,,, showed
the largest increase in admission risk.
Air pollution reported to have little role in
asthma variations during fall and winter.
During spring and summer, however, the
high risk categories are associated with
high concentration of various pollutants
(i.e., PM10, SO2,NO2, O3).

PM,0 positively associated with
pulmonary admissions year-round,
especially in winter. No association with
cerebro-vascular or abdominal control
diseases. However, use of linear
temperature, and with no RH interaction,
may have biased effect estimates
downwards for pollutants here most
linearly related to temperature (i.e., O}
and PM,,,).
Asthma HA's (all ages)
For an increase of 50 /Ltg/m1 PM,0:
ER = 6.11% (not statistically significant)
NR
Pulmonary HA's (>29 vrs )
PM,() = 50/^g/m3
(LagO)ER = 3.3%(CI: 1.7,5)

-------
83
O
O
O
       TABLE 6-17 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
	AND HOSPITAL ADMISSIONS STUDIES

 Reference/Citation
 Location, Duration
 PM Index/Concentrations           Study Description:
                                                                                  Results and Comments
                                                                              PM Index, Lag, Excess Risk %,
                                                                              (95% CI = LCI, UCL) Co-Pollutants
Tl
H
6
o
2
o
H
         United States (cont'd)

         Lipsettetal. (1997)
         Santa Clara County, CA
         Population = NR
         (Winters 88 - 92)
         PMI(I mean = 61.2 jUg/m3
         PM10 Min/Max = 9/165 (Ug/
         Moolgavkaretal. (1997)
         Minneapolis-St. Paul 86-91
         Population.^ NR
         Birmingham, AL '86-'91
         Population. = NR
 PMIO mean =
 PM,0fQR =22-41
                            3 (M-SP)
 PM|U mean =43.4 ,ug/m3(Birm)
 PMIO IQR =26-
                                  Asthma emergency department (ER) visits
                                  from 3 acute care hospitals (mean=7.6/day)
                                  related to CoH, NO2, PM,0, and O3 using
                                  Poisson model with long-wave, day of
                                  week, holiday, and weather controls
                                  (analysis stratified by minimum T).  Every
                                  other day PM10 estimated from CoH.
                                  Residential wood combustion (RWC)
                                  reportedly a major source of winter  PM.
                                  Gastro-enteritis (G-E) ER admissions also
                                  analyzed as a control disease.
Investigated associations between air
pollution (PM10, SO2, NO2 O3, and CO) and
hospital admissions for COPD
(mean/day=2.9 in M-SP; 2.3 in Birm) and
pneumonia (mean=7.6 in M-SP; 6.0 in
Birm) among older adults (>64 yrs.).
Poisson GAM's used, controlling for day-
of-week, season, LOESS of temperature
(but neither RH effects nor T-RH
interaction considered).
Consistent relationships found between
asthma ER visits and PM,U, with greatest
effect at lower temperatures. Sensitivity
analyses supported these findings.  NO2
also associated, but in simultaneous
regressions only PM,0 stayed associated.
ER visits for gastroenteritis not
significantly associated with air pollution.
Results demonstrate an association
between wintertime ambient PM10 and
asthma exacerbations in an area where
RWC is a principal PM source.

In the M-SP area, PM,0 significantly and
positively associated with total daily
COPD and pneumonia admissions among
elderly, even after simultaneous inclusion
of O3. When four pollutants included in
the model (PM,0, SO2, O3, NO2), all
pollutants remained positively associated.
In Birm., neither PM,0 nor O3 showed
consistent associations across lags.  The
lower power (fewer counts) and lack of T-
RH interaction weather modeling in this
Southern city vs. M-SP may have
contributed to the differences seen
between cities.
                                                                              Asthma ED Visits (all ages)
                                                                              PMIO = 50 Mg/m3 (2 day lag):
                                                                              At 20° F, ER = 34.7% (CI: 16, 56.5)
                                                                              At 30° F, ER = 22% (CI: 11, 34.2)
                                                                              At 41 ° F, ER = 9.1% (CI: 2.7, 15.9)
COPD + Pneumonia Admissions (>64vrs.)

In M-SP, For PMIO = 50 ng/m* (max Ig)
ER(lgl) = 8.7%(CI. 4.6, 13)
With O3 included simultaneously:
ER(lgl )= 6.9% (95 CI: 2.7, 11.3)

In Birm, For PM,0=50 /^g/m3 (max Ig.)
ER(lgO)=1.5%(CI:-1.5,4.6)
With 03 included simultaneously.
ER(lgO) = 3.2% (CI: -0.7, 7.2)
s
w
o
o
H
W

-------
p
               TABLE 6-17 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                           AND HOSPITAL ADMISSIONS STUDIES        	^	

         Reference/Citation
         Location, Duration
         PM Index/Concentrations            Study Description:
                                                                                  Results and Comments
                                                                              PM Index, Lag, Excess Risk %,
                                                                              (95% CI = LCI, UCL) Co-Pollutants
oo
H
a
o
2
o
H
O
d
o
H
w
          United States (cont'd)

          Nauenberg and Basu (1999)
          Los Angeles (91 -94)
          Wet Season =11/1-3/1
          Dry Season = 5/1-8/15
          Population .= 2.36 Million
          PMIO Mean = 44.81 ^g/m3
          PM,(1SE=1723,ug/m3
         Norrisetal (1999)
         Seattle, WA (9/95- 12/96)
         Pop. Of Children < 18= 107,816
          PM,()IQR=11.6Mg/m3
          osp mean = 0.4 m~l/10"
          (=12.0Mg/m3PM25)
                                           The effect of insurance status on the
                                           association between asthma-related hospital
                                           admissions and exposure to PM,0 and O3
                                           analyzed, using regression techniques with
                                           same day and 8-day weighted moving
                                           average levels, after removing trends using
                                           Fourier series. Compared results during wet
                                           season for all asthma HA's (mean = 8.7/d),
                                           for the uninsured (mean=0.77/d), for
                                           MediCal (poor) patients (mean = 4.36/d),
                                           and for those with other private health or
                                           government insurance (mean = 3.62/d).
The association between air pollution and
childhood (<18 yrs.) ED visits for asthma
from the inner city area with high asthma
hospitahzation rates (0.8/day, 23/day/10K
persons) were compared with those from
lower hospital utilization areas(l. 1 /day,
8/day/10K persons). Daily ED counts were
regressed against PM,0, light scattering
(osp), CO, SO2, and NO2 using a
semiparametnc Poisson regression model
evaluated for over-dispersion and auto-
correlation.
                                       No associations found between asthma
                                       admissions and O3. No O3 or PMK)
                                       associations found in dry season. PM,0
                                       averaged over eight days associated with
                                       increase in asthma admissions, with even
                                       stronger increase among MediCal asthma
                                       admissions in wet season. The authors
                                       conclude that low income is useful
                                       predictor of increased asthma
                                       exacerbations associated with air
                                       pollution. Non-respiratory HA's showed
                                       no such association with PM10.
Associations found between ED visits for
asthma in children and fine PM and
CO.  CO and PM10 highly correlated with
each other (r=.74) and K, an indicator of
woodsmoke pollution. There was no
stronger association between ED visits for
asthma and air pollution in the higher
hospital utilization area than in the lower
utilization area in terms of RR's.
However, considering baseline risks/1 OK
population indicates a higher PM
attributable risk (AR) in the inner city.
All Age Asthma HA's
PM|0 = 50 /ug/m3, no co-pollutant, during wet
season (Jan. 1 - Mar. 1):

All Asthma Hospital Admissions
0-d lag PM10 ER =  16.2 (CI: 2.0, 30)
8-d avg. PM!0 ER = 20.0 (CI: 5.3, 35)

MediCal Asthma Hospital Admissions
8-d avg. PMI(, ER= 13.7 (3.9, 23.4)

Other Insurance Asthma HA's
8-d avg. PMH)ER = 6.2(-3.6, 16.1)

Children's (<18 yrs.) Asthma ED Visits
Single Pollutant Models:
24hPMl()=50//g/m3
Lagl ER = 75.9% (25.1,  147.4)
For25//g/m3 PM25
Lagl ER = 44.5% (CI: 21.7, 71.4)

Multiple Pollutant Models:
24hPM,0=50,ug/m3
Lagl ER = 75.9%(CI: 16.3, 166)
For25(ug/m3PM25
Lagl ER = 51.2%(CI:23.4, 85.2)
 H
 m

-------
o
O
O
       TABLE 6-17 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
=^====^=	                       AND HOSPITAL ADMISSIONS STUDIES
 Reference/Citation
 Location, Duration
 PM Index/Concentrations            Study Description:
                                                                                   Results and Comments
                                                                               PM Index, Lag, Excess Risk %,
                                                                               (95% CI = LCI, UCL) Co-Pollutants
 T1
 H
 6
 o
 2
 s
o
 c
 s
          United States (cont'd)

          Morris et al. (2000)
          Spokane, WA( 1/95 -3/97)
          Population = 300,000
          PM1() mean. = 27.9 jUg/m3
          PMIO Min/Max =4.7/186.4 //g/m3
                   = 21.4Atg/m3
          Seattle, WA (9/95 -12/96)
          Pop. Of Children <18 = 107,816
          PM|0 mean. = 21.5 jug/m3
          PM10 Min/Max = 8/69.3 ,ug/m3
          PM,,,IQR= 11.7Mg/m3

          Schwartz et al. (1996b)
          Cleveland (Cayahoga County), Ohio
          (88 -  90)
          PM10 mean = 43 Aig/m3
          PM,0 IQR = 26-56 ,ug/m3
 Tolbert et al. (2000b)
 Atlanta, GA (92 - 94 Summers)
 Population = 80% of children in
 total population of 3 million
 PM10 mn. (SE) = 38.9 (15.5
 PM,0 Range = 9, 105 ,ug/m3
Associations investigated between an
atmospheric stagnation index (# of hours
below median wind speed), a "surrogate
index of pollution", and asthma ED visits
for persons <65 yr. (mean=3.2/d) in
Spokane and for children <18 yr.
(mean=1.8/d) in Seattle. Poisson GAM
model applied, controlling for day of week,
long-wave effects, and temperature and dew
point (as non-linear smooths).  Factor
Analysis (FA) applied to identify PM
components associated with asthma HA's.

Review paper including an example drawn
from respiratory hospital admissions of
adults aged 65 yr and older (mean = 22/day)
in Cleveland, OH. Categorical variables for
weather and sinusoidal terms for filtering
season employed.

Pediatric (<17 yrs. of age) ED visits (mean
= 467/day) related to air pollution (PMIO,
O3, NOX, pollen and mold) using GEE and
logistic regression and Bayesian models.
Autocorrelation, day of week, long-term
trend terms, and linear temperature controls
included.
Stagnation persistence index was strongly
associated with ED visits for asthma in
both cities. Factor analysis indicated that
products of incomplete combustion
(especially wood-smoke related K., OC,
EC, and CO) are the air pollutants driving
this association. Multi-pollutant models
run with "stagnation" as the "co-pollutant"
indicated importance of general air
pollution over any single air pollutant
index, but not of the importance of various
pollutants relative to each other.

Hospital admissions for respiratory illness
of persons aged 65 yr and over in
Cleveland strongly associated with PM)0
and O3, and marginally associated with
SO2 after control for season, weather, and
day of the week effects.

Both PM|0 and O3 positively associated
with asthma ED visits using all three
modeling approaches. In models with
both O3 and PM10, both pollutants become
non-significant because of high
collineanty of the variables (r=0.75).
                                                                                                                  Asthma ED Visits
                                                                                                                  Single Pollutant Models

                                                                                                                  Persons<65 years (Spokane)
                                                                                                                  ForPM10IQR = 50Jug/m3
                                                                                                                  Lag 3 ER = 2.4% (CI: -10.9, 17.6)

                                                                                                                  Persons<18 years (Seattle)
                                                                                                                  For PMIO IQR = 50 (Ug/m3
                                                                                                                  Lag3ER=56.2%(95CI: 10.4,121.1)
                                                                                                                  Respiratory HA's for persons 65+ years
                                                                                                                           PM1(,
                                                                                                                       5.8%(CI:0.5, 11.4)
                                                                                                                           Pediatric (<17 vrs. of age) ED Visits
Lag 1 day ER = 13.2% (CI: 1.2, 26.7)
With O3 8.2 (-7.1, 26.1)
 o

-------
o
O
O
      TABLE 6-17 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
	                                        AND HOSPITAL ADMISSIONS STUDIES                           	

 Reference/Citation
 Location, Duration
 PM Index/Concentrations           Study Description:
                                                                                   Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% CI = LCI, UCL) Co-Pollutants
o

H
 O
 H
O
          United States (cont'd)

          Tolbertetal. (2000a)
          Atlanta
          Period 1:  1/1/93-7/31/98
          Mean, median, SD:
          PM,0Gug/m3):  30.1,28.0, 12.4

          Period 2:  8/1/98-8/31/99
          Mean, median, SD:
          PMio Cug/m3):  29.1, 27.6, 12.0
          PM25Oug/m3): 19.4,17.5,9.35
          CP (Mg/m3): 9.39, 8.95, 4.52 10-
          100 nm PM counts
          (count/cm3): 15,200,10,900,
          26,600
          10-100 nm PM surface area
          (unrVcm3): 62.5,43.4,  116
          PM25 soluble metals (,ag/m3):
          0.0327, 0.0226, 0.0306
          PM25Sulfates(pg/m3):  5.59,4.67,
          3.6
          PM2 5 Acidity Cug/m3):  0.0181,
          0.0112,0.0219
          PM2 5 organic PM Cug/m3): 6.30,
          5.90,3.16
          PM2 5 elemental carbon (,ug/m3):
          2.25,1.88,1.74
                                  Preliminary analysis of daily emergency
                                  department (ED) visits for asthma (493),
                                  wheezing (786.09) COPD (491, 492, 4966)
                                  LRI 466.1, 480, 481, 482, 483, 484, 485,
                                  486), all resp disease (460-466, 477, 480-
                                  486, 491, 492, 493, 496, 786.09) for
                                  persons > 16 yr in the period before (Period
                                  1) and during (Period 2) the Atlanta
                                  superstation study. ED data analyzed here
                                  from just 18 of 33 participating hospitals;
                                  numbers of participating hospitals increased
                                  during period 1.  Mean daily ED visits for
                                  dysrhythmias and all DVD in period 1 were
                                  6.5 and 28.4, respectively. Covariates:
                                  NO2, 03, SO2, CO temperature, dewpoint,
                                  and, in period 2 only, VOCs. PM measured
                                  by both TEOM and Federal Reference
                                  Method; unclear which used in analyses.
                                  For epidemiologic analyses, the two time
                                  periods were analyzed  separately. Poisson
                                  regression analyses were conducted with
                                  cubic splines for time, temperature and
                                  dewpoint. Day-of-week and hospital
                                  entry/exit indicators also included.
                                  Pollutants treated a-priori as three-day
                                  moving averages of lags 0, 1, and 2. Only
                                  single-pollutant results reported.
In period 1, observed significant COPD
association with 3-day average PM,0.
COPD was also positively associated with
NO2, O3, CO and SO2.  No statistically
significant association observed between
asthma and PM10 in period 1.  However,
asthma positively associated with ozone
(p=0.03). In period 2, i.e., the first year of
operation of the superstation, no
statistically significant associations
observed with PM10 or PM2 5.  These
preliminary results should be interpreted
with caution given the incomplete and
variable nature of the databases analyzed.
                                                                                                                         Period 1:
                                                                                                                         PMIO (0-2 d):
                                                                                                                           asthma:
                                                                                                                             5.6% (-8.6, 22.1)
                                                                                                                           COPD:
                                                                                                                             19.9% (0.1, 43.7)

                                                                                                                         Period 2: (all 0-2 day lag)
                                                                                                                         PM,0:  asthma
                                                                                                                                18.8% (-8.7, 54.4)
                                                                                                                                COPD
                                                                                                                                -3.5% (29.9- 33.0)
                                                                                                                         PM25:  asthma
                                                                                                                                2.3% (-14.8, 22.7)
                                                                                                                                COPD
                                                                                                                                12.4% (-7.9, 37.2)
                                                                                                                         PMH)25 asthma
                                                                                                                                21.1% (-18.2, 79.3)
                                                                                                                                COPD
                                              x_-v_*» ±-r
                                              -23.0% (50.7 - 20.1)
 o
 HH
 H
 W

-------
I
KJ
O
O
       TABLE 6-17 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
	AND HOSPITAL ADMISSIONS STUDIES              	^====^==

 Reference/Citation
 Location, Duration
 PM Index/Concentrations           Study Description:
                                                                                  Results and Comments
                                      PM Index, Lag, Excess Risk %,
                                      (95% CI = LCI, UCL) Co-Pollutants
w
£
H
o
O
z
O
H
O
O
H
m
O
          United States (cont'd)

          Yang etal( 1997)
          Study Period: 92 - 94
          Reno-Sparks, Nevada
          Population = 298,000
          PM10 mean = 33.
         PM10 range = 2.2, 1 57.3
         Zanobetti, et al. (2000a)
         Study Period: 86-94
         Chicago (Cook Count), IL
         Population = 633,000 aged 65+
         PM,0 mean = 33.6 ,ug/m3
         PMIO range = 2.2, 157.3 ^g/m3
                                  Association between asthma ER visits
                                  (mean = 1.75/d, SD=1.53/d) and PM10, CO
                                  and O3 assessed using linear WLS and
                                  ARIMA regression, including adjustments
                                  for day-of-week, season, and temperature
                                  (but not RH or T-RH interaction). Season
                                  adjusted only crudely, using month dummy
                                  variable.

                                  Analyzed HA's for older adults (65 + yr)
                                  for COPD (mean = 7.8/d), pneumonia
                                  (mean = 25.5/d), and CVD, using Poisson
                                  regression controlling for temperature, dew
                                  point, barometric pressure, day of week,
                                  long wave cycles and autocorrelation,  to
                                  evaluate whether previous admission or
                                  secondary diagnosis for associated
                                  conditions increased risk from air pollution.
                                  Effect modification by race, age, and sex
                                  also evaluated.
Only O3 showed significant associations
with asthma ER visits.  However, the
crude season adjustment and linear model
(rather than Poisson) may have adversely
affected results. Also, Beta-gauge PM,0
mass index used, rather than direct
gravimetric mass measurements.
Air pollution- associated CVD HA's were
nearly doubled for those with concurrent
respiratory infections (RI) vs. those
without concurrent RI. For COPD and
pneumonia admissions, diagnosis of
conduction disorders or dysrhythmias
(Dyshr.) increased PMH1RR estimate. The
PM,d RR effect size did not vary
significantly by sex,  age, or race, but
baseline risks across these groups differ
markedly, making such sub-population RR
inter-comparisons difficult to interpret.
NR
PM|0 = 50 /ug/m3(average of lags 0,1)
COPD (adults 65+ vrs.)
W/o prior RI. ER = 8.8% (CI: 3.3, 14.6)
With prior RI ER = 17.1% (CI: -6.7, 46.9)
COPD (adults 65+ vrs.)
W/o concurrent Dys. ER = 7.2% (CI: 1.3,13.5)
With concurrent Dys. ER = 16.5%(CI: 3.2,
31.5)
Pneumonia (adults 65+ vrs.)
W/o pr. Asthma ER =11% (CI: 7.7, 14.3)
With pr. Asthma ER = 22.8% (CI: 5.1, 43.6)
Pneumonia (adults 65+ vrs.)
W/o pr. Dyshr. ER = 10.4% (CI: 6.9, 14)
With pr. Dyshr. ER = 18.8% (CI: 6.3, 32.7)
O

-------
o

K>
O
O
               TABLE 6-17 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                         AND HOSPITAL ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
                                          Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
s^
Z
•n
H
6
o
o
H
O
c
o
H
W
O
&
o
United States (cont'd)
Lippmann et al. (2000)
Detroit, MI ('92-'94)
Population = 2.1 million
PM10Mean = 31 Atg/m3
(IQR=19, 38,ug/m3;
max=105,ug/m3)
PM25Mean= 18^g/m3
(IQR= 10, 21  ,ug/m3; max=86
,ug/m3)
PMI0.2 5 Mean =
(IQR= 8, 17 ^g/
SO4"Mean = 5
(IQR=1.8, 6.3 ^g/m3;
max=34.5 Mg/m3)
H+ Mean = 8.8 nmol/m3 = 0.4
(IQR=0, 7nmol/m3;max=279)
                                          Respiratory (COPD and Pneumonia) HA's
                                          for persons 65 + yr. analyzed, using GAM
                                          Poisson models, adjusting for season, day
                                          of week, temperature, and relative humidity.
                                          The air pollution vanables analyzed were:
                                          PMIO, PM25, PM10.25, sulfate, H+, O3, SO2,
                                          NO2, and CO. However, this study
                                          site/period had very low acidic aerosol
                                          levels. As noted by the authors 85% of H+
                                          data was below detection limit (8 nmol/m3).
                         ; max=
For respiratory HA's, all PM metrics
yielded RR's estimates >1, and all were
significantly associated in single pollutant
models for pneumonia  For COPD, all
PM metrics gave RR's >1, with H+ being
associated most significantly, even after
the addition of O3 to the regression.
Adding gaseous pollutants had negligible
effects on the various PM metric RR
estimates. The most consistent effect of
adding co-pollutants was to widen the
confidence bands on the PM metric RR
estimates: a common statistical artifact of
correlated predictors. Despite usually
non-detectable levels, H+ had strong
association with respiratory admissions on
the few days it was present. The general
similarity of the PM25 and PM|0_25 effects
per,ug/m3 in this study suggest similarity
in human toxicity of these two inhalable
mass components in study locales/periods
where PM2 5 acidity is usually not present.
Pneumonia HA's for 65+ vrs.
No co-pollutant:
PMlo(50,ug/m3)ldlag
  ER = 22% (CI: 8.3, 36)
PM25(25//g/m3)ldlag:
  ER=13%(CI:3.7,22)
PM25.,(1(25Mg/m3)ldlag:
  ER=12%(CI:0.8,24)
H+ (75 nmol/m3) 3d lag:
  ER=12%(CI:0.8, 23)
 O, co-pollutant (lag 3) also in model:
PM,0(50^g/m3)ldlag,
  ER = 24% (CI: 8.2, 43)
PM25(25Mg/m3)ldlag:
  ER=12%(CI: 1.7,23)
PM254()(25^g/m3)ldlag:
  ER=14%(CI:0.0, 29)
H+ (75 nmol/m3) 3d lag:
  ER=11%(CI:-0.9, 24)
COPD Hospital Admissions for 65+ yrs.
 No co-pollutant:
PMIO (50 Mg/m3) 3d lag
  ER = 9.6%(CI:-5.1,27)
PM2S(25(ug/m3)3dlag:
  ER = 5.5%(CI: -4.7, 17)
PM25.lo(25^g/m3)3dlag:
  ER = 9.3% (CI: -4.4, 25)
H+ (75 nmol/m3) 3d lag:
  ER=13%(CI:0.0,28)
 O, co-pollutant (lag 3) also in model:
PM,0 (50 ^g/m3) 3d lag,
  ER= 1.0% (-15, 20)
PM25(25,ug/m3)3dlag:
  ER = 2.8% (CI: -9.2, 16)
PM25.,0(25Mg/m3)3dlag:
  ER = 0.3%(CI: -14, 18)
H+ (75 nmol/m3) 3d lag:
  ER =13% (CI:-0.6.28)	

-------
£
p
O
o
      TABLE 6-17 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                 AND HOSPITAL ADMISSIONS STUDIES                     	===

Reference/Citation
Location, Duration
PM Index/Concentrations           Study Description:
                                                                                   Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% CI = LCI, UCL) Co-Pollutants
 H
 6
 o
 z
 o
 H
O
 O
 H
 W
 O
 ?o
 o
 H
 W
          United States (cont'd)

          Lumley and Heagerty (1999)
          Seattle (King Cty.), WA (87-94)
          Population = NR
          PM, daily mean = NR
          PM|.|,, daily mean = NR
          From Sheppard et al, 1999:
          PM10 mean = 31.5 Aig/m3
          PM1()IQR=  19-39 ,ug/m3
          PM25 mean = 16.7 jUg/m3
          PM25IQR=8-21 Aig/m3
          Moolgavkaretal. (2000)
          King County, WA (87 - 95)
          Population = NR
          PMHI mean = 30.0 /^g/m3
          PMU)1QR =18.9-37.3 ^g/m3
          PM2 5 mean =18.1 ^g/m3
          PM25IQR=10-23Mg/m3
                                  Estimating equations based on marginal
                                  generalized linear models applied to
                                  respiratory HA's for persons <65 yrs. of age
                                  (mean " 8/day) using  class of variance
                                  estimators based upon weighted empirical
                                  variance of the estimating functions.
                                  Poisson regression used to fit a marginal
                                  model for the log of admissions with linear
                                  temperature, day of week, time trend, and
                                  dummy season variables.  No co-pollutants
                                  considered.
                                  Association between air pollution and
                                  hospital admissions (HA's) for COPD
                                  (all age mean=7.75/day; 0-19 yrs.
                                  mean=2.33/day) investigated using Poisson
                                  GAM's controlling for day-of-week,
                                  season, and LOESS of temperature. Co-
                                  pollutants addressed:  O3, SO2, CO, and
                                  pollens. PM2 5 only had one monitoring site
                                  versus multiple sites averaged for other
                                  pollutants.
PM, at lag 1 day associated with
respiratory HA's in children and younger
adults (<65), but not PMun, suggesting a
dominant role by the submicron particles
in PM2 5-asthma HA associations reported
by Sheppard et al. (1999). 0-day lag PM,
and 0 and 1  day lag PM,.H, had RR near 1
and clearly non-significant. Authors note
that model residuals correlated at r=0.2,
suggesting the need for further long-wave
controls in the model (e.g., inclusion of
the LOESS of HA's).

Of the PM metrics, PMU) showed the most
consistent associations across lags (0-4 d).
PM2 5 yielded the strongest positive PM
metric association at Iag3 days, but gave a
negative association at Iag4 days. That
PM2 5 only had one monitoring site may
have contributed to its effect estimate
variability.  Residual autocorrelations (not
reported) may also be a factor. Adding
gaseous co-pollutants or pollens decreased
the PM2 5 effect estimate less than PM10.
Analyses indicated that asthma HA's
among the young were driving the overall
COPD-air pollution associations.
Respiratory HA's for persons <65 vrs. old
PM, = 25 ,ug/m3, no co-pollutant:

l-dlagER = 5.9(l.l, 11.0)
COPD HA's all ages (no co-pollutant)
PM10 (50 Aig/m3, lag 2)
      ER = 5.1%(CI:0, 10.4)
PM25(25Atg/m3, lag 3)
      ER = 6.4%(CI:0.9, 12.1)

COPD HA's all ages (CO as co-pollutant)
PM,0 (50 /.g/m3, lag 2)
      ER = 2.5% (CI: -2.5, 7.8)
PM25(25Mg/m3,lag3)
      ER=5.6%(CI:0.2, 11.3)

-------
63
O
O
      TABLE 6-17 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                  AND HOSPITAL ADMISSIONS STUDIES             	

Reference/Citation
Location, Duration
PM Index/Concentrations            Study Description:
                                                                                    Results and Comments
                                                                               PM Index, Lag, Excess Risk %,
                                                                               (95% CI = LCI, UCL) Co-Pollutants
ON
a
o
2
o
H
O
C
O
H
W
O
?o
n
t-H
H
          United States (cont'd)

          Moolgavkar (2000a)
          Study Period: 1987-1995

          Chicago (Cook County), IL
          Population = NR
          PMU) median = 35 £/g/m3
          PMHJ IQR = 25-47 ,ug/m3

          Los Angeles (LA County), CA
          Population = NR
          PMUI median = 44 /^g/m3
          PMH)IQR = 33-f
          PM2SIQR=15-31/ug/m3

          Phoenix (Mancopa County). AZ
          Population = NR
          PM,,| median = 41 ^g/m3
          PM10 IQR =32-51 /ng/m3
Investigated associations between air
pollution (PM10, O3, SO,, NO2, and CO)
and COPD Hospital Admissions (HA's).
PM2 5 also analyzed in Los Angeles. HA's
for adults >65 yr.:  median=12/day in
Chicago, =4/d in Phoenix; =20/d in LA.
In LA, analyses also conducted for children
0-19 yr (med.=17/d) and adults 20-64
(med.=24/d). Poisson GAM's used
controlling for day-of-week, season, and
splines of temperature and RH (but not their
interaction) adjusted for overdispersion.
PMdata available only every 6th day
(except for daily PM,,, in Chicago), vs.
every day for gases. Power likely differs
across pollutants, but number of sites and
monitoring days not presented. Two
pollutant models forced to have same lag
for both pollutants. Autocorrelations or
intercorrelations of pollutant coefficients
not presented or discussed.
                                                                          For >64 adults, CO, NO2 and O3 (in
                                                                          summer) most consistently associated with
                                                                          the HA's.  PM effects more variable,
                                                                          especially in Phoenix. Both positive and
                                                                          negative significant associations for PM
                                                                          and other pollutants at different lags
                                                                          suggest possible unaddressed negative
                                                                          autocorrelation. In LA, PM associated
                                                                          with admissions in single pollutant
                                                                          models, but not in two pollutant models.
                                                                          The forcing of simultaneous pollutants to
                                                                          have the same lag (rather than maximum
                                                                          lag), which likely maximizes
                                                                          intercorrelations between pollutant
                                                                          coefficients, may have biased the two
                                                                          pollutant coefficients, but information not
                                                                          presented.. Analysis in 3 age groups in
                                                                          LA yielded similar results. Author
                                                                          concluded that  "the gases, other than
                                                                          ozone, were more strongly associated with
                                                                          COPD admissions than PM, and that there
                                                                          was considerable heterogeneity in the
                                                                          effects of individual pollutants in different
                                                                          geographic areas".
Most Significant Positive ER
Single Pollutant Models:
COPD HA's (>64 yrs.) (50 Mg/m3 PM,,,):
Chicago: Lag 0 ER =2% (CI: -0.2, 4.3)
LA:     Lag2ER = 6.1%(CI: 1.1,11.3)
Phoenix: Lag 0 ER = 6.9% (CI: -4.1, 19.3)

LA COPD HA's

(50 //g/m3 PM,,,, 25 //g/m3 PM25 or PM25.]0)

(0-19 yrs.): PMU) lg2=10.7%(CI: 4.4, 17.3)
(0-19 yrs.): PM25 lgO=4.3%(CI: -0.1, 8.9)
(0-19 yrs.): PM(25.,0) lg2=17.1%(CI: 8.9, 25.8)
(20-64 yrs.): PM10 lg2=6.5%(CI: 1.7, 11 5)
(20-64 yrs.): PM25 lg2=5.6%(Cl: 1.9, 9.4)
(20-64 yrs.): PM25.IO lg2=9%(CI: 3, 15.3)

(>64yrs):PM10lg2 = 6.1%(l.l,11.3)
(> 64 yrs): PM25 Ig2 = 5.1% (0.9, 9.4)

(>64 yrs.): PM25.IO Ig3=5.1% (CI: -0.4, 10.9)

(>64 yr) 2 Poll. Models (CO = co-poll.)

PM,,,: Lag 2 ER = 0.6% (CI: -5.1, 6.7)
PM25: Lag 2 ER = 2.0% (-2.9, 7.1)

-------
g.
O
O
      TABLE 6-17 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                 AND HOSPITAL ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
                                           Study Description:
                                       Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% CI = LCI, UCL) Co-Pollutants
H
6
o
2!
O
H
o
         United States (cont'd)

         Sheppard et al.  (1999)
         Seattle, WA, Pop. = NR
         1987-1994
         PM1() mean = 31.5 ^g/m1
         PMI(1 IQR =19-39 Mg/m3
         PM25 mean = 16.7 ^g/m3
         PM25IQR = 8-21 ^g/m1
         PM25.,,, mean = 16.2 /ug/
         Canada

         Burnett etal. (1997b)
         Toronto, Canada (1992-1994),
         Pop. = 4 mill.
         PM25 mean = 16.8 A NO2, SO2, and CO)
evaluated.
                                       Asthma HA's significantly associated \vith
                                       PM10, PM25> and PMU>_25 mass lagged 1
                                       day, as well as CO.  Authors found PM
                                       and CO to be jointly associated with
                                       asthma admissions.  Highest increase in
                                       risk in spring and fall. Results conflict
                                       with hypothesis that wood smoke (highest
                                       in early study years  and winter) would be
                                       most toxic. Associations of CO with
                                       respiratory HA's taken by authors to be an
                                       index of incomplete combustion, rather
                                       than direct CO biological effect.
Positive air pollution-HA associations
found, with ozone being pollutant least
sensitive to adjustment for co-pollutants.
However, even after the simultaneous
inclusion of O3 in the model, the
association with the respiratory hospital
admissions were still significant for PMU>,
PM2S, PM25.1(), CoH,, SO4, and H+.
                                       Asthma Admissions (ages 0-64)
                                       PM,,,(lag=lday);50^g/m3
                                       ER= 13.7% (CI: 5.5%, 22.6)
                                       PM25 (lag=lday); 25 ,ug/m3
                                       ER = 8.7%(C1:3.3%, 14.3)
                                       PMzs-io (lag=lday); 25 //g/m3
                                       ER=11.1%(CI:2.8%, 201)
Respiratory HA's all ages(no co-pollutant)
PM10 (50 ,ug/m3, 4d avg. lag 0)
   ER = 10.6% (CI: 4.5-17.1)
PM25 (25 Aig/m3, 4d avg. lag 1)
   ER = 8.5% (CI: 3.4, 13.8)
PM25.10(25//g/m3, Sdavg. lag 0)
   ER = 12.5% (CI: 5.2,20.0)
Respiratory HA's all agesffl^ co-pollutant)
PM1(, (50 Aig/m3, 4d avg. lag 0)
   ER = 9.6% (CI: 3.5, 15.9)
PM25 (25 jug/m3, 4d avg., lag 1)
   ER = 6.2%(1.0, 11.8)
PM25.10 (25 Mg/m3, 5d avg. lag 0)
   ER = 10.8% (CI: 3.7, 18.1)
o
H
W

-------
I
K)
O
o
      TABLE 6-17 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                  AND HOSPITAL ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
                                            Study Description:
                                        Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% Cl = LCI, UCL) Co-Pollutants
7*
O
O
2
O
H
O
e!
O
H
m
O
js
O
HH
H
m
          Canada (cont'd)
          Burnett et al. (1999)
          Metro-Toronto, Canada
          1980-1994

          Pollutant: mean, median, IQR.
          FP^O^g/m3):  18,16,10
          CPes, (Mg/m3):  12,10,8
          PMl()cs,Cug/m3): 30,27,15
Delfmoetal. (1997)
Montreal, Canada
Population= 3 million
6-9/92, 6-9/93
1993 Means (SD):
PM10= 21.7 ,ug/m3 (10.2)
PM25=12.2^g/m3(7.1)
SO4== 34.8nmol/m3(33.1)
H+=  4 nmol/m3 (5.2)
Daily hospitalizations for asthma (493,
mean 1 I/day), obstructive lung disease
(490-492, 496, mean 5/day), respiratory
infection (464, 466, 480-487, 494, mean
13/day) analyzed separately in relation to
environmental covariates.  Same geographic
area as in Burnett et al., 1997b. Three size-
classified PM metrics were estimated, not
measured, based on a regression on TSP,
SO4, and COH in a subset of every 6th-day
data. Generalized additive models. Non-
parametric LOESS prefilter applied to both
pollution and hospitalization data.  Day of
week controls. Tested 1-3 day averages of
air pollution ending on lags 0-2. Covanates:
O3, NO2, SO2, CO, temperature, dewpomt
temperature, relative humidity.

Association of daily respiratory emergency
department (ED) visits (mean = 98/day
from 25 of 31 acute care hospitals) with O3,
PM10, PM25, SO4=, and H+ assessed using
linear regression with controls for temporal
trends, auto-correlation, and weather. Five
age sub-groups considered.
                                                                           In univariate regressions, all three PM
                                                                           metrics were associated with increases in
                                                                           respiratory outcome. In multi-pollutant
                                                                           models, there were no significant PM
                                                                           associations with any respiratory outcome
                                                                           (results not shown).  Use of estimated PM
                                                                           metrics limits the interpretation of
                                                                           pollutant-specific results reported.
                                                                           However, results suggest that a linear
                                                                           combination of TSP, SO4, and COH does
                                                                           not have a strong independent association
                                                                           with cardiovascular admissions when a
                                                                           full range of gaseous pollutants are also
                                                                           modeled.
No associations with ED visits in '92, but
33% of the PM data missing then. In '93,
only H+ associated for children <2, despite
very low H+ levels.  H+ effect stable in
multiple pollutant models and after
excluding highest values. No associations
for ED visits in persons aged 2-64 yrs.
For patients >64 yr, O3, PM,0, PM2 5, and
SO4=  positively associated with visits
(p < 0.02), but PM effects smaller than for
                                                                                                                            PM,0
                                                                                                                            PM10.25
                                       Percent excess risk (95% CI) per 50 /^g/
                                       PMIO; 25 //g/m3 PM2 5 and PM(10.2 S}:
                                       Asthma
                                       PM25(0-l-2d): 6.4(2.5, 10.6)
                                       PM,Q(0-1 d): 8.9(3.7, 14.4)
                                       PM,0.25(2-3-4d): 11.1 (5.8, 16.6)
                                       COPD
                                       PM25:  4.8 (-0.2, 10.0)
                                              6.9(1.3,12.8)
                                              (2-3-4d): 12.8(4.9,21.3)

                                       Resp. Infection:
                                       PM25:  10.8(7.2, 14.5)
                                       PM,0:  14.2(9.3, 19.3)
                                       PMlo.25(0-l-2d): 9.3(4.6,14.2)
                                                                                                                            Respiratory ED Visits
Adults >64: (pollutant lags = 1 day)
50 //g/m3 PMI(1ER = 36.6% (10.0, 63.2)
         PM25 ER = 23.9% (4.9, 42.8)

-------
 o
 cr
 O
 O
      TABLE 6-17 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                 AND HOSPITAL ADMISSIONS STUDIES         	_=_^^_
Reference/Citation
Location, Duration
PM Index/Concentrations           Study Description:
                                                                                   Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% CI = LCI, UCL) Co-Pollutants
 L/1
          Canada (cont'd)

          Delfinoetal. (1998)
          Montreal, Canada
          6-8/89,6-8/90
          Mean PM10= 18.6 ,ug/m3
          (SD=9.3, 90lh% = 30.0 ,ug/m3)
          Stiebetal. (1996)
          New Brunswick, Canada
          Population = 75,000
          May-Sept. 84 - 92
          Range: 1-23, 95lh% =1
          TSP Mean = 36.7 //g/m3
          Range:5-108, 95ih% =70 //g/m3
                                  Examined the relationship of daily ED
                                  visits for respiratory illnesses by age
                                  (mean/day: <2yr.=8.9; 2-34yr.=20.1; 35-
                                  64yr =22.6; >64yr.=20.3) with O3 and
                                  estimated PM25. Seasonal and day-of-week
                                  trends, auto-correlation, relative humidity
                                  and temperature were addressed in linear
                                  time series regressions.

                                  Asthma ED visits (mean=l  6/day) related to
                                  daily 03 and other air pollutants (SO2, NO2,
                                  SO42", and TSP). PM measured only every
                                  6th day. Weather variables included
                                  temperature, humidex, dewpomt, and RH.
                                  ED visit frequencies were filtered to remove
                                  day of week and long wave trends.  Filtered
                                  values were regressed on pollution and
                                  weather variables for the same day and the
                                  3 previous days.
There was an association between PM2_,
and respiratory ED visits for older adults
(>64), but this was confounded by both
temperature and O3. The fact that PM2 s
was estimated, rather than measured, may
have weakened its relationship with ED
visits, relative to O3.
Positive, statistically significant (p < 0.05)
association observed between O3 and
asthma ED visits 2 days later; strength of
the association greater in nonlinear
models. Ozone effect not significantly
influenced by addition of other pollutants.
However, given  limited number of
sampling days for sulfate and TSP, it was
concluded that "a particulate effect could
not be ruled out".
Older Adults(>64 vr) Respiratory ED Visits
Estimated PM2 5 = 25 ,ug/m3

Single Pollutant:
(lag 1 PM25) ER = 13.2 (-0.2, 26.6)

With Ozone (lag 1 PM25):
Est. PM25 (lagl) ER = 0.8% (CI: - 14.4, 15.8)

Emergency Department Visits (all ages)
Single Pollutant Model
100 ,ug/m3 TSP = 10.7% (-66.4, 87.8)
 TI
 H
 6
 o
 o
 H
O
 c
 o
 H
 m
 o
 &
 o
 h-H
 H
 W

-------
O
o
                TABLE 6-17 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                            AND HOSPITAL ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
                                            Study Description:
                                        Results and Comments
                                        PM Index, Lag, Excess Risk %,
                                        (95% CI = LCI, UCL) Co-Pollutants
oo
TI
H
•Z
O
H
O
G
O
H
W
O
&
O
N-*
H
m
          Canada (cont'd)

          Stieb et al. (2000)
          Saint John, Canada
          7/1/92-3/31/96
          mean and S.D.:
          PM,0 (,ug/m3): 14.0,9.0
          PM25(Aig/m3): 8.5,5.9
          H+(nmol/m3): 25.7,36.8
          Sulfate (nmol/m3):  31.1,29.7
          COHmean(103lnft): 0.2,0.2
          COH max (103 In ft): 0.6, 0.5
Burnett etal. (1997c)
16 Canadian Cities('81-91)
Population=12.6 MM
CoH mean=0.64(per 103 lin. ft)
CoH IQR=0.3-0.8(per 1031m ft)
Study of daily emergency department (ED)
visits for asthma (mean 3.5/day), COPD
(mean 1.3/day), resp infections (mean
6.2/day), and all respiratory conditions
(mean 10.9/day) for persons of all ages.
Covanates included CO, H2S, NO2, O3,
SO2, total reduced sulfur (TRS), a large
number of weather variables, and 12 molds
and pollens.  Stats: generalized additive
models with LOESS prefiltering of both ED
and pollutant variables, with variable
window lengths. Also controlled for day of
week and LOESS-smoothed functions of
weather.  Single-day, and five day average,
pollution lags tested out to lag 10. The
strongest lag, either positive or negative,
was chosen for final models. Both single
and multi-pollutant models reported. Full-
year and May-Sep models reported.

Air pollution data were compared to
respiratory hospital admissions
(mean=l .46/million people/day) for
16 cities across Canada. Used a random
effects regression model, controlling for
long-wave trends, day of week, weather,
and city-specific effects.
                                                                           In single-pollutant models, significant
                                                                           positive associations were observed
                                                                           between all respiratory ED visits and
                                                                           PM10, PM25, H2S, O3, and SO2.
                                                                           Significant negative associations were
                                                                           observed with H+, and COH max. PM
                                                                           results were similar when data were
                                                                           restricted to May-Sep. In multi-pollutant
                                                                           models, no PM metrics significantly
                                                                           associated with all cardiac ED visits in full
                                                                           year analyses, whereas both O3 and SO2
                                                                           were. In the May-Sep subset, significant
                                                                           negative association found for sulfate. No
                                                                           quantitative results presented for non-
                                                                           significant variables in these multi-
                                                                           pollutant regressions.
The 1 day lag of 03 was positively
associated with respiratory admissions in
the April to December period, but not in
the winter months. Daily maximum 1-hr.
CoH from 11 cities and CO also positively
associated with HA's, even after
controlling forO3.
                                        PM25, (lag 3) 15.1 (-0.2,32.8)

                                        PM10, (lag 3) 32.5 (10.2, 59.3)
Respiratory HA's all ages (with O^.CO)
CoH IQR = 0.5, lag 0:
CoH ER = 3.1% (CI: 1.0-4.6%)

-------
o
O
o
                TABLE 6-17 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                            AND HOSPITAL ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
                                            Study Description:
                                                                          Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% CI = LCI, UCL) Co-Pollutants
 z
 o
 H
O
 c!
 O
 a
 o
 :*
 o
 t—i
 H
 W
          Europe

          Atkinson et al. (1999a)
          London (92 - 94)
          Population = NR
          PM10 Mean = 28.5 Mg/m3
          10lh-90IhlQR= 15.8-46.5 ,ug/m3
          BS mean =12.7 Aig/m3
          10th-90'h IQR = 5.5-21.6 ,ug/m3
          Atkinson et al. (1999b)
          London (92 - 94)
          Population = 7.2 MM
          PM,() Mean = 28.5
          10lh-90'h IQR = 15.8-46.5
          BSmean= 12.7^g/m3
          10th-90lh IQR = 5.5-21.
                                  All-age Respiratory (mean=90/day),
                                  Asthma (25.9/day), and Other Respiratory
                                  (64.1/day) ED visits from 12 London
                                  hospitals considered, but associated
                                  population size not reported. Counts for
                                  ages 0-14, 15-64, and >64 also examined.
                                  Poisson  regression used, controlling for
                                  season, day of week, meteorology,
                                  autocorrelation, overdispersion, and
                                  influenza epidemics.
                                  All-age respiratory (mean=150.6/day), all-
                                  age asthma (38.7/day), COPD plus asthma
                                  in adults >64 yr. (22.9/day), and lower
                                  respiratory (64.1/day) m adults >64 yr
                                  (16.7/day) hospital admissions in London
                                  hospitals considered. Counts for ages 0-14,
                                  15-64, and >64 yr also examined.  Poisson
                                  regression used, controlling for season, day-
                                  of-week, meteorology, autocorrelation,
                                  overdispersion, and influenza epidemics.
PM,0 positively associated, but not BS, for
all-age/all-respiratory category.  PM10
results driven by significant children and
young adult associations, while older adult
visits had negative (but non-significant)
PMU)-ED visit relationship. PM,0
positively associated for all ages, children,
and young adults for asthma ED visits.
However, PM10-asthma relationship
couldn't be separated from SO2 in multi-
pollutant regressions.  Older adult ED
visits most strongly associated with CO.
No Oj-ED visits relationships found (but
no warm season analyses attempted).
Positive associations found between
respiratory-related emergency hospital
admissions and PM10 and SO2, but not for
O3 or BS.  When SO2 and PM|0 included
simultaneously,  size and significance of
each was reduced. Authors concluded that
SO2 and PM10 are both indicators of the
same pollutant mix in this city.  SO2 and
PM,0 analyses by temperature tertile
suggest that warm season effects
dominate.  Overall, results consistent with
earlier analyses for London, and
comparable with those for North America
and Europe.
PM,0 (50 Mg/m3) No co-pollutant:
All Respiratory ED visits
All age(lag ld)ER = 4.9% (CI: 1.3, 8.6)
<15yrs(lag2d)ER = 6.4%(CI: 1, 12.2)
15-64yr(lagld)ER = 8.6% (CI: 3.4, 14)
Asthma ED visits
All age (lag Id) ER = 8.9% (CI: 3, 15.2)
<15yrs (lag 2d) ER = 12.3% (CI: 3.4, 22)
15-64yr (Ig Id) ER = 13% (CI: 4.6, 22.1)

PM,(1(50 Mg/m3) 2d lag & co-pollutant:
Children's (<15 yrs.) Asthma ED Visits:
PM alone: ER= 12.3%(CI: 3.4, 22)
&NO2:  ER = 7.8% (CI: -1.2,17.6)
&O3:   ER=10.5%(CI: 1.6, 20.1)
&SO2:  ER= 8.1% (CI:-1.1, 18.2)
&CO:  ER= 12.1% (CI: 3.2, 21.7)

PM,0 (50 Aig/m3), no co-pollutant.
All Respiratory Admissions:
All age (lag Id) ER = 4.9% (CI: 1.8, 8.1)
0-14 y (lag Id) ER = 8.1% (CI: 3.5, 12.9)
15-64y (lag 2d) ER = 6.9% (CI: 2.1, 12.9)
65+ y (lag 3d) ER = 4.9% (CI: 0.8, 9.3)
Asthma Admissions:
All age (lag 3d) ER = 3.4% (CI: -1.8, 8.9)
0-14 y (lag 3d) ER = 5.4% (CI: -1.2, 12.5)
15-64y(lag3d)ER = 9.4%(CI: 1.1, 18.5)
65+y.(lag Od) ER = 12% (CI: -1.8,27.7)
COPD & Asthma Admissions (65+vrs.)
(lag 3d) ER = 8.6% (CI: 2.6,15)
Lower Respiratory Admissions (65+ yrs.)
(lag 3d) ER = 7.6% (CI: 0.9, 14.8).

-------
o
O
o
      TABLE 6-17 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                              	         AND HOSPITAL ADMISSIONS STUDIES	

Reference/Citation
Location, Duration
PM Index/Concentrations            Study Description:
                                                                                    Results and Comments
                                                                               PM Index, Lag, Excess Risk %,
                                                                               (95% CI = LCI, UCL) Co-Pollutants
H
6
o
2
s
o
o
 o
 t— I
 H
 m
          Europe (cont'd)

          Hajatetal. (1999)
          London, England (92 - 94)
          Population = 282,000
          PMU) mean = 28.2 /ug/m3
          PM1() 10'-90lh%=16.3-46.4^g/m3
          BSmean= 10.1 Mg/m3
          BS 10'-90'h%=4.5-15.9Mg/m3
Wordleyetal. (1997)
Study Period: 4/92 -3/94
Birmingham, UK
Population = NR
PM](I daily values:
Mean = 25.6 //g/m3
range = 2.8, 130.9^g/m3
PM]0 3 day running, mean:
Mean = 25.5 uglm*
range = 7.3, 104.7 /ug/m3
                                  Examined associations of PM,,,, BS, NO2,
                                  O3, SO2, and CO, with primary care general
                                  practitioner asthma and "other LRD"
                                  consultations.  Asthma consultation means
                                  per day = 35.3 (all ages); 14.(0-14 yrs,);
                                  17.7 (15-64 yrs.); 3.6 (>64 yrs.). LRD
                                  means = 155 (all ages); 39.7(0-14 yrs,);
                                  73.8 (15-64 yrs.); 41.1 (>64 yrs.). Time-
                                  series analyses of daily numbers of
                                  consultations performed, controlling for
                                  time trends, season factors, day of week,
                                  influenza, weather, pollen levels, and serial
                                  correlation.
Relation between PMH1 and total HA's for
respiratory (mean = 21.8/d), asthma
(mn.=6.2/d), bronchitis (mn.=2.4/d),
pneumonia (mn.=3.4/d), and COPD
(mn.=3.2/d) analyzed, using linear
regression after adjusting for day of week,
month, linear trend, RH, and T (but not T-
RH interaction). RR's compared for
various thresholds vs. mean risk of HA.
Positive associations, weakly significant
and consistent across lags, observed
between asthma consultations and NO2
and CO in children, and with PM10 in
adults, and between other LRD
consultations and SO2 in children.
Authors concluded that there are
associations between air pollution and
daily concentrations for asthma and other
lower respiratory disease in London. In
adults, the authors concluded that the only
consistent association was with PM,0.
Across all of the various age, cause, and
season categories considered, PMm was
the pollutant most coherent in giving
positive pollutant RR estimates for both
asthma and other LRD (11 of 12
categories positive) in single pollutant
models considered.

PMHI positively associated with all HA's
for respiratory, asthma, bronchitis,
pneumonia, and COPD.  Pneumonia,  all
respiratory, and asthma HA's also
significantly positively associated with the
mean of PMH, over the past three days,
which gave 10 to 20% greater RR's per 10
A«g/m3, as expected given smaller day to
day deviations. Other air pollutants
examined but not presented, as "these did
not have a significant association with
health outcomes independent from that of
PM,,,".
Asthma Doctor's Visits:
50 Mg/m3 PM,0
-Year-round, Single Pollutant:
All ages (Ig 2): ER = 5.4% (CI: -0.6, 11.7)
0-14 yrs.(lgl):ER = 6.4% (-1.5, 14.6)
15-64 yrs.(lg 0): ER = 9.2% (CI: 2.8, 15.9)
>64yrs.(lg 2): ER = 11.7% (-1.8, 26.9)
-Year-round, 2 Pollutant, Children (0, 14):
(PM,0 lag = 1 day) PMIO ER's:
W/NO2: ER = 0.8% (CI: - 8.7, 11.4)
W/O3: ER = 5.5%(-2.1, 13.8)
W/SO2: ER = 3.2% (CI: -6.4, 13.7)
Other Lower Resp. Pis. Doctor's Visits:
50Mg/m3PM10
-Year-round, Single Pollutant:
All ages (Ig 2): ER = 3.5% (CI: 0, 7.1)
0-14 yrs.(lg 1): ER = 4.2% (CI: -1.2, 9.9)
15-64 yrs.(lg 2): ER= 3.7% (CI: 0.0, 7.6)
>64yrs.(lg 2): ER = 6.2% (CI: 0.5, 12.9)

50 Mg/m3 in PMIO
All Respiratory HA's (all ages)
(lagOd) ER = 12.6% (CI: 5.7, 20)
Asthma HA's (all ages)
(Iag2d) ER = 17.6% (CI: 3, 34.4)
Bronchitis HA's (all ages)
(lagOd) ER= 32.6% (CI: 4.4, 68.3)
Pneumonia  HA's (all ages)
(Iag3d) ER= 31.9% (CI: 15, 51.4)
COPD HA's (all ages)
(lagld) ER = 11.5% (CI: -3, 28.2)

-------
I
O
O
       TABLE 6-17 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
=====^==	AND HOSPITAL ADMISSIONS STUDIES\	

 Reference/Citation
 Location, Duration
 PM Index/Concentrations            Study Description:
                                                                                   Results and Comments
                                                                               PM Index, Lag, Excess Risk %,
                                                                               (95% CI = LCI, UCL) Co-Pollutants
          Europe (cont'd)
Os
H
6
o
2
o
H
O
O
H
W
O
&
o
HH
51
          Prescott et al. (1998)
          Edinburgh (10/92-6/95)
          Population = 0.45 MM
          PM,0 mean. =20.7 ngjnf
          PM,0 min/max=5/72 //g/m3
          PM,0 90'h% - 10th% =
 McGregor et al. (1999)
 Birmingham, UK.
 Population = NR
 Mean PM,0 = 30.0
 Hagen et al. (2000)
 Drammen, Sweden( 11/94- 12/97)
 Population = 110,000
 PM,,, mean = 16.8
 PM10IQR = 9.8-20.9 /ug/m3
                                   Poisson log linear regression models used
                                   to investigate relation of daily HA's with
                                   NO2, O3, CO, and PM,0. Adjustments made
                                   for seasonal and weekday variation, daily T
                                   (using 8 dummy variables), and wind speed.
                                   Separate analyses for age<65 yr. (mean resp
                                   HA = 3.4/day) and age >64 yr. (mean resp
                                   HA = 8.7/day), and for subjects with
                                   multiple HA's.
A synoptic climatological approach used to
investigate linkages between air mass types
(weather situations), PM,0, and all
respiratory hospital admissions (mean=
19.2/day) for the Birmingham area.
Examined PM10, SO2, NO2, VOC's, and O3
associations with respiratory hospital
admissions from one hospital (mean =
2.2/day). Used Poisson GAM controlling
for temperature and RH (but not their
interaction), long-wave and seasonality,
day-of-week, holidays, and influenza
epidemics.
The two strongest findings were for
cardiovascular HA's of people aged >64,
which showed a positive association with
PM10 as a mean of the 3 previous days.
PM10 was consistently positively
associated with Respiratory HA's in both
age groups, with the greatest effect size in
those >64, especially among those with
>4 HA's during '81-'95. Weak
significances likely contributed to by low
population size.

Study results show distinct differential
responses of respiratory admission rates to
the six winter air mass types. Two of
three types of air masses associated with
above- average admission rates also favor
high PM|0 levels. This is suggestive of
possible linkage between weather, air
quality, and health.

As a single pollutant, the PM,0 effect was
of same order of magnitude as reported in
other studies. The PMHI association
decreased when other pollutants were
added to the model. However, the VOC's
showed the strongest associations.
                                                                               Single Pollutant Models
                                                                               PM10 = 50 Mg/ni3, mean of lags 1-3

                                                                               Respiratory HA's (age<65)
                                                                               ER = 1.25 (-12.8, 17.5)
                                                                               Respiratory HA's (age>64)
                                                                               ER = 5.33 (-9.3, 22.3)
                                                                               Respiratory HA's (age>64. >4 HA's)
                                                                               ER = 7.93 (-19.0, 43.7)
                                                                                                                          NR
Respiratory Hospital Admissions(all ages)
For IQR=50 ^g/m3
-Single Pollutant Model:
PM10 (lag 0) ER = 18.3% (CI: -4.2, 46)
-Two Pollutant Model (with O3):
PM10 (lag 0) ER = 18.3% (CI: -4.2, 45.4)
-Two Pollutant Model (with Benzene):
PM10 (lag 0) ER = 6.5% (CI:-14 , 31.8)

-------
o
to
o
o
               TABLE 6-17 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                            AND HOSPITAL ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
                                            Study Description:
                                        Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% CI = LCI, UCL) Co-Pollutants
ON
6
O
2
s
o
d
o
H
W
O
a
         Europe (cont'd)

         Dab etal. (1996)
         Paris, France (87 - 92)
         Population = 6.1 MM
         PM,3 mean = 50.8 //g/m3
         PM13 5'h-95'h range = 19.0-137.3
         BS mean = 31.9 ^g/m3
         BS 5lh-95IhRange=11.0-123.3
Medina etal. (1997)
Greater Paris 91 -95
Populations 6.5 MM
Mean PM13 = 25 /ug/m3
PM13 min/max = 6/95 /^g/m3
MeanBS = 21 /^g/m3
BS nun/max = 3/130 /ug/m3
Anderson et al. (1997)
Amsterdam(77 - 89)
Barcelona ( 86- 92)
London (87-91)
Milan ( 80- 89)
Paris ( 87 -  92)
Rotterdam ( 77 - 89)
Populations .= 0.7(A), 1.7(B),
7.2(L),1.5(M),6.5(P),0.6(R)MM
BS Means = 6, 41, 13,-,26, 22
TSP Means = 41,155,-, 105,-,41
Daily mortality and general admissions to
Paris public hospitals for respiratory causes
were considered (means/day:  all
resp.=79/d, asthma=14/d, COPD=12/d).
Time series analysis used linear regression
model followed by a Poisson regression.
Epidemics of influenza A and B,
temperature, RH, holidays, day of week,
trend, long-wave variability, and nurses'
strike variables included. No  two pollutant
models considered.

Evaluated short-term relationships between
PM,j and BS concentrations and doctors'
house calls (mean=8/day; 20% of city total)
in Greater Paris.  Poisson regression used,
with non-parametric smoothing  functions
controlling for time trend, seasonal patterns,
pollen counts, influenza epidemics, day-of-
week, holidays, and weather.
All-age daily hospital admissions (HA's)
for COPD considered in 6 APHEA cities;
Mean/day = 1.1 (A), 11(B), 20(L), 5(M),
11 (P), 1.1 (R). Poisson regression
controlling for day of week, holidays,
seasonal and other cycles, influenza
epidemics, temperature, RH, and
autocorrelation.  Overall multi-city
estimates made using inverse variance wts.,
allowing for inter-city variance.
                                                                           For the all respiratory causes category, the
                                                                           authors found "the strongest association
                                                                           was observed with PM,3" for both hospital
                                                                           admissions and mortality, indicating a
                                                                           coherence of association across outcomes.
                                                                           Asthma was significantly correlated with
                                                                           NO2 levels, but not PM13.
A relationship between all age (0-64 yrs.)
asthma house calls and PM,3, BS, SO2,
NO2, and O3 air pollution, especially for
children aged 0-14 (mean = 2/day).
In two-pollutant models including BS
with, successively, SO2, NO2, and O3, only
BS and O3 effects remained stable.  These
results also indicate that air pollutant
associations noted for hospital ED visits
are also applicable to a wider population
that visits their doctor.

Ozone gave the most consistent
associations across models. Multi-city
meta-estimates also indicated associations
for BS and TSP. The warm/cold season
RR differences were important only for
ozone, having a much stronger effect in
the warm season.  COPD effect sizes
found were much smaller than in U.S.
studies, possibly due to inclusion of non-
emergency admissions or use of less
health-relevant PM indices.
For PM,, = 50
                                                           : BS = 25 ,ag/m3
                                        Respiratory HA's (all ages):
                                        PM,3 Lag 0 ER = 2.2% (CI: 0.2, 4.3)
                                        BS Lag 0 ER = 1.0% (0.2, 1.8)
                                        COPD HA's (all  ages):
                                        PM13 Lag 2 ER = ~2.3% (CI: -6.7, 2.2)
                                        BS Lag 2 ER = -1.1% (-2.9, 0.6)
                                        Asthma HA's (all ages):
                                        PM,3 Lg 2 ER = -1.3% (CI: -4.6, 2.2)
                                        BS Lg 0 ER = 1 .2% (-0.5, 2.9)
Doctor's Asthma House Visits:
50 Aig/m3 PM,3
Year-round, Single Pollutant:
All ages (Ig 2): ER = 12.7% (CI: 4.1, 21.9)
0-14 yrs.(lg 0-3): ER = 41.5% (CI: 20, 66.8)
15-64yrs.(lg2): ER = 6.3%(CI: -4.6, 18.5)
BS (25 jug/m3) Id lag, no co-pollutant:
All Age COPD Hospital Admissions
ER= 1.7% (0.5, 2.97)
TSP (100 ^g/m3) Id lag, no co-pollutant:
All Age COPD Hospital Admissions
ER = 4.45%(CI: -0.53,9.67)

-------
P
O
O
      TABLE 6-17 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                 AND HOSPITAL ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
                                           Study Description:
                                        Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% CI = LCI, UCL) Co-Pollutants
ON

a\
Tl
H
O
O
2
O
H
O
c
O
H
w
O
w
o
HH
H
tn
         Europe (cont'd)

         Diaz etal. (1999)
         Madrid (94 - 96)
         Population = NR
         TSP mean = 40 ^g/
          Spix etal. (1998)
          London (L) (87 -91)
            Pop. =7.2 Million (MM)
Amsterdam (A) (77 - 89)
   Pop. =0.7 MM
   BS Mean = 6 Mg/m3
   TSP mean = 41 Aig/m3
Rotterdam (R) (77 - 89)
   Pop. =0.6MM
   BS Mean = 22 ^g/m3
   TSP mean = 41 ,ug/m3
Paris (P) (87 - 92),
   Pop.= 6.14MM
   BS Mean = 26 ^g/m3
Milano (M) (80 - 89)
   Pop. = 1.5 MM
   TSPMean=120(,ug/m3)
ARIMA modeling used to analyze
emergency respiratory and circulatory
admissions (means/day=7.8,7.6) from one
teaching hospital.  Annual, weekly, and 3
day periodicities controlled, but no time
trend included, and temperature crudely fit
with v-shaped linear relationship.


Respiratory (ICD9 460-519) HA's in age
groups 15-64 yr and 65 + yrs. related to
SO2, PM (BS or TSP), O3, and  NO2 in the
APHEA study cities using standardized
Poisson models with confounder controls
for day of week, holidays, seasonal and
other cycles, temperature, RH,  and
autocorrelation. PM lag considered ranged
from 0-3 day, but  varied from city to city.
Quantitative pooling conducted by
calculating the weighted means of local
regression coefficients using a  fixed-effects
model when no heterogeneity could be
detected; otherwise, a random-effects model
employed.
Although TSP correlated at zero lag with
admissions in winter and year-round, TSP
was never significant in ARIMA models;
so effect estimates not reported for TSP.
Also, found biologically implausible u-
shaped relationship for O3, possibly
indicating unaddressed temperature
effects.

Pollutant associations noted to be stronger
in areas where more than one monitoring
station was used for assessment of daily
exposure. The most consistent finding
was an increase of daily HA's for
respiratory diseases (adults and elderly)
with O3. The SO2 daily mean was
available in all cities, but SO2 was not
associated consistently with adverse
effects. Some significant PM associations
were seen, although no  conclusion related
to an overall particle effect could be
drawn. The effect of BS was significantly
stronger with high NO2 levels on the same
day, but NO2 itself was  not associated with
HA's.  Authors concluded that "there was
a tendency toward an association of
respiratory admissions with BS, but the
very limited number of cities prevented
final conclusions."
                                                                                                                N/A
Respiratory Admissions (BS = 25 ,ug/m3)
BS (L, A, R, P)
15-64 yrs: 1.4% (0.3, 2.5)
  65+yrs: 1.0% (-0.2, 2.2)
TSP (A, R, M)(100Aig/m3)
15-64 yrs: 2.0 (-2.1, 6.3)
  65+yrs: 3.2 (-1.2, 7.9)
Respiratory HA's
BS (L, A, R, P): Warm (25 /^g/m3)
15-64 yrs:-0.5% (-5.2, 4.4)
  65+yrs: 3.4% (-0.1, 7.1)
BS (L, A, R, P): Cold (25 jug/ni3)
15-64 yrs: 2,0% (0.8, 3.2)
  65+yrs: 0% (-2.2, 2.3)
TSP (A, R, M): Warm (100 ,ug/m3)
15-64 yrs: 6.1% (0.1, 12.5)
  65+yrs: 2.0% (-3.9, 8.3)
TSP (A, R, M): Cold (100 /^g/m3)
15-64 yrs:-5.9% (-14.2, 3.2)
  65+yrs: 4.0% (-0.9, 9.2)

-------
O

N>
O
O
               TABLE 6-17 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                            AND HOSPITAL ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
                                            Study Description:
                                        Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% CI = LCI, UCL) Co-Pollutants
•n
H
6
o
z
3
O
H
W
o
&
o
H
m
         Europe (cont'd)

         Vigottietal. (1996)
         Study Period.: 80-89
         Milan, IT
         Population = 1.5 MM
         TSPmean= 139.0,ug/m3
         TSP 1QR = 82.0, 175.7 /ug/m3
          Anderson et al. (1998)
          London (87 - 92)
          Population = 7.2 MM
          BS daily mean = 14.
          BS25-75thIQR = 24-38
Damiaetal. (1999)
Valencia, Spain (3/94-3/95)
Population = NR
BSmean= 101 Aig/m3
BS range = 34-213 Mg/m3
Association between adult respiratory HA's
(15-64 yr mean =11.3/day, and 65 + yr
mean =8.8/day) and air pollution evaluated,
using the APHEA protocol. Poisson
regression used with control for weather
and long term trend, year, influenza
epidemics, and season

Poisson regression used to estimate the RR
of London daily asthma hospital admissions
associated with changes in O3, SO2, NO2
and particles (BS) for all ages and for 0-14
yr. (mean=19.5/d), 15-64 yr (mean=13.1/d)
and 65 + yr. (mean =2.6/d). Analysis
controlled for time trends, seasonal factors,
calendar effects, influenza epidemics, RH,
temperature, and auto-correlation.
Interactions with co-pollutants and
aeroallergens tested via 2 pollutant models
and models with pollen counts (grass, oak
and birch).

Associations of BS and SO2 with weekly
total ED admissions for asthma patients
aged > 12 yrs (mean = 10/week) at one
hospital over one year assessed, using linear
stepwise regression. Season-specific
analyses done for each of 4 seasons, but no
other long-wave controls.  Linear T, RH,
BP, rain, and wind speed included as crude
weather controls in ANOVA models.
Increased risk of respiratory HA was
associated with both SO2 and TSP. The
relative risks were similar for both
pollutants. There was no modification of
the TSP effect by SO2 level. There was a
suggestion of a higher TSP effect on
hospital admissions in the cool months.

Daily hospital admissions for asthma
found to have associations with O3, SO2,
NO2, and particles (BS), but there was
lack of consistency across the age groups
in the specific pollutant. BS association
was strongest in the 65 + group, especially
in winter. Pollens not consistently
associated with asthma HA's, sometimes
being positive, sometimes negative. Air
pollution associations with HA's not
explained by airborne pollens in
simultaneous regressions, and there was
no consistent pollen-pollutant interaction.

Both BS and SO2 correlated with ED
admissions for asthma (SO2: r=0.32; BS:
r=0.35), but only BS  significant in
stepwise multiple regression.  No linear
relationship found with weather variables.
Stratified ANOVA found strongest BS-ED
association in the autumn and during
above average temperatures. Uncontrolled
autocorrelation (e.g.,  within-season) and
weather effects likely remain in models.
Young Adult (15-64 vrs.) Resp. HA's
100 //g/m3 increase in TSP
Lag2ER=5%(CI:0,10)

Older Adult (65+ vrs.) Resp. HA's
100 ^g/m3 increase in TSP
Lagl ER = 5%(CI: -1,  10)

Asthma Admissions.  BS=25 ueJm1
BS Lag = 0-3 day average concentration
All age ER = 5.98% (0.4, 11.9)
<15yr. ER = 2.2%(-4.6, 9.5)
15-64yrER=1.2%(-5.3, 8.1)
65+yr. ER = 22.8% (6.1, 42.5)

BS=50 /ug/m3, 2d lag & co-pollutant:
Older Adult (>64 vrs.) Asthma Visits:
BS alone:  ER = 14.6% (2.7, 27.8)
&O3:     ER = 20.0% (3.0, 39.8)
&NO2:    ER = 7.4% (-8.7, 26.5)
SO2:      ER=11.8% (-2.2, 27.8)

Asthma ED Visits (all ages):
BS = 40 Atg/m3 (single pollutant)
BS as a lag 0 weekly average:
ER = 41.5% (CI = 39.1,  43.9)

-------

O
O
               TABLE 6-17 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                            AND HOSPITAL ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
                                            Study Description:
                                        Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% CI = LCI, UCL) Co-Pollutants
T1
6
O
2
O
H
O
c
O
H
m
O
*>
n
HH
H
W
          Europe (cont'd)

          Kontosetal. (1999)
          Piraeus, Athens GR (87 - 92)
          Population = NR
          BS mean =46.5 /ug/m3
          BS max =200 Mg/m3
          Pantazopoulou et al. (1995)
          Athens, GR (1988)
          Population = NR
          Winter (1/88-3/88,9/88-12/88)
          BS mean. =75 jUg/m3
          BS 5th-95'h %=26 - 161 A^g/ni3
          Summer (3/22/88-3/88,9/21/88)
          BS mean. =55 ,ug/m3
          BS5lh-95th%=19-90Aig/m3
Ponce de Leon et al. (1996)
London (4/87-2/92)
Population = 7.3 million
BS mean. =14.6 ^g/m3
BS 5'h-95lh %=6 - 27 //g/m3
                                  Relation of respiratory HA's for children
                                  (0-14 yrs) (mean = 4.3/day) to BS, SO2,
                                  NO2, and O3 evaluated, using a
                                  nonparametnc stochastic dynamical system
                                  approach and frequency domain analyses.
                                  Long wave and effects of weather
                                  considered, but non-linearity and
                                  interactions of T and RH relation with HA's
                                  not addressed.

                                  Examined  effects of air pollution on daily
                                  emergency outpatient visits and admissions
                                  for cardiac and respiratory causes.  Air
                                  pollutants  included: BS, CO, and NO2.
                                  Multiple linear regression models used,
                                  controlling for linear effects of temperature
                                  and RH, day of week, holidays, and dummy
                                  variables for month to crudely control for
                                  season, separately for winter and summer.
Poisson regression analysis of daily counts
of HA's (means/day:  all ages=125.7; Ages
0-14=45.4; Ages 15-64=33.6; Ages
65+=46.7).  Effects of trend, season and
other cyclical factors, day of the week,
holidays, influenza epidemic, temperature,
humidity, and
autocorrelation addressed.  However,
temperature modeled as linear, with no RH
interaction.  Pollution variables were BS,
SO2, O3, and NO2, lagged 0-3 days.
                                        Pollution found to explain significant
                                        portion of the HA variance.  Of pollutants
                                        considered, BS was consistently among
                                        most strongly explanatory pollutants
                                        across various reported analyses.
Daily number of emergency visits related
positively with each air pollutant, but only
reached nominal level of statistical
significance for N02 in winter. However,
the very limited time for each within-
season  analysis (6 mo.) undoubtably
limited the power of this analysis to detect
significant effects.  Also, possible lagged
pollution effects were apparently not
investigated, which may have reduced
effect estimates.

O3 associated with increase in daily HA's,
especially in the "warm" season.
However, u-shape of the O3 dose-response
suggests that linear temperature control
was not adequate.  Few significant
associations with other pollutants, but
these tended to be positive (especially in
cold season,  Oct-March, and for older
individuals for BS).
                                                                                                                  NR
Single Pollutant Models
For Winter (BS = 25 /^g/rn')
Outpatient Hospital Visits
ER = 1.1% (-0.7, 2.3)
Respiratory HA's
ER = 4.3% (0.2, 8.3)
For Summer, BS = 25 (Ug/m3)
Outpatient Hospital Visits
ER = 0.6% (-4.7, 6.0))
Respiratory HA's
ER = 5.5% (-3.6, 14.7)

Respiratory HA's (all ages)
Single Pollutant Models
For Oct-Mar. BS = 25 ^g/m3
Lagl ER = 0.2% (-1.9, 2.3)
For Apr-Sep. BS = 25 //g/m3
Lagl ER =-2.7% (-6.0, 0.8)

Respiratory HA's (>65)
Single Pollutant Models
For Oct-Mar. BS = 25 /^g/m3
Lag 2 ER= 1.2% (-2.1,4.5)
For Apr-Sep. BS = 25 ^g/m3
Lag 2 ER = 4.5% (-1.0, 10.4)

-------
o
sr
O
O
       TABLE 6-17 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
	AND HOSPITAL ADMISSIONS STUDIES	

 Reference/Citation
 Location, Duration
 PM Index/Concentrations            Study Description:
                                                                                   Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% CI = LCI, UCL) Co-Pollutants
 Tl
 H
 b
 o
 2
 3
/O
 C
 O
 H
 M
 O
 70
 O
 H-H
 H
 W
          Europe (cont'd)

          Schouten et al. (1996)
          Amsterdam/Rotterdam (77 - 89)
          Amsterdam Pop. = 0.69 Million
          Rotterdam Pop. =  0.58 Million
          Amsterdam, NE
          BS mean. =11 ,ug/m3
          BS Slh-95'h% =1-37 Atg/m3
          Rotterdam, NE
          BS mean. =26 Mg/m3
          BS5'h-95'h%=6-61,ug/m3
          Gartyetal. (1998)
          PMHI mean = 45 /^g/m3
          Tel Aviv, Israel (1993)
                                   Daily emergency HA's for respiratory
                                   diseases (ICD 460-519), COPD (490-492,
                                   494, 496), and asthma (493). The mean
                                   HA/d (range) for these were: 6.70 (0-23),
                                   1.74 (0-9) and 1.13 (0-7) respectively in
                                   Amsterdam and 4.79 (0-19), 1.57 (0-9),
                                   and 0.53 (0-5) in Rotterdam. HA
                                   associations with BS, O3, NO,, and SO2
                                   analyzed, using autoregressive Poisson
                                   regression allowing for overdispersion and
                                   controlling for season,  day of week,
                                   meteorological factors, and influenza
                                   epidemics.
                                   Seven day running mean of asthma ED
                                   visits by children (1-18 yrs.) to a pediatnc
                                   hospital modeled in relation to PMH, in Tel
                                   Aviv, Israel.
BS did not show any consistent effects in
Amsterdam; but in Rotterdam BS was
positively related to  HA's. Most
consistent BS associations in adults >64
yrs. in winter. Positive O3 association in
summer in people aged >64 in Amsterdam
and Rotterdam. SO2 and NO2 did not
show any clear effects. Results not
changed in pollutant interaction analyses.
The authors concluded short-term air
pollution-emergency HA's association is
not always consistent at these individual
cities' relatively low counts of daily HA's
and low levels of air pollution. Analyses
for all ages of all the Netherlands gave a
strong BS-HA association in winter.

No PMio associations found with ED
visits. The ER visits-pollutant correlation
increased significantly when the
September peak was excluded. Use of a
week-long average and associated
uncontrolled long-wave fluctuations (with
resultant autocorrelation) likely prevented
meaningful analyses of short -term PM
associations with ED visits.
Single Pollutant Models
For BS=25 ^g/m3, 2 day lag
For all of the Netherlands:
Respiratory HA's (all ages)
Winter:
ER = 2.0% (-1.5, 5.7)
Summer:
ER = 2.4% (0.6, 4.3)
N/A

-------
o
CT
K)
O
O
       TABLE 6-17 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
	AND HOSPITAL ADMISSIONS STUDIES	

 Reference/Citation
 Location, Duration
 PM Index/Concentrations           Study Description:
                                                                                  Results and Comments
                                                                              PM Index, Lag, Excess Risk %,
                                                                              (95% CI = LCI, UCL) Co-Pollutants
  s
O
Z
O
H
/O
s
H
 O
 HH
 H
 W
 Europe (cont'd)

 Sunyeretal. (1997)
 Barcelona (86 - 92)
    Population = NR
    BS Median:  40 ^g/m3
    BS Range: 11-258 (B
 Helsinki (86 - 92)
    Population = NR
    BS Median:  -
    BS Range: -
 Paris (86 - 92)
    Population = NR
    BS Median:  28 ,ug/m3
    BS Range: 4-186^g/m3
 London (86 - 92)
    Population = NR
    BS Median:  13/ug/m3
    BS Range: 3-95 ^g/m3
 Teniasetal (1998)
 Study Period.: 94-95
 Valencia, Spain
 Hosp. Cachment Pop. =200,000
 BS mean = 57.7 ,ug/m3
 BS IQR = 25.6-47.7
Evaluated relations of BS, S02, NO2, and
O3 to daily counts of asthma HA's and ED
visits in adults [ages 15-64 years: mean/day
= 3.9 (B); 0.7 (H); 13.1 (H); 7.3 (P)] and
children [ages < 15 years: mean/day = 0.9
(H);19.8(L);4.6(P)]. Asthma
(ICD9=493) studied in each city, but the
outcome examined differed across cities:
ED visits in Barcelona; emergency hospital
asthma admissions in London and Helsinki,
and total asthma admissions in  Pans.
Estimates from all cities obtained for entire
period and also by warm or cold seasons,
using Poisson time-series regression,
controlling for temperature and RH, viral
epidemics, day of week effects, and
seasonal and secular trends. Combined
associations were estimated using meta-
analysis.
Associations between adult (14+ yrs.)
emergency asthma ED visits to one city
hospital (mean =1.0/day) and BS, NO2, O3,
SO2 analyzed, using Poisson auto-
regressive modeling, controlling for
potential confounding weather and time
(e.g., seasonal) and trends using the
APHEA protocol.
Daily admissions for asthma in adults
increased significantly with increasing
ambient levels of NO2, and positively (but
non-sigmficantly) with BS. The
association between asthma admissions
and pollution varied across cities, likely
due to differing asthma outcomes
considered.  In children, daily admissions
increased significantly with SO2 and
positively (but non-significantly) with BS
and NO2, though the latter only in cold
seasons. No association observed in
children for O3. Authors concluded that
"In addition to particles, NO2 and SO2 (by
themselves or as a constituent of a
pollution mixture) may be important in
asthma exacerbations".
Association with asthma was positive and
more consistent for NO2 and O3 than for
BS or SO2.  Suggests that secondary
oxidative-environment pollutants may be
more asthma relevant than primary
reduction-environment pollutants (e.g.,
carbonaceous particles). NO2 had greatest
effect on BS in co-pollutant models, but
BS became significant once 1993 was
added, showing power to be a limitation of
this study.
ER per 25 ^g/m3 BS (24 h Average)
Asthma Admissions/Visits:
<15 yrs.:
  London ER = 1.5% (Ig Od)
  PansER=1.5%(lg2d)
  Total ER=1.5%(-1.1,4.1)
15-64 yrs:
  Barcelona ER = 1.8% (Ig 3d)
  London ER = 1.7% (Ig Od)
  Pans ER = 0.6% (Ig Od)
  Total ER=  1.0% (-0.8, 2.9)
Two Pollutant (per 25 Mg/m3 BS)
Asthma Admissions (24 h Avg)
<15yrs, (BS&NO2):
  London ER = 0.6% (Ig Od)
  Paris ER = 2.9% (Ig 2d)
  Total ER =  1.8% (-0.6, 4.3)
<15yrs,(BS&SO2):
  London ER=-1.1% (Ig Od)
  ParisER=-1.4%(lg2d)
  Total ER=-1.3 (-5.0,  2.5)
15-64 yrs, (BS & NO2):
  Barcelona ER = 1.5% (Ig Od)
  London ER = -4.7% (Ig Od)
  Paris ER=-0.7% (Ig Id)
  Total ER =  -0.5% (-5.1,4.4)

Adult Asthma HA's. BS =
For 1993-1995:
Lag OER= 10.6% (0.9, 21.1)
For 1994-1995:
Lag OER = 6.4% (-4.8, 18.8)

-------
g.
tsJ
O
o
      TABLE 6-17 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                   AND HOSPITAL ADMISSIONS STUDIES  	^^^
Reference/Citation
Location, Duration
PM Index/Concentrations            Study Description:
                                                                                     Results and Comments
                                                                                 PM Index, Lag, Excess Risk %,
                                                                                 (95% CI = LCI, UCL) Co-Pollutants
O\
ON
OO
Tl
H
6
o
2
o
H
O
c!
O
H
W
O
&
o
          Latin America

          Bragaetal. (1999)
          Sao Paulo, Brazil (92 - 93)
          Population = NR
          PM|(I mean = 66.3 ,ug/m3
          PMH)Std. Deviation = 26.1
          PMU1 Min./Max. = 26.7/165.4
Gouveia and Fletcher (2000)
Study Period. :92-94
Sao Paulo, Brazil
Population = 9.5 MM x 66%
PM,,, mean = 64.9 ,wg/m3
PM10IQR = 42.9-75.5 ^g/m3
PMl(110/90'h%=98.1 //g/m3
PM,(195'h%= 131.6 Mg/m3
Lin etal. (1999)
Sao Paulo, BR (91 - 93)
Population = NR
PM]0 mean =65 ^g/m3
PM10SD = 27/ug/m3
PMH) range = 15-193 ,ug/m3
Pediatnc (<13 yrs.) hospital admissions
(mean=67.6/day) to public hospitals serving
40% of the population were regressed
(using both Poisson and maximum
likelihood methods) on air pollutants,
controlling for month of the year, day-of-
week, weather, and the daily number of
non-respiratory admissions
(mean=120.7/day). Air pollutants
considered included PMI(I, O3, SO2, CO,
and NO2.

Daily public hospital respiratory disease
admissions for children (mean resp. < 5y =
56.1/d; mean pneumonia <5y =40.8/d;
mean asthma <5 y = 8.5/d; mean
pneum.
-------
B
o

o
o
               TABLE 6-17 (cont'd). ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                          AND HOSPITAL ADMISSIONS STUDIES
         Reference/Citation
         Location, Duration
         PM Index/Concentrations
                                           Study Description:
                                                                                  Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% CI = LCI, UCL) Co-Pollutants
Tl
H
6
o
2
o
H
O
c
o
         Latin America (cont'd)
         Ostroetal. (1999b)
         Santiago, CI (7/92—12/93)
         <2 yrs. Population = 20,800
         3-14 yrs. Population = 128,000
         PM,0 mean. =108.6 //g/m3
         PM,0 Min/Max=18.5/380 Mg/m3
         PM10IQR = 70.3 - 135.5 //g/m3
         Rosas etal. (1998)
         SW Mexico City (1991)
         Population = NR
         PM|() mean. =77 AJg/nv1
         PMIO min/max= 25/183 /
                                           Analysis of daily visits to primary health
                                           care clinics for upper (URS) or lower
                                           respiratory symptoms (LRS) for children 2-
                                           14 yr (mean LRS=111 .I/day) and < age 2
                                           (mean LRS=104 3/day). Daily PM10 and O3
                                           and meteorological variables considered.
                                           The multiple regression GAM included
                                           controls for seasonality (LOESS smooth),
                                           temperature, day of week, and month.
                                           Log-regression analysis of relations
                                           between emergency hospital admissions for
                                           asthma for children <15 yrs
                                           (mean=2.5/day), adults (mean=3.0/day),
                                           and adults >59 yrs (mean=0.65/day) and lag
                                           0-2 d pollen, fungal spores, air pollutants
                                           (O3, NO2, SO2, and PM1()) and weather
                                           factors.  Long wave controlled only by
                                           separating the year into two seasons:  "dry"
                                           and "wet". Day-of-week not included in
                                           models.
Analyses indicated an association between
PMH, and medical visits for LRS in
children ages 2-14 and in children unde.
age 2 yr. PM1(I was not related to non-
respiratory visits (mean =208/day).
Results unchanged by eliminating high
PM10 (>235 Mg/m3) or coldest days
(<8°C). Adding O3 to the model had little
effect on PMm-LRS associations.
Few statistical associations were found
between asthma admissions and air
pollutants. Grass pollen was associated
with child and adult admissions, and
fungal spores with child admissions.
Authors conclude that aeroallergens may
be more strongly associated with asthma
than air pollutants, and  may act as
confounding factors in epidemiologic
studies. Results are limited by low power
and the lack of long-wave auto-correlation
controls in the models.
                                                                                                                         Lower Resp. Symptoms Clinic Visits
   Single Pollutant Models:
-Children<2 years
Lag 3 ER = 2.5% (CI: 0.2, 4 8)
-Children 2- 14 years
Lag 3 ER = 3.7% (CI: 0.8, 6.7%)
   Two Pollutant Models (with 03):
-Children<2 years
Lag 3 ER = 2.2% (CI: 0, 4.4)
-Children 2- 14 years
Lag3ER = 3.7%(CI:0.9, 6.5)

NR
 o
 H
 w

-------
o
K>
o
o
                TABLE 6-17 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
                                                            AND HOSPITAL ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
                                            Study Description-
                                                                          Results and Comments
                                       PM Index, Lag, Excess Risk %,
                                       (95% CI = LCI, UCL) Co-Pollutants
 O
 H
O

 O
 H
 O
 HH
 H
 W
          Australia
          Smith etal. (1996)
          StdyPd.:  12/92-1/93,12/93-1/94
          West Sydney, AU
          Population = 907,000
          -Period 1 (12/92-1/93)
          Bbtdtt median = 0.25 KT4/m
          BSLal,IQR = 0. 18-0.39 10-4/m
          Bsum 95'"% = 0.861 0~4/m
          -Period 2 (12/93-1/94)
          B,tan median = 0.19 10~4/m
          BsumIQR = 0.1 -0.38 10-4/m
          Biuu, 95lh5% = 3.26 10'4/m PMH)
          median =
          PMI()IQR =11. 5-28.8 /
          Morgan etal. (1998)
          Sydney, AU (90 - 94)
          Population = NR
          PM2 5 24 h mean = 9.
          PM2, 10th-90'h% = 3.6-18
          PM25 max-1 h mean = 22.8 ^g/m3
          PM25 10lh-90th% = 7.5-44.4 /ug/m3
                                  Study evaluated whether asthma visits to
                                  emergency departments (ED) in western
                                  Sydney (mean«10/day) increased as result
                                  of bushfire-generated PM ( B5altl from
                                  nephelometry) in Jan., 1994 (period 2) Air
                                  pollution data included nephelometry
                                  (Bscalt), PMU), SO,, and NO2. Data analyzed
                                  using two methods: (1) calculation of the
                                  difference in proportion of all asthma ED
                                  visits between the time periods, and; (2)
                                  Poisson regression analyses. Control
                                  variables included T, RH, BP, WS, and
                                  rainfall.
                                  A Poisson analysis, controlled for
                                  overdispersion and autocorrelation via
                                  GEE, of asthma (means: 0-14
                                  yrs.=15.5/day; 15-64=9/day)), COPD (mean
                                  65+yrs =9.7/day), and heart disease HA's.
                                  PM2 5 estimated from nephelometry.
                                  Season and weather controlled using
                                  dummy variables.
No difference found in the proportion of
all asthma ED visits during a week of
bushfire-generated air pollution, compared
with the same week  12 months before,
after adjusting for baseline changes over
the 12-month period. The max. B,^,,
reading was not a significant predictor of
the daily asthma ED visits in Poisson
regressions.  However, no long-wave
controls applied, other than indep. vars.,
and the power to detect differences was
weak (90% for a 50% difference). Thus,
the lack of a difference may be due to low
statistical strength or to lower toxicity of
particles from burning vegetation at
ambient conditions vs. fossil fuel
combustion.

Childhood asthma was primarily
associated with NO2, while COPD was
associated with both NO2  and PM.  1 -hr.
max PM2 5 more consistently positively
related to respiratory HA's than 24-h avg
PM2 5. Adding all other pollutants
lowered PM effect sizes, although
pollutant inter-correlations makes many
pollutant model interpretations difficult.
No association found between asthma and
O3 or PM. The authors cited the error
introduced by estimating PM2 5 and the
low PM levels as possible reasons for the
weak PM-respiratory HA  associations.
ED Asthma Visits (all ages)
Percent change between bushfire and non
bushfire weeks:
ER = 2.1%(CI: -0.2,4.5)
Asthma HA's
Single Pollutant Model:
1-14 yrs.(lagl) ER = -1.5% (CI: -7.8, 5.3)
1 5-64 yrs.(lagO) ER = 2.3% (CI: -4, 9)
ForlhPM2S=25Mg/m3
1-14 yrs.(lagl) ER = + 0.5% (CI: - 1.9, 3.0)
15-64 yrs.(lagO) ER = 1.5% (CI: -0.9, 4)
Multiple Pollutant Model:
                                                                                                                  1-14 yrs (lagl) ER= -0.6% (CI: -7.4, 6.7)
                                                                                                                  COPD (65+vrs.)
                                                                                                                  Single Pollutant Model:
                                                                                                                 (IagO)ER=4.2%(CI:-1.5, 10.3)
                                                                                                                 ForlhPM25 = 25,ug/m3
                                                                                                                 (lag 0) ER = 2% (CI: -0.3, 4.4)
                                                                                                                 Multiple Pollutant Model:
                                                                                                                 ForlhPM25 = 25/ug/m3
                                                                                                                 (lagO)ER=1.5%(CI: -0.9,4)

-------
O
o
       TABLE 6-17 (cont'd).  ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
	                                     AND HOSPITAL ADMISSIONS STUDIES

 Reference/Citation
 Location, Duration
 PM Index/Concentrations            Study Description.
                                                                                    Results and Comments
                                                                               PM Index, Lag, Excess Risk %,
                                                                               (95% CI = LCI, UCL) Co-Pollutants
 H
 b
 o
 z
 o
 H
XD
 O
 H
 O
 HH
 H
 W
          Asia

          Tanakaetal. (1998)
          StdyPd.: 1/92-12/93
          Kushiro, Japan
          Pop. = 102 adult asthmatics
          PM,o mean = 24.0
          PM,0 IQR = NR
          Wong etal. (1999)
          Study Period.: 94-95
          Hong Kong
          Population = NR
          PM10 mean = 50.1 A1.0 for both atopies and
non-atopies.  PMH, associated with a
reduction in risk (OR<1.0) for both
atopies and non-atopies.  Poisson
regression gave a non-significant effect by
PM,,, on asthma HA's. However, no long-
wave or serial auto-correlation controls
applied, so the opposing  seasonahties of
PM vs. HA's indicated in time series data
plots are likely confounding these results.

Positive associations were found for HA's
for all respiratory diseases and COPD with
all four pollutants.  PMU, results for lags
0-3 cumulative.  Admissions for asthma,
pneumonia, and influenza were associated
with NO2, O3, and PM,0. Those aged > or
= 65 years were at higher risk, except for
PM10. No significant respiratory HA
interactions with PMU, effect were found
for high NO2, high O3, or cold season.
Positive associations found between TSP,
SO2, and NO2, and daily HA and ED visits
for asthma in children, but only with ED
visits-among adolescents. Lack of power
(low counts) for adolescents' HA's
appears to have been a factor in the lack of
associations. When ED visits stratified by
year, SO2 and TSP remained associated in
every year, but not NO2. Analyses for
control diseases (appendicitis and urinary
tract infections)  found no associations.
                                                                                For same-day (lag=0) PM,,,
                                                                                Adult Asthma HA's
                                                                                OR for <30 vs >30 Aig/m3 PM1():
                                                                                Non-atopic OR = 0.77 (CI: 0.61, 0.98)
                                                                                Atopic OR = 0.87 (CI: 0.75,  1.02)

                                                                                Poisson Coefficient for PMH, > 30
                                                                                Non-atopic B = -0.01 (SE = 0.15)
                                                                                Atopic B = -0.002 (SE = 0.09)
PMI() = 50 A
-------
  1      distributed lag model, the maximum lag model is deemed here to provide the closest available
  2      estimate of the full pollutant-health effects impact.
  3          Among the numerous epidemiological studies published on PM:o morbidity, many
  4      evaluated effects of relatively high PMIO concentrations. This likely reflects the fact that large
  5      populations are required to have enough power for such studies, and large population areas may
  6      tend to have higher PM10 levels, thus reducing the number of cities available for such
  7      evaluations.  Despite this, quite a number did evaluate associations at low PM10 concentration
  8      levels and associations between acute PM10 exposures and total respiratory-related hospital
  9      admissions have been reported by several investigators for numerous U.S. cities with annual
10      mean ambient concentrations extending to below 50 /wg/m3.
11          The recent NMMAPS multi-city study (Samet et al., 2000a,b) of PM10 concentrations and
12      hospital admissions by persons 65 and older in 14 U.S. cities is of particular interest. As noted in
13      Table  6-17 and shown in more detail in Table 6-18, this study indicates PM10 effects similar to
14      other cities, but with narrower confidence bands, due to its greater power derived by combining
15      multiple cities in the same analysis.  This allows significant associations to be identified, despite
16      the fact that many  of the cities considered have relatively small populations and that each of the
17      14 cities had mean PM10 below 50 ,ug/m3. The cities considered and their respective annual
18      mean/daily maximum. PM,0 concentrations (in/^g/m3) are:  Birmingham (34.8/124.8); Boulder
19      (24.4/125.0); Canton (28.4/94.8); Chicago (36.4/144.7); Colorado Springs (26.9/147.2); Detroit
20      (36.8/133.6);Minneapolis/St Paul (36.8/133.6); Nashville (31.6/128.0); New Haven (29.3/95.4);
21      Pittsburgh (36.0/139.3); Provo/Orem (38.9/241.0); Seattle (31.0/145.9); Spokane (45.3/605.8);
22      and Youngstown (33.1/104.0). As seen in Table 6-18, the PM10 association remained even when
23      only those days with PM10 less than 50 //g/m3 were considered.
24          Other U.S. studies finding associations of respiratory-related hospital admissions or medical
25      visits with PM,0 levels extending below 50 //g/m? include: Schwartz (1995) in Tacoma;
26      Schwartz (1994b)  in Minneapolis; Schwartz et al. (1996b) in Cleveland; Sheppard et al. (1999)
27      in Seattle; Gwynn  et al. (2000) in Buffalo, NY; Linn et al. (2000) in Los Angeles, Nauenberg and
28      Basu (1999) in Los Angeles; and Moolgavkar et al. (1997) in Minneapolis-St. Paul, MN, but not
29      in Birmingham, AL.  The excess risk estimates appear to most consistently fall in the range of
30      5-25% per 50 /ug/m3 PM10 increment, with those for asthma visits and hospital admissions
31      usually being higher than for COPD and pneumonia hospital admissions.

        March 2001                               6-172       DRAFT-DO NOT QUOTE OR CITE

-------
           TABLE 6-18.  PERCENT INCREASE IN HOSPITAL ADMISSIONS PER 10-//g/m3
                                INCREASE IN PM,,, IN 14 U.S. CITIES


% Increase
CVD
(95% CI)
COPD
%Increase
(95% CI)
Pneumonia
% Increase
(95% CI)
Constratined lag models (Fixed Effect Estimates)
One day mean (lag 0)
Previous day mean
Two day mean
(for lag 0 and 1 )
PM10<50,wg,m3
(two day mean)
Quadratic distributed lag
Unconstrained distributed
Fixed effects estimate
Random effects estimate
1.07
0.68
1.17
1.47
1.18
Lag
1.19
1.07
(0.93, 1.22)
(0.54,0.81)
(1.01, 1.33)
(1.18, 1.76)
(0.96, 1.39)

(0.97, 1.41)
(0.67, 1.46)
1.44
1.46
1.98
2.63
2.49

2.45
2.88
(1 00, 1.89)
(1.03, 1.88)
(1.49,2.47)
(1.71,3.55)
(1.78,3.20)

(1.75,3.17)
(0.19,5.64)
1.57
1.31
1.98
2.84
1.68

1.90
2.07
(1.27, 1.87)
(1.03, 1.58)
(1.65,2.31)
(2.21,3.48)
(1.25,2.11)

(1.46,2.34)
(0.94, 3.22)
         Source: Samet et al (2000a,b)
 1          Similar associations between increased respiratory related hospital admissions/medical
 2      visits and relatively low short-term PMIO levels were also reported by various investigators for
 3      several non-U.S. cities. Wordley et al. (1997), for example, reported positive and significant
 4      associations between PMi0 (mean = 25.6 Aig/m3, max. = 131 /ug/m3) and respiratory admissions in
 5      Birmingham, UK; and Atkinson et al. (1999b) found significant increases in hospital admissions
 6      for respiratory disease to be associated with PM,0 (mean = 28.5 /^g/m3) in London, UK. Hagen
 7      et al. (2000) and Prescott et al. (1998) also found positive but non-significant PM10 associations
 8      with hospital admissions in Drammen, Sweden (mean =16.8 //g/m3) and Edinburgh, Scotland
 9      (mean = 20.7 /ug/m3), respectively; but both considered relatively small populations, limiting
10      statistical power in these studies.
11
12      6.3.2.3.1 Paniculate Matter Mass Fractions and Composition  Comparisons
13          While PM10 mass is the metric most often employed as the particle pollution index in the
14      U.S. and Canada, some new studies have begun to examine the relative roles of various PM10
15      mass fractions  and chemical constituents (such as SO4=) in the PM-respiratory hospital

        March 2001                               6-173        DRAFT-DO NOT QUOTE OR CITE

-------
 1      admissions association.  Several new studies report significant associations of increased
 2      respiratory-cause medical visits and/or hospital admissions with ambient PM2 5 and/or PMI0.2 5
 3      ranging to quite low concentrations. These include the Lippmann et al. (2000) study in Detroit,
 4      where all PM metrics (PM10, PM2 5, PM10_2 5, H+) were positively related to pneumonia and COPD
 5      admissions among the elderly (aged 65+ yr) in single pollutant models, with their RR values
 6      generally remaining little changed (but with broader confidence intervals) in multipollutant
 7      models including one or more gaseous pollutant (e.g., CO, O3, NO2, SO2). Excess risks for
 8      pneumonia admissions in the one pollutant model were 13% (3.7, 22) and 12% (0.8, 24) per
 9      25 yUg/m3 of PM2 5 and PM,0_2 5, respectively; those for COPD admissions were 5.5% (-4.7,  1 7)
1 0      and 9.3% (-4.4, 25) per 25 Mg/m3 PM2 5 and PM10.2 5, respectively.  Also of note, Moolgavkar
1 1      found ca. 5.0% excess risk for COPD  hospital admissions among the elderly (64+ yr) in Los
12      Angeles  to be significantly related to both PM2 5 and PM10.2 5 in one pollutant models; but the
13      magnitudes of the risk estimates dropped by more than half to non-statistically significant levels
14      in two-pollutant models including CO. In the same study, similar magnitudes of excess risk (i.e.,
15      in the range of ca. 4 to 7%) were found in one-pollutant models to be associated  with PM2 5  or
16      PM10_25 for other age groups  (0-19 yr; 20-64 yr) in Los Angeles, as well. Moolgavkar et al.
1 7      (2000) also found 5.6% (0.2, 11.3) excess risk for all-ages COPD hospital admissions per
18      25 /ug/m3 PM2 5 increase in King County,  WA.
19           In contrast to the above Lippmann and Moolgavkar findings, Tolbert reported no significant
20      associations of PM2 5 or PM10_2 5 with COPD emergency department visits in Atlanta, based on
21      data from less than half of all participating hospitals and ca. 1 yr of supersite air quality data.
22           Gwynn et al. (2000) considered a 2.5 yr period (May 1 988-Oct.  1 990) in the Buffalo, NY
23      region in a time-series analysis of daily mortality and hospital admissions for total, respiratory,
24      and circulatory hospital admissions categories. Pollutants considered included: PM10, H+, SO4=
25      COH, O3, CO, SO2, and NO2. The H+ and SO4= concentrations were determined  from daily  PM2 5
26      samples  not analyzed for mass (in order to avoid possible acid neutralization). Various modeling
27      techniques were applied to control for confounding of effect estimates due to seasonality,
28      weather  and day-of-week effects. They found multiple significant pollutant-health effect
29      associations, the most significant being between SO4= and respiratory hospital admissions.  When
30      calculated in terms of increments employed across analyses in this report, various PM RR's
31      were: PMIO RR=1.11, 95% C.I.=1.05-1.18(for 50 ^g/m3); H+ RR=1.06, 95% €.1=1.03-1.09 (for
        March 2001                              6-1 74        DRAFT-DO NOT QUOTE OR CITE

-------
  1      75 nmoles/m3= 3.6 //g/m3, if as H2SO4); and SO4= RR=1.08, 95% C.I.=1.04-1.12 (for
  2      155 nmoles/m3=15 /ug/m3).  As in the Burnett et al. (1997b) study described below, H+ yielded
  3      the highest RR per /ag/m3 of concentration.  These various PM metric associations were not
  4      significantly affected by inclusion of gaseous co-pollutants in the regression model. Thus, all
  5      PM components considered except COH were found to be associated with increased hospital
  6      admissions, but H+, SO4= and O3 had the most coherent associations with respiratory admissions.
  7           Lumley and Heagerty (1999) illustrate the effect of reliable variance estimation on data
  8      from hospital admissions for respiratory disease on King County, WA for eight years (1987-94),
  9      together with air pollution and weather information.  However, their weather controls were
 10      relatively crude (i.e., seasonal dummy variables and linear temperature terms).  This study is
 11      notable for having compared sub-micron PM (PM, 0) versus coarse PM10_i „ and for finding
 12      significant hospital admission associations only with PM, „. This may suggest that the PM2 5 vs.
 13      PM10 separation may not always  be sufficient to differentiate submicron fine particle vs. coarse-
 14      particle toxicities.
 15          Asthma hospital admission studies conducted in U.S. various communities provide
 16      additional important new data. Of particular note is a unique study by Sheppard et al. (1999)
 17      which evaluated relationships between measured ambient pollutants (PM10, PM2 5, PM10.2 5, SO2,
 18      O3 and CO) and non-elderly adult (<65 years of age) hospital admissions for asthma in Seattle,
 19      WA. Both PM and CO were found to be jointly associated with asthma admissions.  An
 20      estimated 4 to 5% increase in the rate of asthma hospital admissions (lagged 1 day) was reported
 21      to be associated with interquartile range changes in PM indices (19 ,ug/m3 for PM10, 11.8 //g/m3
 22      for PM2 5, and 9.3 /^g/rn3 for PM10.2 5), equivalent to excess risk rates as follows: 13% (95% CI
 23      05, 23) per 50 /^g/m3 for PM10; 9% (95% CI 3, 14) per 25 jug/m3 PM2 5; 11 % (95% CI 3, 20) per
 24      25 /ug/m3 PM10.25. Also of note is the Norris et al. (1999) study showing associations of low
 25      levels of PM2 5 (mean = 12 /wg/m3) with markedly increased asthma hospital admissions, i.e.,
 26      excess risk = 44.5% (CI 21.7, 71.4) per 25 /ug/m3 PM2 5.
 27          Turning to non-U.S. studies, Burnett et al. (1997b) evaluated the role that the  ambient air
 28      pollution mix, comprised of gaseous pollutants and PM indexed by various physical and
29      chemical measures, plays in exacerbating daily admissions to hospitals for cardiac diseases and
30      for respiratory diseases (tracheobronchitis, chronic obstructive long disease, asthma, and
31      pneumonia).  They employed daily measures of fine and coarse particulate mass, aerosol

        March 2001                               6-175        DRAFT-DO NOT QUOTE OR CITE

-------
 1      chemistry (sulfates and acidity), and gaseous pollutants (ozone, nitrogen dioxide, sulfur dioxide,
 2      and carbon monoxide) collected in Toronto, Ontario, Canada, during the summers of 1992, 1993,
 3      and 1994. Positive associations were observed for all ambient air pollutants for both respiratory
 4      and cardiac diseases.  Ozone was the most consistently significant pollutant and least sensitive to
 5      adjustment for other gaseous and particulate measures. The PM associations with the respiratory
 6      hospital admissions were significant for: PM10 (RR=1.11 for 50 /wg/m3; CI=1.05-1.17); PM2 5
 7      (fine) mass (RR=1.09 for 25 /ug/m3; CI=1.03-1.14); PM10.25 (coarse) mass (RR=1.13  for 25
 8      Mg/m3; CI=1.05-1.20); sulfate levels (RR=1.11 for 155 nmoles/m3=15 //g/m3; CI=1.06-1.17); and
 9      aerosol acidity (RR=1.40 for 75 nmoles/m3= 3.6 Aig/m3, if as H2SO4; CI=1.15-1.70).  After
10      simultaneous inclusion of ozone in the model, the associations with the respiratory hospital
11      admissions remained significant for: PM10 (RR=1.10; CI=1.04-1.16); fine mass (RR=1.06;
12      CI=1.01-1.12); coarse mass (RR=1.11; CI=1.04-1.19); sulfate levels (RR=1.06; €1=1.0-1.12);
13      and aerosol acidity (RR=1.25; CI=1.03-1.53), using the same increments. Of the PM metrics
14      considered here, aerosol acidity yielded the highest RR estimate, despite having the lowest mass
15      concentration increment, suggesting a higher toxicity per /u.g of exposure to acidic aerosols.
16      Regression models that included all recorded pollutants simultaneously (with high
17      intercorrelations among the pollutants) were also presented.
18           There have also been numerous new time-series  studies examining associations between air
19      pollution and respiratory-related hospital admissions in Europe, as summarized in Table 6-17;
20      but most of these studies relied primarily on black smoke (BS) as their PM metric. BS is a
21      particle reflectance measure that provides an indicator of particulate blackness and is highly
22      correlated with airborne carbonaceous particle concentrations (Bailey and Clayton, 1982). In the
23      U.S., Coefficient of Haze (COH) is a metric of particle transmittance that similarly most directly
24      represents a metric of particle blackness and ambient elemental carbon concentration (Wolff
25      et al., 1983) and has been found to be highly correlated with BS (r = 0.9) (Lee et al., 1972).
26      However, the relationship between airborne carbon and total mass of overall aerosol  (PM)
27      composition varies over time and from locality to locality, so the BS-mass ratio is less reliable
28      than the BS-carbon relationship (Bailey and Clayton,  1982).  This means that the BS-mass
29      relationship is likely to be very different between Europe and the U.S., largely due to differences
30      in local PM source characteristics (e.g., percentages of diesel powered motor vehicles).
31      Therefore, while these European BS-health effects studies are of qualitative use for evaluating

        March 2001                               6-176        DRAFT-DO NOT QUOTE OR CITE

-------
  1     the PM-health effects associations, they are not as useful for quantitative assessment of PM
  2     effects relevant to the U.S.
  3          The most recent European air pollution health effects analyses have mainly been conducted
  4     as part of the APHEA study, which evaluated 15 European cities from 10 different countries with
  5     a total population of over 25 million.  All studies used a standardized data collection and analysis
  6     approach which included: consideration of the same suite of air pollutants (BS, SO2, NO2, SO2,
  7     and O3) and the use of time-series regression addressing: seasonal and other long-term patterns;
  8     influenza epidemics; day of the week; holidays; weather; and, autocorrelation (Katsouyanni et al.,
  9     1996). The general coherence of the APHEA results with other results gained under different
 10     conditions strengthens the argument for causality in the air pollution-health effects association.
 11     Unfortunately, the general use of the less comparable suspended particle (SPM) measures and BS
 12     as PM indicators in some of the APHEA locations and analyses lessens the quantitative
 13     usefulness of such analyses in evaluating associations between PM and health effects most
 14     pertinent to the U.S. situation.
 15
 16     6.3.2.3.2 Identification of Potential Susceptible Subpopulations
 17          Associations between ambient PM measures and respiratory admissions have been found
 18     for all age groups, but older adults and children have been indicated by a number of hospital
 19     admissions  studies  to exhibit the most consistent PM-health effects associations in the literature.
 20     As reported in this and previous PM AQCDs, numerous studies of older adults (e.g., those 65+
 21      years of age) have related acute PM exposure with an increased incidence of hospital admissions
 22     (e.g., see Anderson et al, 1998). However, only a limited number have specifically studied
 23     children as a subgroup.  Burnett et al.  (1994) examined the differences in air pollution-hospital
 24     admissions  associations as a function of age in the province of Ontario, reporting that the largest
 25      percentage increase in admissions was found among infants (neonatal and post-neonatal, one year
 26      or less in age).
27           There are more than a dozen recent respiratory-related hospital admissions studies
 28      summarized in Table 6-17 that include children. Looking in detail at these study results reveals
29      that the PM RR's for all children (e.g., 0-18) are not usually noticeably larger than those for
30      adults, but such comparisons of RR's must adjust for differences in the baseline risks for each
31      group. For example, if hospital admissions  per 100,000 per day for young children are double

        March 2001                               6-177        DRAFT-DO NOT QUOTE OR CITE

-------
 1      the rate for adults, then they will have a pollution relative risk (RR) per //g/m3 that is half that of
 2      the adults given the exact same impact on admissions/100,0007^g/m3/day. Thus, it is important
 3      to adjust RR's or Excess Risks (ER's) for each different age groups' baseline, but this
 4      information is usually not available (especially regarding the population catchment for each age
 5      group in each study).
 6           One of the few indications that is notable when comparing children with other age group
 7      effect estimates in Table 6-17 is the higher excess risk estimate for infants (i.e., the group <1 yr.
 8      of age) in the Gouveia and Fletcher (2000) study, an age group that has estimated risk estimate
 9      roughly twice as large as for other children or adults.  This is confirmatory of the excess risk
10      pattern previously found in the above-noted Burnett et al. (1994) study for respiratory-related
11      hospital admissions.
12
13      6.3.2.3.3  Evaluation of Pollen as Potential Confounder
14           Pollen is an atmospheric constituent that has a large effect on asthma incidence and might
15      potentially be a factor that may confound PM-admissions associations, if it is correlated with
16      both PM and hospital admissions.  In a London study, airborne pollen did not confound the
17      analysis of air pollution (including black smoke) and daily admissions for asthma during the time
18      period 1987-1992 (Anderson et al., 1998).  However, in a study by Moolgavkar et al.  (2000) for
19      Seattle, adding pollens to PM time-series regressions of respiratory admissions was found to
20      diminish the PM effect estimates somewhat, but more for PM10 than that for PM2 5.  This latter
21      finding would not be unexpected, given that in general, confounding by bioaerosols is unlikely to
22      account for PM2 5-related health impacts due to lack of correlation between daily PM2 5 levels and
23      seasonal pollution events and weather-related specification events. Overall, then, pollens do not
24      appear to significantly confound the PM-admissions relationship, despite their large effect on
25      respiratory admissions.
26
27      6.3.2.4 Key New Respiratory Medical Visits Studies
28           As discussed above, medical visits  include both hospital emergency department (ED) visits
29      and doctors' office visits. As in the past PM AQCD's, most of the available morbidity studies
30      presented in Table 6-17 are of ED visits and their associations with air pollution. These studies
31      collectively confirm the results provided in the previous AQCD,  indicating a positive and

        March 2001                               6-178        DRAFT-DO NOT QUOTE OR CITE

-------
  1      significant association between ambient PM levels and increased respiratory-related hospital
  2      visits.
  3          Of the medical visit and hospital admissions studies since the 1996 PM AQCD, the most
  4      informative are those that evaluate health effects associations at levels below previously well-
  5      implicated PM concentrations. In the case of medical visits, the Norris et al. (1999, 2000) studies
  6      of asthma ED visits found significant PM- associated health effects among children in Seattle,
  7      even at quite  low average PM levels and  even after incorporating the effects of other air
  8      pollutants (study mean PM10 = 21.7 /ug/m3; estimated mean PM2 5 = 12 /ug/m3). Tolbert et al.
  9      (2000b) reported a significant PM,0 association with pediatric ED visits  in Atlanta where the
 10      maximum PMIO concentration was 105 yUg/m3. Also, Delfino et al. (1997) found significant PM10
 11       and PM2 5 associations for respiratory ED visits among older adults in Montreal when mean
 12      PM10= 21.7 Aig/m3 and mean PM2 5 = 12.2 /ug/m3. Medina et al. (1997) reported significant
 13       associations between doctor's asthma house visits and PM13 (which would have a slightly higher
 14      concentration value than PM10) in Paris when mean PM13 = 25 /ug/m3 and maximum daily
 15       PM,3 = 95 //g/m3, Hajat et al. (1999) reported significant PM10 associations with asthma doctor's
 16      visits for children and young adults in London when mean PM10 = 28.2 yUg/m3 and the PM10 90th
 17       percentile was only 46.4 /ug/m3. Overall, then, numerous new medical visits studies indicate
 18       PM-health effects associations at lower PM2 5 and PM10 levels than demonstrated previously for
 19       this health outcome.
 20
 21       6.3.2.4.1 Scope of Medical Visit Morbidity Effects
 22           Several of these recent medical visit studies consider a new endpoint for comparison with
 23      ED visits: visits in the primary care setting.  In particular, key studies showing PM-health effects
 24      associations for this health outcome include:  the study by Medina et al. (1997) for Paris, France
 25      which evaluated doctors' visits to patients in that city; the study by Hajat et al. (1999) that
 26      evaluated the  relationship between daily General  Practice (GP) doctor consultations for asthma
 27      and other lower respiratory disease (LRD) and air pollution in London, UK; the study by
28      Choudhury et al. (1997) of private asthma medical visits in Anchorage, Alaska; and the study by
29      Ostro et al. (1999b) of daily visits by young children to primary care health clinics in Santiago,
30      Chile for upper or lower respiratory symptoms.


        March 2001                               6-179        DRAFT-DO NOT QUOTE OR CITE

-------
 1          While limited in number, the above studies collectively provide new insight into the fact
 2      that there is a broader scope of severe morbidity associated with PM air pollution exposure than
 3      previously documented. As the authors of the London study note: "There is less information
 4      about the effects of air pollution in general practice consultations but, if they do exist, the public
 5      health impact could be considerable because of their large numbers." Indeed, the Paris doctors'
 6      house calls and the London doctors' GP office visits studies both indicate that the effects of air
 7      pollution, including PM, can affect many more people than indicated by hospital admissions
 8      alone.
 9          These new studies  also provide indications as to the quantitative nature of medical visits
10      effects, relative to those for hospital admissions. In the London case, comparing the number  of
11      admissions from the authors' earlier study (Anderson et al., 1996) with those for GP visits in the
12      1999 study (Hajat et al., 1999) indicates that there are  approximately 24 asthma GP visits for
13      every asthma admission in that city.  Also, comparing the PMIO coefficients indicates that the
14      all-ages asthma effect size for the GP visits (although not statistically different) was about 30%
15      larger than that for hospital admissions. Similarly, the number of doctors' house calls for asthma
16      approximated 45/day in  Paris (based on an average 9 asthma house calls in the SOS-Medocina
17      data base, representing 20% of the total; Medina et al., [1997]), versus an average 14 asthma
18      admissions/day (Dab et al., 1996), or a factor of 3 more doctors' house calls than hospital
19      admissions.  Moreover, the RR for Paris asthma doctors' house calls was much higher than
20      asthma admissions (RR=1.18 for 25 /ug/m3 BS for house calls vs. RR=1.01 per 25 /ug/m3 BS  for
21      hospital admissions).  Thus, these new studies suggest that looking at only hospital admissions
22      and emergency hospital  visit effects  may greatly underestimate numbers of respiratory morbidity
23      events in a population due to acute ambient PM exposure.
24
25      6.3.2.4.2  Evaluation of Extraneous Factors Potentially Affecting Respiratory Medical Visit
26               Study Outcomes
27          Some recent studies have examined certain factors that might extraneously affect the
28      outcomes of PM-medical visit  studies. Stieb et al. (1998a) examined the occurrence of bias and
29      random variability in diagnostic classification of air pollution  and daily cardiac respiratory
30      emergency department visits such as asthma, COPD, respiratory infection and cardiac. They
31      concluded that there was no evidence of diagnostic bias in relation to daily air pollution levels.

        March 2001                               6-180       DRAFT-DO NOT QUOTE OR CITE

-------
  1     Also, Stieb et al. (1998b) reported that for a population of adults visiting an emergency
  2     department with cardiac respiratory disease, fixed site sulfate monitors appear to accurately
  3     reflect daily variability in average personal exposure to particulate sulfate, whereas particulate
  4     acid exposure was not as well represented by fixed site monitors. Another study investigated
  5     possible confounding of respiratory visit effects due to pollens. In London, Atkinson et al.
  6     (1999b) studied the association between the number of daily visits to emergency departments for
  7     respiratory complaints and measures of outdoor air pollution for PM10, NO2, SO2 and CO. They
  8     examined different age groups and reported the strongest association for children for visits for
  9     asthma, but were unable to separate the effects of PM10 and SO2.  Pollen levels did not influence
 10     the results, similar to results from the asthma panel studies described below in Section 6.3.3.
 11
 12     6.3.2.5  Summary of Key Findings on Acute Particulate Matter Exposure and
 13             Respiratory-Related Hospital Admissions  and Medical Visits
 14          The results of new studies discussed above are generally consistent with and supportive of
 15     findings presented in the previous PM AQCD (U.S. Environmental Protection Agency, 1996),
 16     with regard to ambient PM associations of short-term exposures with respiratory-related hospital
 17     admissions/medical visits.  Excess risk estimates for specific subcategories of respiratory-related
 18     hospital admissions/medical visits  for U.S.  cities are graphically depicted in Figure 6-7. The
 19     excess risk estimates fall most consistently in the range of 5 to 25% per 50 ywg/m3 PM,0
 20     increments, with those for asthma visits and hospital admissions tending to be somewhat higher
 21     than for COPD and pneumonia hospital admissions.
 22          More limited new evidence substantiates increased risk of respiratory-related hospital
 23     admissions due to ambient fine particles (PM2 5, PM, „, etc.), and it also points towards
 24     associations of such admissions with ambient coarse particles (PMi0_2 5). Excess risk estimates
 25     tend to fall in the range of ca. 5.0 to 15.0% per 25 /ug/m3 PM25 or PM10.2 5 for overall respiratory
 26     admissions or for COPD admissions, whereas larger estimates are found for asthma admissions
 27     (ranging upwards to ca. 40 to 50%  for children < 18 yr.  old in one study).
28           Various new medical visits studies (including non-hospital physician visits) indicate that the
29      use of hospital admissions alone can greatly understate the total clinical morbidity effects of air
30      pollution.  Thus, these results support the hypothesis that considering only hospital admissions
31      and emergency hospital visit effects may greatly underestimate the numbers of medical visits

        March 2001                               6-181        DRAFT-DO NOT QUOTE OR CITE

-------
              Tolbertetal (2000a) Atlanta -
               Morris et al (2000) Seattle -
              Noms et al. (2000) Spokane -
               Morris etal (1999) Seattle -
          Choudhury et al (1997) Anchorage -
         Nauenberg and Basu (1999) LA.CA -
             Sheppard et a) (1999) Seattle -
           Zonobetti et al (2000b) Chicago -
           Samet et al (2000b) 14 US Cities -
              Moolgavkar (2000c) Phoenix -
               Moolgavkar (2000c) LA.CA -
              Moolgavkar (2000c) Chicago -
            Moolgavkar et al. (2000) King C -
           Moolgavkar et al (1997) Minn-SP -
             Moolgavkar et al  (1997) Birm -
              Chen et al. (2000) Reno.NV -
           Zanobetti et al (2QOOb) Chicago -
           Samet et al. (2000b)  14 US Cities -
                                  -25







._:

f


Asthma Visits





i Asthma Hospital Admissions
^ t
f*H
^ COPD Hospital Admissions
t-»H
•H

w Pneumonia Hospital Admissions
                                           25       50       75
                                                 Excess Risk, %
             100
125
150
        Figure 6-7. Maximum excess risk of respiratory-related hospital admissions and visits per
                    50-jLtg/m3 PM10 increment in selected studies of U.S. cities.
 1
 2
 3
 4
 5
 6
 9
10
11
occurring in a population as a result of acute ambient PM exposure. Those groups identified in
these morbidity studies as most strongly affected by PM air pollution are older adults and the
very young.

6.3.3  Effects of Particulate Matter Exposure  on Lung Function and
       Respiratory Symptoms
     In the 1996 PM AQCD, the available respiratory disease studies used a wide variety of
designs examining pulmonary function and respiratory symptoms in relation to PM10.  The
models for analysis varied and the populations included several different subgroups. Pulmonary
function studies were suggestive of short term effects resulting from ambient PM exposure.  Peak
expiratory flow rates showed decreases in the range of 2 to 5  1/min resulting from an increase of
        March 2001
                                           6-182
DRAFT-DO NOT QUOTE OR CITE

-------
  1     50 /ug/m3 in 24-h PM10 or its equivalent, with somewhat larger effects in symptomatic groups
  2     such as asthmatics. Studies using FEV, or FVC as endpoints showed less consistent effects. The
  3     chronic pulmonary function studies were less numerous than the acute studies, and the results
  4     were inconclusive.
  5
  6     6.3.3.1  Effects of Short-Term Particulate Matter Exposure on Lung Function and
  7             Respiratory Symptoms
  8          The available acute respiratory symptom studies discussed in the 1996 PM AQCD included
  9     several different endpoints, but typically presented results for: (1) upper respiratory symptoms,
 10     (2) lower respiratory symptoms, or (3) cough. These respiratory symptom endpoints had similar
 11     general patterns of results.  The odds ratios were generally positive, the 95% confidence intervals
 12     for about half of the studies being statistically significant (i.e., the lower bound exceeded 1.0).
 13          The earlier studies of morbidity health outcomes of PM10 exposure on asthmatics were
 14     limited in terms of conclusions that could be drawn because of the few available studies on
 15     asthmatic subjects. Lebowitz et al. (1987) reported a relationship with TSP exposure and
 16     productive cough in a panel of 22 asthmatics but not for peak flow or wheeze. Pope et al. (1991)
 17     studied respiratory symptoms in two panels of asthmatics in the Utah Valley. The 34 asthmatic
 18     school children panel yielded estimated odd ratios of 1.28 (1.06, 1.56) for lower respiratory
 19     illness (LRI) and the second panel of 21 subjects aged 8 to 72 for LRI of 1.01 (0.81, 1.27) for
 20     exposure to PM10.  Ostro et al. (1991) reported no association for PM2 5 exposure in a panel of
 21     207 adult asthmatics in Denver; but, for a panel of 83 asthmatic children age 7 to 12 in central
 22     Los Angeles, reported a relationship of shortness of breath to O3 and PM10, but could not  separate
 23     effects of the two pollutants. These few studies did not indicate a consistent relationship  for
 24     PM10 exposure and health outcome in asthmatics.
 25         Numerous new studies of short-term PM exposure effects on lung function and respiratory
 26     symptoms have been published since early  1996. Most of these followed a panel of subjects over
 27     one or more periods and evaluated daily lung function and/or respiratory symptom associations
28     with changes in ambient PM,0 and/or PM2 5. Lung function was usually measured daily with
29     many studies including forced expiratory volume (FEV), forced vital capacity (FVC)  and  peak
30     expiratory flow rate (PEF).  Most analyses included both morning and afternoon measurements.
31      A variety of respiratory symptoms were measured, including cough, phlegm, difficulty breathing,

        March 2001                              6-183        DRAFT-DO NOT QUOTE OR CITE

-------
 1      wheeze, and bronchodilator use. Finally, several measures of airborne particles were used,
 2      including: PM10, PM25, TSP, BS, and sulfate fraction of ambient PM.
 3          These various studies are summarized in several tables presented below. Data on physical
 4      and chemical aspects of ambient PM levels (especially for PM10 and PM2 5 and smaller size
 5      fractions) are of particular interest, as are new studies examining health outcome effects and/or
 6      exposure measures not studied as much in the past. Each table is organized by study location,
 7      PM measure, etc.  Where possible, results are presented in terms of the units described earlier.
 8      Specific analyses were selected for summarization based on the following criteria:
 9      • Peak flow was used as the primary lung function measurement of interest. While FEV1 would
10       be a good measure, peak flow (PF) is most often measured in these panel studies.
11      • Cough, phlegm, difficulty breathing, wheeze, and broncho-dilator use were summarized as
12       measures of respiratory symptoms when available.
13      • Preference was given to results reported for PM10 and PM2 5 and/or smaller PM.
14      • The  analyses were also restricted to include a short-term lag (zero or one day), a longer-term
15       lag (2- to 5- day), and a moving average analysis, when available. If both  0- and a 1-day lag
16       analyses were presented, the 0-day lag analysis was used for all, but the AM PF results.
17       For longer lags, the measure coming closest to average of 2 to 5 days was  selected.
18
19      6.3.3.1.1 Lung Function and Respiratory Symptom Effects in Asthmatic  Subjects
20          Tables 6-19 and 6-20 summarize short-term PM exposure effects on lung function and
21      respiratory symptoms, respectively, in asthmatic subjects.  The peak flow analyses results for
22      asthmatics tend to  show small decrements for both PM10 and PM2 5. For PM10, 2 of 4 of the
23      newly available point estimates showed decreases, but the majority of the studies  were not
24      statistically significant, as shown in Figure 6-8 as an example of PEF outcomes.  Lag 1 may be
25      more relevant for morning measurement of asthma outcome from the previous day. The figure
26      presents studies which provided such data.  The results were consistent for both AM and PM
27      peak flow analyses. The effects using 2 to five-day lags averaged about the  same as did the zero
28      to one-day lags, but the effects had wider confidence limits. Similar results  were  found for the
29      PM2 5 studies, although there were fewer studies.  Several studies included PM2 5 and PMIO
30      independently in their analyses of peak flow.  Of these, Gold et al. (1999), Naeher et al. (1999),
31      Tiittanen et al. (1999), Pekkanen et al. (1997), and Romieu et al. (1996) all found similar results

        March 2001                               6-184        DRAFT-DO NOT QUOTE OR CITE

-------
3
S3*
K)
O
o
  TABLE 6-19. SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION TESTS
 	  IN STUDIES OF ASTHMATICS	

                                                                                                                                  Effect measures standardized to 50 /ag/m3
                                                                                                                                  PM10 (25 Aig/m3 PM2.S) Negative coefficients
                                                                                                                                  for lung function and ORs greater than 1 for
                                                                                                                                  other endpomts suggest PM effects
          Reference citation, location, duration,
          pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc
Results and Comments
Effects of co-pollutants
oo
Tl
H
6
o
2
o
H
O
s
£1
o
*3
O
          United States

          Schwartz and Neas (2000)
          Reanalyses of three recent longitudinal diary studies
          Harvard Six Cities, Uniontown, PA, State College, PA
          Thurston et al. (1997)
          Summers of 1991-93.
          Ozone, H+, and sulfates fraction.
Canada

Vedaletal. (1998)
Port Alberni, BC
Europe

Gielenetal. (1997)
Amsterdam, NL
Mean PM,() level.  30.5 ^g/m3 (16, 60 3).
Mean maximum 8 hr O3  67 ,ug/m3
                                                 Analysis of Harvard Six City Study data for
                                                 1,844 school children (grades 2-5) from
                                                 6 eastern U S urban areas only included
                                                 respiratory symptom diary information.
                                                 Uniontown and State College analyses also
                                                 included PEF data.

                                                 Three 5-day summer camps conducted in 1991,
                                                 1992, 1993. Symptoms and change in lung
                                                 function (morning to evening) measured.
                                                 Linear regression analysis adjusting for pollen
                                                 and daily maximum temperature was used for
                                                 analysis of lung function.
Study 206 children aged 6 to 13 years living in
Port Albemi, British Columbia. 75 children had
physician-diagnosed asthma, 57 had an
exercised induced fall in FEV1, 18 children
with airway obstruction, and 56 children
without any symptoms. Peak flow measured
twice daily  An autoregressive model was fitted
to the data using GEE methods. Covariates
included temperature, humidity, and
precipitation.
Study evaluated 61 children aged 7 to 13 years
living in Amsterdam, The Netherlands.  77
percent of the children were taking asthma
medication and the others were being
hospitalized for respiratory problems. Peak
flow measurements were taken twice daily.
Associations of air pollution were evaluated
using time senes analyses. The analyses
adjusted for pollen counts, time trend, and day
of week.
                                           In both the Harvard Six City and the
                                           Uniontown-State College diary studies,
                                           fine particle measures were more strongly
                                           related to asthma-related responses (i.e.,
                                           increased lower respiratory symptoms and
                                           decreased peak flow).

                                           No relationship between lung function
                                           and pollutants was found.
                                                                                                       PMIO was associated with changes in
                                                                                                       morning peak flow for lags 0, 1, 2, 3, and
                                                                                                       4 day average.
The strongest relationships were found
with ozone, although some significant
relationships found with PMm
LagO,PM10:
 Evening PEF = -0 08 (-2.49, 2.42)
Lagl,PM10-
 Morning PEF = 1.38 (-0 58, 3.35)
Lag 2, PM,0:
 Morning PEF = 0.34 (-1.78, 2.46)
 Evening PEF = -1 46 (-3.23, 0.32)

-------
p
3
N>
O
o
  TABLE 6-19 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
 =========	TESTS IN STUDIES OF ASTHMATICS                           	

                                                                                                                                  Effect measures standardized to 50 ^g/m3
                                                                                                                                  PM,o (25 Mg/mJ PM2.5). Negative coefficients
                                                                                                                                  for lung function and ORs greater than 1 for
                                                                                                                                  other endpoints suggest PM effects
          Reference citation, location, duration,
          pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
ON
oo
ON
O
H
O
o
H
W
O
&
o
I—<
H
W
          Europe (cont'd)

          Hiltermannetal. (1998)
          Leiden, ML
          July-Oct, 1995
           Peters etal. (1996)
           Erfurt and Weimar, Germany
           SO2, TSP, PM ,„, sulfate fraction, and PSA
           Mean PM|0 level was 112 /j.g/m\
           Peters etal. (1997b)
           Erfurt, Germany
           PM fractions measured over range of sizes from
           ultrafme to fine, including PM,,,
           Particles measured using size cuts of 0.01 to 0 1, 0 1 to
           0.5, and 0 5 to 2.5 Mm.
           Mean PMIO level- 55 A(g/m3 (max 7)).  Mean SO2
           100Mg/m3(max383).
Peters etal. (1997c)
Sckolov, Czech Republic
Winter 1991-1992
PM,0, SO2, TSP, sulfate, and particle strong acid
Median PM,0 level: 47 ^g/m3 (29, 73)
Median SO2:  46 /jg/m3 (22, 88).
                                                 270 adult asthmatic patients from an out-patient
                                                 clinic in Leiden, The Netherlands were studied
                                                 from July 3 to October 6, 1995. Peak flow
                                                 measured twice daily. An autoregressive model
                                                 was fitted to the data  Covariates included
                                                 temp, and day of week  Individual responses
                                                 not modeled

                                                 Panel of 155 asthmatic children in the cities of
                                                 Erfurt and Weimar, E. Germany studied. Each
                                                 panelist's mean PEF over the entire period
                                                 subtracted from the PEF value to obtain a
                                                 deviation Mean deviation for all panelists on
                                                 given day was analyzed using an autoregressive
                                                 moving average.  Regression analyses done
                                                 separately for adults and children in each city
                                                 and winter, then combined results calculated

                                                 Study of 27 non-smoking adult asthmatics
                                                 living m Erfurt, Germany dunng winter season
                                                 of 1991-1992. Morning and evening peak flow
                                                 readings recorded. An auto-regressive model
                                                 was used to analyze deviations in individual
                                                 peak flow values, including terms for time
                                                 trend, temp , humidity, and wind speed and
                                                 direction,
89 children with asthma in Sokolov, Czech
Republic studied.  Subjects kept diaries and
measured peak flow for seven months during
winter of 1991 -2.  The analysis used linear
regression for PFT. First order autocorrelations
were observed and corrected for using
polynomial distributed lag (PDL) structures.
                                           No relationship between ozone and PFT
                                           was found
                                           Five day average SO2 was associated with
                                           decreased PEF. Changes in PEF were not
                                           associated with PM levels.
                                           Strongest effects on peak flow found with
                                           ultrafme particles. The two smallest
                                           fractions, 0.01 to 0 1 and 0.1 to 0.5 were
                                           associated with a decrease of PEF.
Five day mean SO2, sulfates, and particle
strong acidity were also associated with
decreases in PM PFT as well as PM,0.
                                      Lag 0, PM10:
                                       Average PEF = -0.80 (-3.84, 2.04)
                                      7 day ave., PMIO:
                                       Average PEF = -1.10 (-5.22, 3.02)
LagO, PM,,,:
 Evening PEF =-0.38 (-1.83, 1 08)
Lagl,PM,0:
 Morning PEF = - 1 30 (- 2 36, 0.24)
5 Day Mean, PMH)
 Morning PEF = -1 51 (-3.20,0 19)
 Evening PEF = -2.31 (-4.54, -008)
LagO, PM,5:
 Evening PEF = -0 75 (-1.66, 0 17)
Lag 1,PM25-
 Morning PEF =-071 (-1.30,0.12)
5 Day Mean, PM, 5.
  Morning PEF=-1 19 (-1.81, 0 57)
  Evening PEF = -1.79 (-2.64, -0.95)

Lag 0, PM,,,:
 Morning PEF =-0.71 (-2 14,070)
 Evening PEF = -0.92 (- 1.96, 0.12)
5 Day mean PMIO-
 Evening PEF = -1.72 (-3.64, 0.19)
 Morning PEF = -0.94 (-2.76, 0 91

-------
O
O
             TABLE 6-19 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
                                                                   TESTS IN STUDIES OF ASTHMATICS
          Reference citation, location, duration,
          pollutants measured, summary of values
                                                 Type of study, sample size, health outcomes
                                                 measured, analysis design, covanates included,
                                                 analysis problems, etc
                                          Results and Comments
                                          Effects of co-pollutants
                                     Effect measures standardized to
                                     PM,0 (25 ^g/m3 PM2.,). Negative coefficients
                                     for lung function and ORs greater than 1 for
                                     other endpomts suggest PM effects
OO
 •n
 H
 6
 o
 2
 o
 H
O
 o
 3
 o
 w
 o
 H
          Europe (cont'd)

          Timonen and Pekkanen (1997)
          Kupio, Finland
          PM,0, BS, NO2, and S02.
          Pekkanen et al. (1997)
          Kuopio, Finland
          PM fractions measured over range of sizes from
          ultrafine to fine, including PM,0
          Mean PM,,, level  18/^g/m3 (10, 23)
          Mean NO2 level
Segalaetal. (1998)
Paris, France
Nov. 1992-May 1993.
BS, SO,, NO2, PM,3 (instead of PM,,,), measured.
Mean PM13 level:  34.2 ,ug/m' (range 8.8, 95).
Mean SO2 level-  21.7 ,ug/m3 (range 4 4, 83.8).
Mean NO, level: 56.9 ^g/m3 (range 23 8, 121 9)
Agocsetal. (1997)
Budapest, Hungary
Australia

Rutherford etal. (1999)
Brisbane, Australia
PM 10, TSP, and particle diameter.
                                                 Studied 74 asthmatic children (7 to 12 yr) in
                                                 Kuoio, Finland. Daily mean PEF deviation
                                                 calculated for each child.  Values were
                                                 analyzed, then using linear first-order
                                                 autoregressive model.

                                                 Studied 39 asthmatic children aged 7-12 years
                                                 living in Kuopio, Finland Changes in peak flow
                                                 measurements were analyzed using a linear
                                                 first-order autoregressive model.
                                                            Study of 43 mildly asthmatic children aged 7-15
                                                            years living in Pans, France from Nov. 15, 1992
                                                            to May 9, 1993. Peak flow measured three
                                                            times a day.  Covariates in the model included
                                                            temperature and humidity An autoregressive
                                                            model was fitted to the data using GEE
                                                            methods.
Panel of 60 asthmatic children studied for two
months in Budapest, Hungary Mixed model
used relating TSP to morning and evening
PEFR measurements, adjusting for SO2, time
trend, day of week, temp., humidity
Study examined effects of 11  dust events on
peak flow and symptoms of people with asthma
in Brisbane, Australia.  PEF data for each
individual averaged for a period of 7 days prior
to the identified event  This mean was
compared to the average for several days of PEF
after the event, and the difference was tested
using a paired t-test
                                          Lagged concentrations of NO, related to
                                          declines m morning PEF as well as PMIO
                                          andBS
                                          Changes in peak flow found to be related
                                          to all measures of PM, after adjusting for
                                          minimum temperature. PNO.032-0.10
                                          (I/cm3) and PN1.0-3.2 (I/cm3) were most
                                          strongly associated with morning PEF
                                          deviations
                                           Effects found related to PM,,, were less
                                           than those found related to the other
                                           pollutants
The paired t-tests were stat. significant for
some days, but not others. No general
conclusions could be drawn.
                                      Lagged concentrations of PM,0 and BS related
                                      to declines in morning PEF.
                                      LagO, PM,0:
                                       Evening PEF = -0 35 (- 1.14, 0.96)
                                      Lag 1,PMIO:
                                       Morning PEF = -2.70 (-6.65, 1.23)
                                      Lag 2, PM,,,:
                                       Morning PEF = -4 35 (-8 02, -0.67)
                                       Evening PEF = - 1.10 (-4 70, 2.50)

                                      Small sized particles had relationships similar
                                      to those of PM,,, for morning and evening
                                      PEF.

                                      Lag 4, PM,,-
                                       Morning PEF = -0.62 (- 1.52, 0.28)
                                      No significant TSP-PEFR relationships found

-------
  .
 O
 o
  TABLE 6-19 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
 	TESTS IN STUDIES OF ASTHMATICS	

                                                                                                                            Effect measures standardized to 50 ^g/m3
                                                                                                                            PM|0 (25 jug/m3 PM2.5)  Negative coefficients
                                                                                                                            for lung function and ORs greater than 1 for
                                                                                                                            other endpomts suggest PM effects
          Reference citation, location, duration,
          pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covanates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
oo
oo
          Latin America

          Romieuetal. (19%)
          Mexico City, Mexico
          During study period, maximum daily 1 -h O3 ranged
          from 40 to 370 ppb (mean 190 ppb, SD = 80 ppb).
          24 h ave, PM,0 levels ranged from 29 to 363 ,ug/m3
          (mean 166.8 Mg/m3, SD 72.8 Mg/m3).
          For 53 percent of study days, PM|0 levels exceeded
Romieuetal. (1997)
Mexico City, Mexico
During study period, maximum daily 1-h ozone ranged
from 40 to 390 ppb (mean 196 ppb SD = 78 ppb)
PM|() daily average ranged from 12 to  1
                                               Study of 71 children with mild asthma aged 5-7
                                               years living in the northern area of Mexico City.
                                               Morning and evening peak flow measurements
                                               recorded by parents. Peak flow measurements
                                               were standardized for each person and a model
                                               was fitted using GEE methods. Model included
                                               terms for minimum temperature
Study of 65 children with mild asthma aged 5-
13 yr in southwest Mexico City. Morning and
evening peak flow measurements made by
parents  Peak flow measurements standardized
for each person and model was fitted using GEE
methods. Model included terms for minimum
temperature
                                         Ozone strongly related to changes in
                                         morning PEF as well as PM10.
Strongest relationships were found
between ozone (lag 0 or 1) and both
morning and evening PFT.
LagO, PMI0.
 Evening PEF =
Lag 2, PM,0:
 Evening PEF =
LagO,PM25:
 Evening PEF =
Lag 2, PM25:
 Evening PEF =
Ugl,PM10
 Morning PEF =
Lag 2, PMU)
 Morning PEF =

LagO, PM10:
 Evening PEF =
Lag 2, PM10.
 Evening PEF =
 Morning PEF =
Lag 0, PM,0
 Morning PEF =
 -4.80 (-8.00,-1.70)

 -3.65 (-7.20, 0.03)

 -4.27 (-7.12,-085)

 -2.55 (-7 84, 2.74)

•• -4.70 (-7.65, -1 7)

> -4.90 (-8 4, -1.5)


 -1.32 (-6.82, 4.17)

 -004 (-4.29, 4.21)
<2.47(-l 75,6.75)

•• 0.65 (-3 97, 5.32)
 O
 O
 2
 O
 H
/O
 s
 o
 H
 w

-------
I
K)
O
o
 TABLE 6-20.  SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON SYMPTOMS IN STUDIES
===^==	==	       OF ASTHMATICS

                                                                                                                                Effect measures standardized to 50 ^g/m'
                                                                                                                                PM10 (25 ^g/m3 PMW). Negative coefficients
                                                                                                                                for lung function and ORs greater than 1 for
                                                                                                                                other endpomts suggest PM effects
          Reference citation, location, duration,
          pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates
included, analysis problems, etc.
Results and Comments
Effects of co-pollutants
00



0

Tl

O
z;
o
c
o
H
W
O
X>
o
h—(
H
W
          United States

          Delfmoetal. (1996)
          So. California
          Summer of 1995.
          Ozone and PM10 measured
          Delfmoetal. (1997)
          San Diego County, CA
           Delfmoetal (1998)
           So. California community
           Aug.-Oct. 1995
           Highest 24-hour PM10 mean:  54/^g/m3
           Ostroetal. (1995)
           Los Angeles, CA
           TSP, sulfates, nitrates, 0,, SO2, N02, and PM,0
           measured.
                                               Study of 9 adults and 13 children with history
                                               of bronchodilator use living in Southern
                                               California  An autoregressive logistic model
                                               was fitted to symptoms adjusting for
                                               temperature and relative humidity

                                               A panel of 9 adults and 13 children were
                                               followed during late spring 1994 in semi-rural
                                               area of San Diego County at the inversion
                                               zone elevation of around 1,200 feet. A
                                               random effects model was fitted to ordinal
                                               symptom scores, bronchodilator use, and PEF
                                               in relation to 24-hour PM10. Temp., relative
                                               humidity, fugal spores, day of week and O3
                                               evaluated

                                               Relationship of asthma symptoms to O, and
                                               PMIO examined in a So. California
                                               community with high Oj and low PM  Panel
                                               of 25 asthmatics ages 9-17 followed daily,
                                               Aug -Oct., 1995  Longitudinal regression
                                               analyses utilized GEE model controlling for
                                               autocorrelation, day of week, outdoor fungi
                                               and weather.

                                               Study  of African-American children, ages
                                               7-12 years  with confirmed asthma.
                                               109 children were eligible, most of whom
                                               lived central and south-central Los Angeles.
                                               Analyses done using "daily reporting of
                                               respiratory symptoms including cough,
                                               shortness of breath, and wheeze". General
                                               logistic regression models used, with GEE
                                               corrections for autocorrelation.
                                          Pollen not associated with asthma symptom
                                          scores.
                                        No significant relationships with PM]0.
                                          Although PM10 never exceeded 51
                                          bronchodilator use was significantly
                                          associated with PM,0(0.76 [0.027, 0.27])
                                          puffs per 50 Mg/m3. Fungal spores were
                                          associated with all respiratory outcomes.
                                          Asthma symptoms scores significantly
                                          associated with both outdoor O3 and PMIO
                                          in single pollutant and co-regressions  1 -hr
                                          and 8-hr maxi PM10 had larger effects than
                                          24-hr mean.
                                          relationships found between shortness of
                                          breath and PM10 or ozone, with symptoms
                                          estimated to increase about 9% per each 10
                                               3 increase in PM,0.
                                         Lag 0, Symptoms:
                                          Short. Breath OR = 1 51 (1 04, 2.17)
                                         Others not significant

-------
o
O
o
   TABLE 6-20 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON SYMPTOMS IN STUDIES
  	OF ASTHMATICS	

                                                                                                                                   Effect measures standardized to 50 ^g/m3
                                                                                                                                   PM,0(25 ^g/m' PM2.5).  Negative coefficients
                                                                                                                                   for lung function and ORs greater than 1  for
                                                                                                                                   other endpomts suggest PM effects
          Reference citation, location, duration,
          pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covanates
included, analysis problems, etc
Results and Comments
Effects of co-pollutants
O
O
2:
3
/O
c
o
H
m
O
&
O
H—<
H
W
           United States (cont'd)

           Thurston et al. (1997)
           Summers 1991-1993.
           O3, H+, sulfate, pollen, daily max temp, measured.
           Canada

           Vedaletal (1998)
           Port Albemi, BC
          Europe

          Gielenetal. (1997)
          Amsterdam, NL
Hiltermann et al. (1998)
Leiden, NL
July-Oct 1995.
                                                  Three 5-day summer camps conducted in
                                                  1991, 1992, 1993. Study measured
                                                  symptoms and change in lung function
                                                  (morning to evening). Poisson regression for
                                                  symptoms.
                                                  Study of 206 children aged 6 to 13 years
                                                  living in Port Alberni, British Columbia
                                                  75 children had physician-diagnosed asthma,
                                                  57 had an exercised induced fall in FEV1,
                                                  18 children with airway obstruction, and
                                                  56 children without any symptoms.
                                                  Respiratory symptom data obtained from
                                                  diaries An autoregressive model was fitted to
                                                  the data, using GEE methods. Covanates
                                                  included temp , humidity, and precipitation.
Study of 61 children aged 7 to 13 years living
in Amsterdam, NL. 77 percent were taking
asthma medication and the others were being
hospitalized for respiratory problems
Respiratory symptoms recorded by parents in
diary. Associations of air pollution evaluated
using time series analyses, adjusted for pollen
counts, time trend, and day of week.

Study of 270 adult asthmatic patients from an
out-patient clinic in Leiden, NL from July 3,
to October 6,  1995. Respiratory symptom
data obtained from diaries. An autoregressive
model was fitted to the data. Covanates
included temperature and day of week.
                                         Ozone related to respiratory symptoms
                                         No relationship between symptoms and
                                         other pollutants.
                                         In general, PM,0 was associated with
                                         changes in both peak flow and respiratory
                                         symptoms.
                                                                                           Strongest relationships found with O3,
                                                                                           although some significant relationships
                                                                                           found with PM10.
PMlo> O3, and NO2 were associated with
changes in respiratory symptoms.
                                        Lag 0, Symptoms:
                                         Cough OR = 1.40 (1.04, 1.88)
                                         Phlegm OR = 1 28 (0.86, 1 89)
                                        Lag 2, Symptoms:
                                         Cough OR = 1 40(1.13, 1.73)
                                         Phlegm OR =1.40 (1.03, 190)
                                        Lag 0, Symptoms
                                         Cough OR = 2.19 (0.77, 6.20)
                                         Branch. Dial. OR = 0.94 (0 59, 1.50)
                                        Lag 2, Symptoms:
                                         Cough OR = 2.19 (047, 10.24)
                                         Bronch. Dial. OR = 2 90 (1.80, 4 66)
Lag 0, Symptoms:
 Cough OR = 0.93 (0 83, 1.04)
 Short, breath OR = 1.17 (1 03, 1 34)
7 day average, Symptoms:
 Cough OR = 0.94 (0.82, 1.08)
 Short, breath OR = 1 01 (0.86, 1.20)

-------
tB
i-t
o
sr
K>
O
O
   TABLE 6-20 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON SYMPTOMS IN STUDIES
  	OF ASTHMATICS	

                                                                                                                                      Effect measures standardized to 50 A'g/m3
                                                                                                                                      PM,0 (25 ^g/m3 PM2.5). Negative coefficients
                                                                                                                                      for lung function and ORs greater than 1 for
                                                                                                                                      other endpomts suggest PM effects
          Reference citation, location, duration,
          pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covanates
included, analysis problems, etc.
Results and Comments
Effects of co-pollutants
0\
 Tl
 H
 6
 o
 z
 o
 H
O
 c
 o
 H
 w
 n
 h-H
 H
 W
          Europe (cont'd)

          Hiltermann et al (1997)
          The Netherlands
Peters et al (1997b)
Erfurt, Germany
PM fractions measured over range of sizes from ultrafine
to fine, including PM,,,.
Mean PM,,, level  55 ^g/m3 (max 71)
Mean SO2.  100 ^g/m3 (max 383)
Peters et al (1997c)
Sokolov, Czech Republic
Winter 1991-1992
PM,0, SO,, TSP, sulfate, and particle strong acid
Median PM,,,  47 ^g/m3 (29, 73)
Median SO,' 46 /ug/m3 (22, 88)
Peters et al  (1997c)
Sokolov, Czech Republic
PMHI one central site  SO4 reported.
Mean PM,,,- 55 ,ug/m3, max 177 /Jg/
SO4 - fine,  mean 8 8 //g/m3, max 23
Sixty outpatient asthmatics examined for
nasal inflammatory parameters in The
Netherlands from July 3 to October 6, 1995
Associations of log transformed inflammatory
parameters to 24-h PM,,, analyzed, using a
linear regression model  Mugwort-pollen and
Oj were evaluated

Study of 27 non-smoking adult asthmatics
living in Erfurt, Germany during winter
season  1991 -1992  Diary used to record
presence of cough. Symptom information
analyzed using multiple logistic regression
analysis
Study of 89 children with asthma in Sokolov,
Czech Republic Subjects kept diaries and
measured peak flow for seven months during
winter of 1991 -2  Logistic regression for
binary outcomes used  First order
autocorrelations were observed and corrected
for using polynomial distributed lag
structures

Role of medication use evaluated in panel
study of 82  children, mean ages 9.8 yr., with
mild asthma in Sokolov, Czech Republic
Nov 1991 -Feb 1992.  Linear and logistic
regression evaluated PM,,,, SO,, temp, RH
relationships to respiratory symptoms.
                                                                                              Inflammatory parameters in nasal lavage of
                                                                                              patients with intermittent to severe
                                                                                              persistent asthma were associated with
                                                                                              ambient O3 and allergen exposure, but not
                                                                                              with PM,,, exposure
                                                                                                        Weak associations found with 5 day mean
                                                                                                        sulfates and respiratory symptoms
Significant relationships found between
TSP and sulfate with both phlegm and
runny nose
Medicated children, as opposed to those not
using asthma medication, increased their
beta-agonist use in direct association with
increases in 5-day mean of SO4 particles
<2 5 //m, but medication did not prevent
decrease in PEF and increase in prevalence
of cough attributable to PM air pollution
Lag 0, PM,,,-
 Cough OR = 1.32(1 16, 1.50)
 Feeling ill OR = 1 20(1 01, 1.44)
5 Day Mean,  PM,,,:
 Cough OR = 1 30(1 09, 1 55)
 Feeling ill OR = 1.47(1 16, 1 86)
LagO, PM,,:
 Cough OR = 1.19(1.07, 1.33)
 Feeling ill OR = 1 24(1 09, 1 41)
5 Day Mean,  PM2,:
 Cough OR = 1 02(0.91, 1 15)
 Feeling ill OR = 1 21 (1.06, 1 38)

Lag 0, Symptoms
 Cough OR = 1 01 (097, 1.07)
 Phlegm OR  = 1.13(1 04, I 23)
5 Day Mean,  Symptoms
 Cough OR= 1.10(1.04, 1 17)
 Phlegm OR = 1  17(1 09, 1.27)
Cough 1.16(1 00, 1 34) 6.5 //g/m' increase
5-day mean SO4
5-d Mean S04/mcrease of 6 5 ,ug/m3
Beta-Agonist Use       1.46 (1.08, 1 98)
Theophyllme Use       0.99 (0 77, 1 26)
No PM,,, analysis

-------
 g.
 O
 O
   TABLE 6-20 (cont'd). SHORT-TERM PARTICULATE  MATTER EXPOSURE EFFECTS ON SYMPTOMS IN STUDIES
                                                                         OF ASTHMATICS
                                                                                                                                     Effect measures standardized to 50 ^g/m3
                                                                                                                                     PMH1 (25 ^g/m3 PM,.S). Negative coefficients
                                                                                                                                     for lung function and ORs greater than 1  for
                                                                                                                                     other endpomts suggest PM effects
           Reference citation, location, duration,
           pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covanates
included, analysis problems, etc.
Results and Comments
Effects of co-pollutants
 T)
 H
 6
 o
 2
 o
 H
/O
 G
 O
 H
 tfl
 O
 &
 O
 H-H
 H
 W
           Europe (cont'd)

           Neukirch et al. (1998)
           Pans, France
           SO2, NO2, PMI3 and BS.
           Segalaetal. (1998)
           Paris, France
           SO,, NO,, PM,3 (instead of PM,,,), and BS
           Guntzeletal. (1996)
           Switzerland
           Taggartetal. (1996)
           Northern England
           SO2, NO2 and BS
Latin America

Romieuetal  (1997)
Mexico City, Mexico
During study period, max daily 1 -h O3 range: 40 to 390
ppb (mean 196 ppb SD = 78 ppb)
PMH) daily average range:  12 to 126 ^g/m3.
Romieuetal  (1996)
During study period, max daily range  40 to 370 ppb
(mean 190 ppb, SD = 80 ppb).
24 h ave PM]0 levels range:  29 to 363 /^g/m3 (mean
166.8 Mg/m3, SD 72.8 //g/m3).
PM i,, levels exceeded 150 ^g/m3 for 53% of study days.
24-h ave. PM2S levels range 23-177 ^g/rn3 (mean
85.7
                                                    Panel of 40 nonsmoking adult asthmatics in
                                                    Pans studied.  GEE models used to associate
                                                    health outcomes with air pollutants Models
                                                    allowed for time-dependent covanates,
                                                    adjusting for time trends, day of week, temp
                                                    and humidity.

                                                    Study of 43 mildly asthmatic children aged
                                                    7-1 Syr in Pans  Patients followed Nov  15,
                                                    1992 to May 9, 1993  Respiratory symptoms
                                                    recorded daily in diary.  An autoregressive
                                                    model fitted to data using GEE methods
                                                    Covanates included temp, and humidity

                                                    An asthma reporting system was used in
                                                    connection with pollutant momtonng in
                                                    Switzerland from fall of 1988 to fall 1990.
                                                    A  Box-Jenkins ARIMA time senes model was
                                                    used to relate asthma to TSP, O3, SO2> and
                                                    NO2 after adjusting for temperature.

                                                    Panel of 38 adult asthmatics studied July 17
                                                    to  Sept 22, 1993 in northern England.  Used
                                                    generalized linear model to relate pollutants to
                                                    bronchial hyper-responsiveness, adjusting for
                                                    temperature.
Study of 65 children with mild asthma aged
5-13 yr living in southwest Mexico City
Respiratory symptoms recorded by the
parents in daily diary An autoregressive
logistic regression model used to analyze
presence of respiratory symptoms
Study of 71 children with mild asthma aged
5-7 yr living in northern Mexico City.
Respiratory symptoms recorded by parents in
daily diary. An autoregressive logistic
regression model was used to analyze the
presence of respiratory symptoms.
                                          Significant relationships found for
                                          incidence of respiratory symptoms and
                                          three or more day lags of SO,, and NO2.
                                          Only selected results were given.
                                          Effects found related to PM,3 were less than
                                          those found related to the other pollutants
                                          No significant relationships found.
                                          Small effects seen in relation to NO2 and
                                          BS
Strongest relationships found between O3
and respiratory symptoms.
Cough and LRI were associated with
increased O3 and PMIO levels.
                                         Significant relationships found between
                                         incidence of respiratory symptoms and three
                                         or more day lags of PMI3
                                        Lag 2, Symptoms:
                                          Short. Breath OR = 1.22(0.83, 1.81)
                                          Resp. Infect. OR = 1 66 (0 84, 3 30)
Lag 0, Symptoms:
 Cough OR = 1.05 (0.92, 1 18)
 Phlegm OR = 1.05 (0 83, 1  36)
 Diff Breath OR = 1 13 (0.95, 1 33)
Lag 2, Symptoms
 Cough OR = 1 00(0.92, 1 10)
 Phlegm OR = 1.00(086, 1  16)
 Diff Breath OR = 1.2(1 1, 1.36)

PM,0 (lag 0) increase of 50 ^g/m3 related to
LRI= 1.21 (1.10, 1.42)
Cough = 1.27(1.16, 1.42)
Phlegm = 1.21 (1 00, 1 48)
PM25 (lag 0) increase of 25 ^g/m1 related to'
LR1= 1.18(1.05, 1.36)
Cough = 1 21 (1 05, 1 39)
Phlegm = 1.21 (1.03, 1 42)

-------
            Romieuetal.(1996)
                 (Mexico)
          Pekkanen etal.(1997)
                (Finland)
             Gielenetal.(1997)
               (Netherlands)
            Romieuetal.(1997)-
                 (Mexico)
                             -10               -5                0                 5
                                       Change in Pulmonary Function, L/min

       Figure 6-8. Selected acute pulmonary function change studies of asthmatic children.
                   Effect of 50 yug/m3 PM,0 on morning Peak flow lagged one-day.
 1      for PM2 5 and PM10. The study of Peters et al. (1997b) found slightly larger effects for PM2 5.
 2      The study of Schwartz and Neas (2000) found larger effects for fine particle measures (PM2 5,
 3      sulfate, etc.) than for the coarse mode. Naeher et al. (1999) found that H+ was significantly
 4      related to a decrease in morning PEF.  Overall, then,  PM10 and PM2 5 both appear to affect lung
 5      function in asthmatic, but there is only limited evidence for a stronger effect of fine- versus
 6      coarse-mode particles. Also, of the studies provided, few if any analyses were able to separate
 7      out the effects of PMi0 and PM2 5 from other pollutants. Gold et al. (1999) attempted to study the
 8      interaction of PM2 5 and ozone on PEF. The authors found independent effects of the two
 9      pollutants, but found that the joint effect was slightly less than the sum of the independent
10      effects.
11             The effects  on respiratory symptoms in asthmatics tended  to be positive, although they
12      were much less consistent than the effects on lung function. Most studies showed increases in
13       cough, phlegm, difficulty breathing, and bronchodilator use, although these increases were
14       generally not statistically significant as shown in Figure 6-9 for cough as an example.  Cough is a
       March 2001
6-193
DRAFT-DO NOT QUOTE OR CITE

-------
            Vedaletal. (1998)
                (Canada)
           Romieuetal. (1997)
                (Mexico)
            Gielenetal. (1997)
              (Netherlands)
           Peters etal. (1997a)
            (Czech Republic)
                             01234567
                                               Odds Ratios for Cough

       Figure 6-9.  Odds ratios with 95% confidence interval for cough per SO-^g/m3 increase in
                   PM10 for selected asthmatic children studies at lag 0.
 1     typical symptom outcome studied.  Several studies included both PM10 and PM2 5 in their
 2     analyses. The studies of Peters et al. (1997b) and Tiittanen et al. (1999) found similar effects for
 3     the two PM measures, whereas the Romieu et al. (1996) study found slightly larger effects for
 4     PM2 5.  Also, the Schwartz and Neas (2000) analyses indicated stronger effects of fine particle
 5     measures (PM2 5, sulfate) than coarse particles on respiratory symptoms in asthmatic school
 6     children in eastern United States urban areas.
 7
 8     6.3.3.1.2 Lung Function and Respiratory Symptom Effects in Nonasthmatic Subjects
 9          Results of the PM10 peak flow analyses in non-asthmatic studies (see Table 6-21) were
10     inconsistent, with fewer studies reporting results in the same manner as for the asthmatic studies.
11     Many of the point estimates showed increases rather than decreases. Similar results were found
12     in the PM2 5 studies.  The effects  on respiratory symptoms in non-asthmatics (see Table 6-22)
13     were similar to those in asthmatics. Most studies showed that PM10 increases cough, phlegm,
14     difficulty breathing,  and bronchodilator use, although these increases were generally not
       March 2001
6-194
DRAFT-DO NOT QUOTE OR CITE

-------
I
to
o
o
 TABLE 6-21.  SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION TESTS IN
                                                             STUDIES OF NONASTHMATICS

                                                                                                                                    Effect measures standardized to 50 /^g/m3
                                                                                                                                    PMH) (25 Mg/m3 PMM)  Negative coefficients
                                                                                                                                    for lung function and ORs greater than  1 for
                                                                                                                                    other endpoints suggest PM effects
          Reference citation, location, duration,
          pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covanates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
 Tl
 H
 z;
 o
 H
O
 G
 O
 H
 W
 O
          United States
          Hoeketal. (1998)
           Lee and Shy (1999)
           North Carolina
           Mean 24 h PM,0 cone, over two years: 25 1
Komcketal. (1998)
Mt. Washington, NH
O3 levels measured at 2 sites near top of the mountain
PM2 5 measured near base of the mountain

Naeheretal. (1999)
Virginia
PM10, PM,,, sulfate fraction, H+, and ozone
Neasetal (1996)
State College, PA
PM2,: mean 23 5, max 85
Results summarized from several other studies
reported in the literature.  These included.
asymptomatic children in the Utah Valley (Pope
et al, 1991), children in Bennekom, NL (Roemer
et al., 1993), children in Uniontown, PA (Neas et
al, 1995), and children in State College, PA
(Neas et al., 1996)  Analyses done using a first-
order autoregressive model with adjustments for
time trend and ambient temp.

Study of the respiratory health status of residents
whose households lived in six communities near
an incinerator in southwestern North Carolina.
Daily PEFR measured in the afternoon was
regressed against 24 hour PMU) level lagged by
one day. Results were adjusted for gender, age,
height, and hypersensitwity.

Study of the effects of air pollution on adult
hikers on Mt. Washington, NH Linear and non-
linear regressions used to evaluate effects of
pollution on lung function

Daily change in PEF studied in 473 non-smoking
women in Virginia during summers 1995-1996.
Separate regression models  run, using normalized
morning and evening PEF for each individual

Study of 108 children in State College, PA, during
summer of 1991 for daily variations in symptoms
and PEFRs in relation to  PM2, An autoregressive
linear regression model was used  The regression
was weighted by reciprocal  number of children of
each reporting period.  Fungus spore cone., temp.,
O3 and SO, were examined.
                                                                                                 Other pollutants not considered
                                                                                                  PMH) was not related to variations in
                                                                                                  respiratory health as measured by
                                                                                                  PEFR.
                                                                                                            PM2 5 had no effect on the O,
                                                                                                            regression coefficient.
Ozone was only pollutant related to
evening PEF
Spore concentration associated with
deficient in morning PERF.
                                   Significant decreases in peak flow found to be
                                   related to PM,,, increases
Morning PEF decrements were associated with
PM,0, PM,5, and H+.  Estimated effect from
PM25 and PM10 was similar No PM effects
found for evening PEF

PM,, (25 Mg/m1) related to RR of
PM'PFER (lag o) = - o 05 (-1.73, o.63)
PM PEFR (lag 1) =-064 (-1.73, 044)
 o
 h— I
 H
 W

-------
to
3
tr
O
O
   TABLE 6-21  (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
                                                       TESTS IN STUDIES OF NONASTHMATICS
                                                                                                                                     Effect measures standardized to 50 ^g/m3
                                                                                                                                     PM]0 (25 Aig/mJ PM2.5). Negative coefficients
                                                                                                                                     for lung function and ORs greater than 1 for
                                                                                                                                     other endpomts suggest PM effects
           Reference citation, location, duration,
           pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covanates included,
analysis problems, etc
Results and Comments
Effects of co-pollutants
 o
 z
 o
 H
O
 d
 o
 o
 H
           United States (cont'd)

           Neasetal. (1999)
           Philadelphia, PA
           Median PM,0 level' 31  6 in SW camps,
           27.8 in NE camps (1QR ranges of about 18).
           Median PM,5 level 22.2 in the SW camps,
           20.7 in NE camps (1QR ranges about 16.2 and 12.9,
           respectively).
           Particle-strong acidity,  fine sulfate particle, and O3 also
           measured.
           Schwartz and Neas (2000)
           Eastern U S.
           PM,5 and CM (PM,,,.,5) measured.
           Summary levels not given.
           Linnetal. (1996)
           So California
           NO, ozone, and PM5 measured.
Europe

Boezenetal (1999)
Netherlands
PM,,,, BS, SO2, and NO2 measured.
                                                     Panel study of 156 normal children attending
                                                     YMCA and YWCA summer camps in greater
                                                     Philadelphia area in 1993. Children followed for
                                                     at most 54 days. Morning and evening deviations
                                                     of each child's PEF were analyzed using a mixed-
                                                     effects model adjusting for autocorrelation
                                                     Covanates included time trend and temp.  Lags
                                                     not used in the analysis
                                                     Analyses for 1844 school children in grades 2-5
                                                     from six urban areas in eastern U S. and from
                                                     separate studies from Uniontown and State
                                                     College, PA. Lower resp symptoms, cough and
                                                     PEF used as endpomts.  The authors replicated
                                                     models used in the original analyses. CM and
                                                     were used individually and jointly in the analyses.
                                                     Sulfate fractions also used in the analyses.
                                                     Details of models not given.

                                                     Study of 269 school children in Southern
                                                     California twice daily for one week in fall, winter
                                                     and spring for two years. A repeated measures
                                                     analysis of covanance was used to fit an
                                                     autoregressive model, adjusting for year, season,
                                                     day of week, and temperature
Data collected from children during three winters
(1992-1995) m rural and urban areas of The
Netherlands.  Study attempted to investigate
whether children with bronchial
hyperresponslveness and high serum Ige levels
were more susceptible to air pollution.  Prevalence
of a 10 percent PEF decrease  was related to
pollutants for children with bronchial
hyperresponslveness and high serum Ige levels.
                                             Analyses that included sulfate fraction
                                             and Oj separately also found
                                             relationship to decreased flow. No
                                             analyses reported for multiple
                                             pollutant models.
                                             Sulfate fraction was highly correlated
                                             with PM, s (094), and, not
                                             surprisingly, gave similar answers.
                                             Morning FVC was significantly
                                             decreased as a function of PM5 and
                                             NO,
No consistent pattern of effects
observed with any of the pollutants
for 0, 1, and 2 day lags
                                   LagO,PM10-
                                    MommgPEF = -8.16(-1481, -1.55)
                                    Evening PEF = - 1.44 (-7.33, 4.44)
                                   5 day ave, PM10
                                    Morning PEF = 2 64 (-6 56, 11.83)
                                    Evening PEF= 1.47 (-7.31, 1022)
                                   LagO,PM25
                                     Morning PEF = -3 28 (-6 64, 0.07)
                                     Evening PEF =-0.91 (-404,221)
                                   5 day ave , PM2 s
                                    Morning PEF = 3.18 (-2.64, 9.02)
                                    Evening PEF = 0.95 (-4.69, 6.57)

                                   Uniontown Lag 0,PM, 5 :
                                    Evening PEF = - 1 52 (-2 80, -0.24)
                                   State College Lag 0, PM25
                                    Evening PEF = -0.93 (-1.88, 0.0!)

                                   Results presented for CM showed no effect
                                   Results for PM,,, were not given

-------
e
o
o
o
             TABLE  6-21 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
            =	                                        TESTS IN STUDIES OF NONASTHMATICS

                                                                                                                                               Effect measures standardized to 50 ^g/m'
                                                                                                                                               PM10 (25 fj.j>/m3 PM2.S).  Negative coefficients
                                                                                                                                               for lung function and ORs greater than 1 for
                                                                                                                                               other endpomts suggest PM effects
          Reference citation, location, duration,
          pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covanates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
 O
 O
 2
 o
 H
XD
 c
 o
 w
 o
          Europe (cont'd)

          Prisoner et al. (1999)
          Austna
          PM,,, measured gravimettncally for 14-d periods.
          Annual mean PM,0 levels range:  136-229 ,ug/m3.
          O, range. 39 1  ppb - 18.5 pbs between sites.
          Gnevmketal (1999)
          Netherlands
          PM,(J and BS.
           Kiinzhetal (2000)
          Roemeretal.(2000)
          PM10 means for 17 panels ranged 11 2 to 98.8
          SO,, NO,, and elemental content of PM also measured
At nine sites in Austria during 1994, 1995, and
1996, a longitudinal study designed to evaluate O3
was conducted  During 1994 - 1996, children
were measured for FVC, FEV, and MEF5() six
times, twice a year in spring and fall  1060
children provided valid function tests  Mean age
7.8 ± 0 7 yr GEE models used. PM,,,, SO2, NO2,
and temp, evaluated.

A panel of adults with chronic respiratory
symptoms studied over two winters in The
Netherlands starting in 1993/1994. Logistic
regression analysis was used to model the
prevalence of large PEF decrements.
Individual linear regression analysis of PEF on
PM was calculated and adjusted for time trends,
influenza incidence, and meteorological variables

Ackermann-Liebnch et al  (1997) data reanalyzed.
Authors showed that a small change in FVC
(-3.14 percent) can result in a 60% increase in
number of subjects with FVC less than 80 percent
of predicted.

Combined results from 1208 children divided
among 17 panels studied.  Separate results
reported by endpomts included symptoms as
reported in a dairy and PEF. Individual panels
were analyzed using multiple linear regression
analysis on deviations from mean PEF adjusting
for auto-correlation.  Parameter estimates were
combined using a fixed-effects model where
heterogeneity was not present and a random-
effects model where it was present
                                                                                                            Small but consistent lung function
                                                                                                            decrements m cohort of school
                                                                                                            children associated with ambient O,
                                                                                                            exposure
                                                                                                            Subjects with low levels of serum
                                                                                                            P-carotene more often had large PEF
                                                                                                            decrements when PM,0 levels were
                                                                                                            higher, compared with subjects with
                                                                                                            high serum P-carotene.
                                                                                                            Results suggested serum P-carotene
                                                                                                            may attenuate the PM effects on
                                                                                                            decreased PEF.

                                                                                                            The results were for two hypothetical
                                                                                                            communities, A and B
Daily concentrations of most elements
were not associated with the health
effects.
                                   PM,0 showed little variation in exposure
                                   between study site.  For PM,0, positive effect
                                   seen for winter exposure but was completely
                                   confounded by temperature

                                   PMIO Summertime
                                   P = 0.003 SE0.012p=0.77
                                                                                                                                                PM10 analyses not focus of this paper.
 o
 HH
 H
 tfl

-------
 3
 to
 o
 o
TABLE 6-21  (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
                                                    TESTS IN STUDIES OF NONASTHMATICS

                                                                                                                                  Effect measures standardized to 50 ^g/mj
                                                                                                                                  PM1(,(25Mg/m3 PMM).  Negative coefficients
                                                                                                                                  for lung function and ORs greater than 1  for
                                                                                                                                  other endpomts suggest PM effects
           Reference citation, location, duration,
           pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covanates included,
analysis problems, etc
Results and Comments
Effects of co-pollutants
CO
 H
 a
 o
 2
 o
 H
O
 c
 o
           Europe (cont'd)

           Scarlett et al. (1996)
           PMH1, 03, and NO2 measured.
           van der Zee et al (1999)
           Netherlands
           PM1(I averages ranged 20 to 48 ,ug/m5
           BS, sulfate fraction, SO,, and NO2 also measured
           van der Zee et al. (2000)
           Netherlands
           PM|0 averages ranged 24 to 53 £ig/m3.
           BS, sulfate fraction, S02, and NO2 also measured.
                                                  In study of 154 school children, pulmonary
                                                  function was measured daily for 31 days.
                                                  Separate autoregresswe models for each child
                                                  were pooled, adjusting for pollen, machine,
                                                  operator, time of day, and time trend.

                                                  Panel study of 795 children aged 7 to 11 years,
                                                  with and without chronic respiratory symptoms
                                                  living in urban and nonurban areas in the
                                                  Netherlands Peak flow measured for three
                                                  winters starting in 1992/1993. Peak flow
                                                  dichotomized at 10 and 20% decrements below
                                                  the  individual median  Number of subjects was
                                                  used as a weight. Minimum temperature day of
                                                  week, and time trend variables were used as
                                                  covanates. Lags of 0, 1 and 2 days were used, as
                                                  well as  5 day moving average

                                                  Panel study of 489 adults aged 50-70 yr, with and
                                                  without chronic respiratory symptoms, living in
                                                  urban and nonurban areas in the Netherlands.
                                                  Resp symptoms and peak flow measured for three
                                                  winters starting in 1992/1993. Symptom
                                                  variables analyzed as a panel instead of using
                                                  individual responses.  The analysis was treated as
                                                  a time series, adjusting for first order
                                                  autocorrelation. Peak flow dichotomized at 10
                                                  and 20% decrements below the individual
                                                  median  The number of subjects used as a weight.
                                                  Minimum temp , day of week, and time trend
                                                  variables used  as covanates Lags of 0, 1 and 2
                                                  days used, as well as 5  day moving average.
                                             PMIO was related to changes in FEV
                                             and FVC
                                             In children with symptoms,
                                             significant associations found
                                             between PM10, BS and sulfate fraction
                                             and the health endpoints.  No multiple
                                             pollutant models analyses reported.
                                             BS tended to have the most consistent
                                             relationship across endpoints Sulfate
                                             fraction also related to increased
                                             respiratory effects. No analyses
                                             reported for multiple pollutant
                                             models  Relationship found between
                                             PM ,„ and the presence of 20%
                                             decrements in symptomatic subjects
                                             from urban areas
                                   Lag 0, PM,,,, Urban areas
                                    Evening PEF OR = 1 15(1.02,1.29)
                                   Lag 2, PM10, Urban areas
                                    Evening PEF OR = 1.07 (0.96, 1.19)
                                   5 day ave, PM|0, Urban areas
                                    Evening PEF = 1.13(0.96, 1.32)
                                   Lag 0, PM,0, Urban areas
                                    Morning large decrements
                                    OR =1.44 (1.02, 2.03)
                                   Lag 2, PM,0, Urban areas
                                    Morning large decrements
                                    OR= 1 14(0.83, 1.58)
                                   5 day ave, PMIO, Urban areas
                                    Morning large decrements
                                    OR= 1.16(0.64,2 10)

                                   Results should be viewed with caution because
                                   of problems in analysis
 n
 H

-------
O
O
TABLE 6-21 (cont'd).  SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
                                                  TESTS IN STUDIES OF NONASTHMATICS

                                                                                                                               Effect measures standardized to 50 ^g/mj
                                                                                                                               PM]0 (25 //g/m1 PM2.S). Negative coefficients
                                                                                                                               for lung function and ORs greater than 1 for
                                                                                                                               other endpomts suggest PM effects
          Reference citation, location, duration,
          pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
                                                                                                          Results and Comments
                                                                                                          Effects of co-pollutants
w
?
Tl
H
6
o
 o
 H
O
 o
 H
          Europe (cont'd)

          Tnttanenetal. (1999)
          Kupio, Finland
          Median PM,0 level: 28 (25*, 75* percentiles = 12, 43)
          Median PM2 5 level  15 (25*, 75lh percentiles = 9, 23)
          Black carbon, CO, SO2, NO2, and 0, also measured.
          Ward et al. (2000)
          West Midlands, UK
          Daily measurements of PM,0, PM,5, SO2, CO, O3, and
          oxides of nitrogen.
          Cuijpers et al. (1994)
          Maastricht, NL
          SO2, NO,, BS, ozone, and H+ measured
                                                 Six-week panel study of 49 children with chronic
                                                 respiratory disease followed in the spring of 1995
                                                 in Kuopio, Finland  Morning and evening
                                                 deviations of each child's PEF analyzed, using a
                                                 general linear model estimated by PROC MIXED.
                                                 Covariates included a time trend, day of week,
                                                 temp , and humidity Lags of 0 through 3 days
                                                 were used, as well as a 4-day moving average
                                                 Vanous fine particles were examined
                                                 Panel study of 9 yr old children in West Midlands,
                                                 UK for two 8-week penods representing winter
                                                 and summer conditions  Individual PEF values
                                                 converted to z-values. Mean of the z-values
                                                 analyzed in a linear regression model, including
                                                 terms for time trend, day of week, meteorological
                                                 variables, and pollen count. Lags up to four days
                                                 also used

                                                 Summer episodes in Maastricht, The Netherlands
                                                 studied  Paired t tests used for pulmonary
                                                 function tests
                                            Ozone strengthened the observed
                                            associations. Introducing either NO2
                                            or SO, in the model did not change
                                            the results markedly  Effects vaned
                                            by lag Separating effects by size was
                                            difficult.
                                            Results on effects of pollution on lung
                                            function to be published elsewhere.
                                            Small decreases in lung function
                                            found related to pollutants.
                                                                                                                                            Lag 0, PM10.
                                                                                                                                             Morning PEF= 1.21 (-043,2.85)
                                                                                                                                             Evening PEF = 0.72 (-0.63, 1 26)
                                                                                                                                            4 day ave, PMI()
                                                                                                                                             Morning PEF = -1 26 (-5 86, 3.33)
                                                                                                                                             Evening PEF = 2  33 (-2.62, 7 28)
                                                                                                                                            LagO,PM2!
                                                                                                                                             Morning PEF = 1  11 (-0 64, 2.86)
                                                                                                                                             Evening PEF = 0  70 (-0 81, 2 20)
                                                                                                                                            4 day ave., PM, 5
                                                                                                                                             Morning PEF=-1 93 (-7 00, 3.15)
                                                                                                                                             Evening PEF = 1.52 (- 3 91, 6 94)
 n
 HH
 H
 m

-------


z
o
HH
H
U
Z
P
fa
^
£
^^
Z
0
I^H
J
P
o.
br
O

H
^
H
tt. en
fe u
p
S "^
^ S
C/2 ffi
O H
X ^
td Z
*M PARTICULATE MATTER
TESTS IN STUDIES OF NO
^•^
H
I
H
OS
O
C/3

?
O
u

M
^»
w
J
CO
•<
H

2
S^ ^
(2
^ O X
o ill
T3 ™ E U
£^ £^
a ~;O W
"c J?"g M
» a- c "
S ^c 2 2
3 -JoS .|
Islf
O o_S u.

pi] £1, t2 5






11
i -
J1-
^ 2
1 "s
3 tS
^ ;H
tt- LL4


Type of study, sample size, health outcomes
measured, analysis design, covanates included,
analysis problems, etc.






- i
0 M
•O &>
g E
'-y C
CJ  t>
f §iS
CQ -o U4


Is
o S
Ij ^00 ^

C c S"
e "™ *O
S | s
-^ a> 03
"c o «
8-8 s
118
•- S 'S
o"ll


Peak flow studied in a panel of 40 school-aged
children living in southwest Mexico City. Daih
deviations from morning and afternoon PEFs








^
a.
a.
r^
to
£,!!
a S 2 ~
•| O^-O"
c "M O
^ " O 

c J
t= ^ •
lgl
» ll
cii M a.


Study of 40 subjects aged over 55 years with
COPD living in Chnstchurch, New Zealand
conducted during winter of 1994. Subjects
recorded their peak flow measurements. A log-
linear regression model with adjustment for firsl
order auto-correlation was used to analyze peak
flow data and a Poisson regression model was







T3
is
£
0
U
PN "!.
0\ Z 2
C..E i £
"« 3 A
•e ^ z-
xSS
dat
                                        5.




























•S
^
f
u
.§
*o
o
c
5
t
CL>
o
u
i
1
; one-pollutant model,
£
S

In 3 Taiwan communities m 1995, PM,U by B-








~
0
•«
tj
1
6











OS
rt
2
•5
o
o
Q
rt
S.





OO
£•
rvi
a
a.
•0
(3
£
ficantly affected both I
3
2?
55
,
gauge measured at selected pnmary schools in
each community. Spirometry tests (FVC, FEV,











§
£
(2
03
CL
1=
a1
p
a
oa
G
S
Q,
00
00
"5
»



Ch
o
(^
o
1
y
u.
u
>
_L
•8
1
ra
8
o"
C/}
0*
>~
u.

FEF2W5.,,, PEF) obtained m penod May 1995 to









c
o
S
1
o
c
s












i showed nonsigmfican
S
a.
-a










0>
~- ~S
li
;ment. No significant i
mstrated in the model :
u U
T3 T3

Jan. 1996 using ATS protocol m study pop. age
8 to 13 yr. 895 children were analyzed. Study
was designed to investigate short-term effect of







































C
00
—
Cfl
03
•a
r-

-------
p
g.
O
O
      TABLE 6-22.  SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON SYMPTOMS IN STUDIES  OF
                                                                      NONASTHMATICS

                                                                                                                                    Effect measures standardized to 50 ^g/m1
                                                                                                                                    PM10 (25 Aig/m' PMM) Negative coefficients
                                                                                                                                    for lung function and ORs greater than 1 for
                                                                                                                                    other endpomts suggest PM effects
          Reference citation, location, duration,
          pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covanates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
           United States

           Schwartz and Neas (2000)
           Eastern U.S.
           PM25 and CM (PM2, to 10) measured
           Summary levels not given
          Zhang et al. (2000)
          Vmton,  Virginia
          24- h PM,,,, PM, s, sulfate and strong acid measured in
          1995
           Canada
                                                     Reported on analysis of 1844 school children in
                                                     grades 2-5 from six urban areas m the eastern
                                                     U.S., and from separate studies from Umontown
                                                     and State College, PA. Lower respiratory
                                                     symptoms, and cough used as endpomts  The
                                                     authors replicated the models used in the original
                                                     analyses  CM and PM, 5 were used individually
                                                     and jointly in the analyses. Sulfates fractions
                                                     were also used in the analyses  Details of the
                                                     models were not given

                                                     In southwestern Virginia, 673 mothers were
                                                     followed June 10 to Aug 31, 1995 for the daily
                                                     reports of present or absence of runny or stuffy
                                                     nose. PM indicator, O,, NO, temp , and  random
                                                     sociodemographic charactenstlcs considered.
                                             Sulfate fraction was highly correlated
                                             with PM25 (0 94), and not
                                             surprisingly gave similar answers
                                             Of all pollutants considered, only the
                                             level of coarse particles as calculated
                                             (PM10 - PM2 5) independently related
                                             to incidence of new episode of runny
                                   PM25 was found to be significantly related to
                                   lower respiratory symptoms even after
                                   adjusting for CM, whereas the reverse was not
                                   true However, for cough, CM was found to be
                                   significantly related to lower respiratory
                                   symptoms even after adjusting for PMZ 5,
                                   whereas the reverse was not true.
g
H
O
           Longetal (1998)
           Winnepeg, CN
           PM,0, TSP, and VOC measured.
Europe

Boezenetal (1998)
Amsterdam, NL
PMIO, SO,, and NO, measured.
                                                     Study of 428 participants with mild airway
                                                     obstruction conducted dunng a Winnepeg
                                                     pollution episode Gender specific odds ratios of
                                                     symptoms were calculated for differing PMH,
                                                     levels using the Breslow-Day test
Study of 75 symptomatic and asymp. adults near
Amsterdam for three months dunng winter 1993-
1994  An autoregressive logistic model was used
to relate PM,,, to respiratory symptoms, cough,
and phlegm, adjusting for daily mm. temp , time
trend, day of week
                                             Cough, wheezing, chest tightness, and
                                             shortness of breath were all increased
                                             dunng the episode
No relationship found with pulmonary
function. Some significant
relationships with respiratory disease
found in subpopulations
 n
 h— <
 H
 M

-------
g.
O
O
TABLE 6-22 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON SYMPTOMS IN STUDIES OF
                                                                       NONASTHMATICS

                                                                                                                                   Effect measures standardized to 50 //g/m3 PM,,,
                                                                                                                                   (25 ^g/m3 PM2.5) Negative coefficients for
                                                                                                                                   lung function and ORs greater than 1 for other
                                                                                                                                   endpomts suggest PM effects
          Reference citation, location, duration,
          pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covanates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
Tl
H
6
O
2!
O
H
O
H
W
O
t— (
H
tn
          Europe (cont'd)

          Roemeretal (1998)
          Mean PMI(, levels measured at local sites ranged
          11 2 to 98.8 /^g/m3 over the 28 sites.
          Roemer et al. (2000)
          PMH1 means for the  17 panels ranged 11 2 to
          98 8 ,ug/m3.
          SO2, NO2, and PM elemental content also measured
van der Zee et al (1999)
Netherlands
PM,,, averages ranged 20 to 48 Aig/m!.
BS, sulfate fraction, S0;, and NO2 also measured
Pollution Effects on Asthmatic Children in
Europe (PEACE) study was a multi-center study
of PM,,,, BS, S02, and NO2 on respiratory health
of children with chronic respiratory symptoms
Results from individual centers were reported by
Kotesovec et al  (1998), Kalandidi et al (1998),
Haluszka et al. (1998), Forsberg et al (1998),
Clench-Aas et al (1998), and Beyer et al. (1998).
Children with chronic respiratory symptoms
were selected into the panels. The symptom with
one of the larger selection percentages was dry
cough (range over sample of study communities
29 to 92% [22/75; 84/91] with most values over
50%)  The group as a whole characterized as
those with chronic respiratory disease, especially
cough.

Combined results from 1208 children divided
among  17 panels studied Endpomts included
symptoms as reported in a dairy and PEF.
Symptom vanables analyzed as a panel instead of
using individual responses.  The  analysis was
treated as a time series, adjusting for first order
autocorrelation. Parameter estimates were
combined using a fixed-effects model where
heterogeneity was not present and a random-
effects model where it was present

A panel study of 795 children aged 7 to 11 yr,
with and without chronic respiratory symptoms,
living in urban and nonurban areas in the
Netherlands.  Respiratory symptoms measured for
3 winters starting 1992/1993  Symptom vanables
analyzed as a panel instead of using individual
responses. The analysis was treated as a time
senes, adjusting for first order autocorrelation.
The number of subjects was used as a weight
Minimum temp., day of week, and time trend
vanables used as covanates  Lags of 0, 1 and 2
days used, as well as 5 day moving average.
                                                                                              These studies modeled group rates and
                                                                                              are an example of the panel data
                                                                                              problem.
                                                                                              Daily concentrations of most elements
                                                                                              were not associated with the health
                                                                                              effects
                                      The analysis of PM](I was not a focus of this
                                      paper
In children with symptoms, significant
associations found between PM,,,, BS
and sulfate fraction and the health
endpomts  No analyses reported with
multiple pollutant models
Lag 0, PM,,,, Urban areas
 Cough OR = 1 04(0.95,  1.14)
Lag 2, PMIO, Urban areas
 Cough OR = 0.94 (0 89,  1 06)
5 day ave, PM,0, Urban areas
 Cough OR = 0.95 (0 80,  1.13)

-------
to
O
C/5
W
S
ta^
H


g
c^
O
SYMPT
O
05
H
U
to
to
r_
to
to


p
O5
0
0,
X
to
RM PARTICULATE MATTER
H
H
»
o
=

^^
"P
+*
o
u

fS
TABLE 6-2























O5
y

5
NONASTHM^














s
-C
,O *Q
"e 2 j3

° | S
o w "S
•o u S 2
y y « o
- -c 2 iJJ

S 00 M "
!*ll
OJ -^ n) OO
3 S C M
S OH 0 =
HII
III!





Cfl W
u .2
H ~-
0 g.
"c j
^2 vi
11
&£ Llj

Type of study, sample size, health outcomes
measured, analysis design, covanates included,
analysis problems, etc.




0

2 >
§1
•a c
si
'-£ £
8S
^ 	 	
Reference citation,
pollutants measure




%
t-
U oo 't' 1> ^ r*^ C — I
S3 p — ! ra ~ ~~. .2 — : ^

i-oocr\ t- o\ o^ c/T oc o
-3o£,-3o££oc.
S" oo ^ 2~ — •» « m f~
c ^ P c o ° p ^ — :
&° ~ 1-7 "ii" <^° ~
Rs! « £0! o* u B! *
c/3 O O <" O O g O O
o" i/> ">  >^ c/3 ^
Sfj § g?J 3 -0 J 3
« _] j_j ra-j— ) w _j i— J
_ 1 i-J m



§_
&
3
•S
1 |
*" 'p.
1 1

rt O
(XI *-J
« 2
m §

Panel study of adults aged 50 to 70 yr dunng 3
consecutive winters starting in 1992/1993.
Symptom variables analyzed as a panel instead of
using individual responses Analysis treated as a
time senes, adjusting for first order
autocorrelation Number of subjects used as a
weight Mm. temp., day of week, time trend
vanables used as covanates. Lags 0, 1 and 2
days used, as well as 5 day moving average.
u"
1
H
ra
w
C

ffl

s|
f — •• p.., ^J
0 *» ?P
§. s §
C t!
9 « S^
1 S.s§
1 «18I
k. S -c E i
§• -o a >> g
S c £ ~ S
3 BJ OJ eg C
m > z Q i





^-V ^-V -— V ^— V
— OO fN OO
"L ^ "^ <^l

00 OO OO O
2. S- 2- i-
§OC rf —
•AI O ^ O
_j o _j ^_; «N (s
11 2 II ll g II
SO , 30 3 >, 3
O
O'-E
1 ? s>
S J §
^ s •§
o oo *-
« S 2
111
C i— •— r '
a^ •§ ^
g|||
||||
CD a w K

Six-week panel study of 49 children with chronic
respiratory disease followed in spnng 1 995 in
Kuopio, Finland Cough, phlegm, URS, LRS and
medication use analyzed, using a random effects
logistic regression model (SAS macro
GLIMMLX). Covanates included a time trend,
day of week, temp , and humidity. Lags of 0 to 3
days used, as well as 4-day moving average.
•o
N i
~ O1-
II t»- O
(/] ° —
~  X

Cu & ^
•S O
>ri - z
r~- ••
ti -^ O"
fs| J3 C/5
^_> rt
oo - o"
ST ^ 1Q o

"^ "u ly-i C
— > — O
^•°- ••§
--re- ^i rt
2 .2 ^ •«' U
0 1- 2 2 ^ .
§£^ c|l
C o ^ rt Cu 3
S a-5 _•§ „ s
. £ 3 S m 5 m g
t— X 2 Tt 2 (N C













C b
sa ~
I i
™ tS
M.-3
O 43
rt "ti
O
C rt
O ^"
*S ^"S
o S >
g> 4J
J3 ~
ll!
"° 2 3
Z "S o.

Symptoms of rhinitis and atopic status were
evaluated in 386 students grades 9 and 10 using
statistical package for the social sciences, Fisher











S
^
Kelesetal. (1999)
Istanbul, Turkey
Nov 199 6 to Jan


























tests, and multiple regression model as
Spearman's coefficient of correlation.
c

£
a. «
a-s
"s |
c o

E ?
3 ^
C 00
S :i-
2 0°
1 °°
<£ —
TSP levels ranged
unpolluted area to





































13
§
•1
Australia/New Zet

1
£
5
"S
^
o
c


S
"rt
'5 g
ll
rt CX
II
si
0- u

"2
Cfl
S
{J
s
•S
1 „
"2 3
o o
i —
§1
••* u
o I
z £

Study of 40 subjects aged over 55 yr, with COPD
living in Chnstchurch, New Zealand dunng
winter 1994 Subjects completed dianes twice






"S
c
3
ra
1
O
U
"*2
Harreetal (1997)
Chnstchurch, NZ
S02, NO,, PM,,,, ai


























daily Poisson regression model used to analyze
symptom data






















I
1




S
'o


§
w
s
bh
s
"5
es
o
c
00
•<

A cohort of 664 preschool children studied for












^
Asia
Awasthietal (19?













c
3

3
£
£
0
o.
1
Cfl
"O
C
S
e
^

two weeks each in northern India Ordinary least
squares was used to relate a respiratory symptom
complex pollutants


1
°_
u
I
c

C« 3
^-" ^
H 4>
« £
C -i
S g
5 s
i*
li
|s
"S0-
U __'
2 o-l?
•a a s
£ vi S
March 2001
6-203
DRAFT-DO NOT QUOTE OR CITE

-------
  1      statistically significant.  Three authors, Schwartz and Neas (2000), Tiittanen et al. (1999) and
  2      Neas et al. (1999), used PM10.2_5 as a coarse-mode (CM) particulate measure.  Schwartz and Neas
  3      (2000) found that CM was  significantly related to cough. Tiittanen found that one day lag of
  4      CM was related to morning PEF, but there was no effect on evening PEF. Neas et al. found no
  5      effects of CM on PEF.
  6
  7      6.3.3.2 Long-Term Particulate Matter Exposure Effects on Lung Function and
  8             Respiratory Symptoms
  9      6.3.3.2.1  Summary of the 1996 Particulate Matter Air Quality Criteria Document Key
10               Findings
11           In the 1996 PM AQCD, the available respiratory disease studies were limited in terms of
12      conclusions that could be drawn. At that time, three studies based on a similar type of respiratory
13      symptom questionnaire administered at three different times as part of the Harvard Six-City and
14      24-City Studies provided data on the relationship of chronic respiratory disease to PM.  All three
15      studies suggest a long-term PM exposure effect on chronic respiratory disease. The analysis  of
16      chronic cough, chest illness and bronchitis tended to be significantly positive for the earlier
17      surveys described by Ware et al. (1986) and Dockery et al. (1989). Using a design similar to the
18      earlier one, Dockery et al. (1996) expanded the analyses to include 24 communities in the United
19      States and Canada.  Bronchitis was found to be higher (odds ratio = 1.66) in the community with
20      the highest particle strong acidity when compared with the least polluted community. Fine
21      particulate sulfate was also associated with higher reporting of bronchitis (OR =  1.65, 95%
22      CI 1.12, 2.42).
23           Interpretation of such studies requires caution in light of the usual difficulties ascribed to
24      cross sectional studies. That is, evaluation of PM effects is based on variations in exposure
25      determined by a different number of locations.  In the  first two studies, there were six locations
26      and, in the third, twenty-four. The results seen in all studies were consistent with a PM gradient,
27      but it was impossible to separate out effects of PM and any other factors  or pollutants having the
28      same gradient.
29           Chronic pulmonary function studies by Ware et al. (1986), Dockery et al. (1989), and Neas
30      et al. (1994) had good monitoring data and well-conducted standardized pulmonary function
31      testing over many years, but showed no effect for children from airborne particle pollution
32      indexed by TSP, PM15, PM2 5 or sulfates. In contrast, the Raizenne et al.  (1996) study did find
        March 2001                               6-204       DRAFT-DO NOT QUOTE OR CITE

-------
  1      significant associations of FEV, or FVC effects in U.S. and Canadian children with both acidic
  2      particles and other fine PM indicators. Overall, the available studies provided only limited
  3      evidence suggestive of pulmonary lung function decrements being associated with chronic
  4      exposure to PM indexed by various measures (TSP, PM10, sulfates, etc.). However, it was noted
  5      that cross  sectional studies require very large sample sizes to detect differences because they
  6      cannot eliminate person to person variation, which is much larger than the within person
  7      variation.  Thus, lack of statistical significance cannot be taken as proof of no effect.
  8
  9      6.3.3.2.2  New Studies of Long-Term Particulate Matter Exposure Respiratory Effects
 10           Numerous studies have been published since  1996 which evaluate effects of long-term PM
 11      exposure on lung function and respiratory illness, as summarized in Table 6-23. The  new studies
 12      examining PM10 and PM2 5 in the United States include McConnell et al. (1999), Abbey et al.
 13      (1998), Berglund et al. (1999), Peters et al. (1999a,b), and Gauderman et al. (2000), which all
 14      examined effects in California cohorts but produced inconsistent results.  Probably most notable
 15      among these California study results, are those of McConnell et al. (1999) indicating that, as
 16      PM,0 increased across communities, a corresponding increase in bronchitis risk per interquartile
 17      range occurred, results consistent with those reported by Dockery et al. (1996), although the high
 18      correlation of PM10,  acid, and NO2 precludes clear attribution of the McConnell et al.  bronchitis
 19      effects specifically to PM alone.
20           As for other non-U.S. studies, particularly interesting results were obtained by Leonardi
21      et al. (2000) as part of the Central European Air Quality and Respiratory Health (CESAR) study.
22      Blood and serum samples were collected from school children aged 9-11 yrs. in each of 17
23      communities in Central Europe (N = 10 to 61 per city).  Numbers of lymphocytes increased as
24      PM concentrations increased across the cities. Regression slopes, adjusted for confounder
25      effects, were largest and statistically significant for PM2 5,  but small and non-significant for
26      PM10_2 5. A similar positive relationship was found between IgG concentration in serum and
27      PM2 5 gradient, but not for PM10 or PM10_2 5.  These results  tend to suggest a PM effect on immune
28      function more strongly due to ambient fine particle  than coarse particle exposure.
29           Other non-U.S. studies of interest examined other PM measures such as TSP and BS in
30      European countries.  In Germany, Heinrich et al. (2000) reported a cross-sectional survey of
31      children, conducted twice (with the same 971 children included in both surveys).  TSP levels

        March 2001                               6-205        DRAFT-DO NOT QUOTE OR CITE

-------
K)
O
o
  TABLE 6-23. LONG-TERM PARTICULATE MATTER EXPOSURE: RESPIRATORY SYMPTOM, LUNG FUNCTION,
                                                           AND BIOMARKER EFFECTS

                                                                                                                         Effect estimates as reported by study
                                                                                                                         authors.  Negative coefficients for lung
                                                                                                                         function and ORs greater than 1 for other
                                                                                                                         endpoints suggest effects of PM
          Reference citation, location, duration, type of
          study, sample size, pollutants measured,
          summary of values
Health outcomes measured, analysis design,
covanates included, analysis problems
Results and Comments
Effects of co-pollutants
          United States
ON
NJ
O
"fl
H
6
O
2
O
H
O
c
o
H
          Abbey etal. (1998)
          California Communities
          20 year exposure to respirable particulates,
          suspended sulfates, ozone, and PMU).
          BerglundetaL (1999)
          California communities
Peters etal. (1999a,b)
12 demographically similar communities in
So. California.
O3, PM acids, and NO2 evaluated.
Gauderman et al. (2000)
12 So. California communities 1993 to 1997
Pollutants:  O3, NO2, PM,,,, and PM25.
PM,0 levels ranged from 16.1 to 67.6
across the communities.
                                           Sex specific multiple linear regressions were
                                           used to relate lung function measures to
                                           various pollutants in long-running cohort study
                                           of Seven Day Adventists (ASHMOG Study).
Cohort study of Seventh Day Adventists.
Multivanate logistic regression analysis of risk
factors (e.g , PM) for chronic airway disease in
elderly non-smokers, using pulmonary function
test and respiratory symptom data.

Stepwise logistic regression was used to relate
prevalence rates for symptoms to community-
specific ambient pollutants after adjustment for
race, sex, asthma, body mass, hay fever, and
membership in an insurance plan.

Studies of lung function growth of 3035
children m 12 communities within 200-mile
radius of Los Angeles during 1993 to 1997.
Cohorts of fourth, seventh, and tenth-graders
studied. By grade cohort, a sequence of linear
regression models were used to determine over
the 4yr of follow-up, if average lung function
growth rate of children was associated with
average pollutant  levels  Adjustment were
made for height, weight, body mass index,
height by age interaction, report of asthma
activity or smoking.  Two-pollutant models
also used.
                                            Sulfates were associated with
                                            decreases in FEV.
                                                                                       Significant risk factors identified:
                                                                                       childhood respiratory illness,
                                                                                       reported ETS exposure, age, sex and
                                                                                       parental history.
Wheeze prevalence was associated
with both acid and NO2.
Lung growth rate for children in
most polluted community, as
compared to least polluted, was
estimated to result in cumulative
reduction of 3.4% in FEV, and
5.0% in MMEF over 4-yr study
period. Estimated deficits mostly
larger for children spending more
time outdoors.  Due to the high
correlation in concentrations across
communities, not able to separate
effects of each pollutant.  No sig.
associations seen with O3.
                                   Frequency of days where PMlo >
                                   100 ,ug/m3 associated with FEV
                                   decrement in males whose parents had
                                   asthma, bronchitis, emphysema,  or hay
                                   fever. No effects seen in other subgroups.

                                   For PM,0 > 100,ug/m3, 42 d/yr:
                                   RR=-1.09CT(0.92, 1.30) for
                                   obstructive disease determined by
                                   pulmonary function tests.
No significant relationships were found
between PM,() and symptoms.
From the lowest to highest observed
concentration of each pollutant, the
predicted differences in annual growth
rates were:  -0.85% for PM,0 (p = 0.026);
-0.64% for PM25 (p =  0 052);  -0.90% for
PM,,,.2 5 (p = 0.030); - 0.77% for NO, (p =
0.019); and -0.73% for inorganic acid
vapor (p = 0.042).
 O
 H
 tfl

-------
O
o
      TABLE 6-23 (cont'd).  LONG-TERM PARTICULATE MATTER EXPOSURE:  RESPIRATORY SYMPTOM, LUNG
                                                   FUNCTION, AND BIOMARKER EFFECTS

                                                                                                                         Effect estimates as reported by study
                                                                                                                         authors.  Negative coefficients for lung
                                                                                                                         function and ORs greater than 1 for other
                                                                                                                         endpoints suggest effects of PM
          Reference citation, location, duration, type of
          study, sample size, pollutants measured,
          summary of values
Health outcomes measured, analysis design,
covanates included, analysis problems
Results and Comments
Effects of co-pollutants
          United States (cont'd)
K)
O
 •m
 H
 6
 o
 z;
 o
 H
O
 c
 o
 H
 W
 O
 &
 o
 k—H
 H
 CD
          McConnell et al. (1999)
          12 Southern California communities
          1994 air monitoring data.
          PMHI (mean 34.8; range 13.0 - 70.7 Mg/m3).
          PM25 (yearly mean 2 week averaged mean
          15.3 Mg/ni3; range 6.7 - 31.5 yUg/m1).
          Dockeryetal. (1996)
          24 communities in the U. S. and Canada.
Raizenne et al. (1996)
24 communities in the U.S. and Canada
Pollutants measured for at least one year prior
to lung function tests: PMIO, PM2,, particle
strong acidity, O3, NO2> and SO2.
                                           Cross sectional study of 3,676 school children
                                           whose parents completed questionnaires in
                                           1993 that characterized the children's history
                                           of respiratory illness.  Three groups examined:
                                           (1) history of asthma; (2) wheezing but no
                                           asthma; and (3) no history of asthma or
                                           wheezing.  Logistic regression model used to
                                           analyze PM, O3, NO2, acid vapor effects. This
                                           study also described in Peters et al. (1999b,c).
Respiratory health effects among 13,369 white
children aged 8 to 12 yrs analyzed in relation
to PM indices.  Two-stage logistic regression
model used to adjust for gender, history of
allergies, parental asthma, parental education,
smoking in home.

Cross-sectional study of lung function.  City
specific adjusted means for FEV and FVC
calculated by regressing the natural logarithm
of the measure on sex, In height, and In age.
These adjusted means were then regressed on
the annual pollutant means for each city.
Positive association between air
pollution and bronchitis and phlegm
observed only among children with
asthma. As PMH> increased across
communities, a corresponding
increase in risk of bronchitis per
interquartile range occurred.
Strongest association with phlegm
was for NO2. Because of high
correlation of PM air pollution,
NO2, and acid, not possible to
distinguish clearly which  most
likely responsible for effects.

Although bronchitis endpoint was
significantly related to fine PM
sulfates, no endpoints were related
to PM10 levels.
PM measures (e.g., particle strong
acidity) associated with FEV and
FVC decrement.
                                                                              PM,o
                                                                                Asthma
                                                                                   Bronchitis MCI (1.1 -  1.8)
                                                                                   Phlegm     2.1(1.4-3.3)
                                                                                   Cough     1.1 (10.8- 1.7)
                                                                                No Asthma / No Wheeze
                                                                                   Bronchitis 0.7 (0.4 - 1.0)
                                                                                   Phlegm     0.8(0.6- 1.3)
                                                                                   Cough     0.9(0.7  - 1.2)

-------
p
3
sr
O
O
 TABLE 6-23 (cont'd). LONG-TERM PARTICULATE MATTER EXPOSURE:  RESPIRATORY SYMPTOM, LUNG
	FUNCTION, AND BIOMARKER EFFECTS	

                                                                                                                   Effect estimates as reported by study
                                                                                                                   authors. Negative coefficients for lung
                                                                                                                   function and ORs greater than 1  for other
                                                                                                                   endpoints suggest effects of PM
         Reference citation, location, duration, type of
         study, sample size, pollutants measured,
         summary of values
Health outcomes measured, analysis design,
covanates included, analysis problems
Results and Comments
Effects of co-pollutants
O
CO
o
52
Tl
6
o
z;
o
H
O
o
H
         Europe

         Ackermann-Liebrich et al. (1997)
         Eight Swiss regions
         Pollutants:  SO2, NO2, TSP, O3, and PMI0.
         Braun-Fahrlander et al. (1997)
         10 Swiss communities
         Pollutants:  PM,,,, NO2, SO2, and O3.
         Zempetal.(1999)
         8 study sites in Switzerland.
         Pollutants:  TSP, PM]0, SO2, NO,, and O3.
                                      Long-term effects of air pollution studied in
                                      cross-sectional population-based sample of
                                      adults aged 18 to 60 yrs. Random sample of
                                      2,500 adults in each region drawn  from
                                      registries of local inhabitants. Natural
                                      logarithms of FVC and FEV, regressed against
                                      natural loganthms of height, weight, age,
                                      gender, atopic status, and pollutant variables.

                                      Impacts of long-term air pollution  exposure on
                                      respiratory symptoms and illnesses were
                                      evaluated in cross-sectional study of Swiss
                                      school children, (aged 6 to 15 years).
                                      Symptoms analyzed using a logistic regression
                                      model including covanates of family history
                                      of respiratory and allergic diseases, number of
                                      siblings, parental education, indoor fuels,
                                      passive smoking, and others.

                                      Logistic regression analysis of associations
                                      between prevalences of respiratory symptoms
                                      in random sample of adults and  air pollution.
                                      Regressions adjusted for age, BMI, gender,
                                      parental asthma, education, and foreign
                                      citizenship.
                                           Significant and consistent effects on
                                           FVC and FEV were found for PM,0,
                                           NO2 and SO2.
                                           Respiratory endpoints of chronic
                                           cough, bronchitis, wheeze and
                                           conjunctivitis symptoms were all
                                           related to the various pollutants.
                                           The colineanty of the pollutants
                                           prevented any causal separation.
                                           Chronic cough and chronic phlegm
                                           and breathlessness were related to
                                           TSP, PMH) and NO2.
                                  Estimated regression coefficient for PM1(>
                                  versus FVC = -0.035 (95% CI -0.041,
                                  - 0.028). Corresponding value for FEV,
                                  -0.016 (95% CI -0.023 to -0.01).  Thus,
                                  10 Aig/m3 PMH) increase estimated to lead
                                  to estimated 3.4 percent decrease in FVC
                                  and 1.6 percent decrease in FEV,.
                                  PM10
                                  Chronic cough OR 11.4 (2.8, 45.5)
                                  Bronchitis OR 23.2 (2.8, 45.5)
                                  Wheeze OR 1.41 (0.55,3.58)
                                  Chronic cough, chronic phlegm and
                                  breathlessness were related to PMI(), and
                                  TSP.
n
H
m

-------
I
o
o
               TABLE 6-23 (cont'd).  LONG-TERM PARTICULATE MATTER EXPOSURE:  RESPIRATORY SYMPTOM, LUNG
              	FUNCTION, AND BIOMARKER EFFECTS
                                                                                                                                  Effect estimates as reported by study
                                                                                                                                  authors. Negative coefficients for lung
                                                                                                                                  function and ORs greater than 1  for other
                                                                                                                                  endpomts suggest effects of PM
Reference citation, location, duration, type of
study, sample size, pollutants measured,
summary of values
                                                    Health outcomes measured, analysis design,
                                                    covanates included, analysis problems
Results and Comments
Effects of co-pollutants
T1
H
6
o
2
o
H
/O
c
o
H
W
O
&
O
t—t
H
          Europe (cont'd)

          Heinnchetal. (1999)
          Bitterfeld, Zerbstand Hettstedt areas of former
          East Germany,
          During Sept. 1992 to July 1993 TSP ranged
          from 44 to 65 ,ug/m3;
          PM10 measured October 1993 - March 1994
          ranged from 33 to 40; and BS ranged from 26
          to 42 ^g/m3
          Hemrich et al. (2000)
          Three areas of former E. Germany
          Pollution measures: SO2, TSP, and some
          limited PMHI data.  TSP decreased from 65,
          48, and 44 ,ug/m3 to 43, 39, and 36 ^g/m3 in
          the three areas
Kramer etal. (1999)
Six East and West Germany communities
(Leipzig, Halle, Maddeburg, Altmark,
Duisburg, Borken)
Between 1991 and 1995 TSP levels in six
communities ranged from 46 to 102 jUg/m3.
Each East Germany community had decrease
in TSP between 1991 and 1995.
                                           Parents of 2470 school children ( 5-14 yr)
                                           completed respiratory health questionnaire.
                                           Children excluded from analysis if had lived <
                                           2 years in their current home, yielding an
                                           analysis group of 2,335 children.  Outcomes
                                           studied:  physician diagnosis for asthma,
                                           bronchitis, symptom, bronchial reactivity, skin
                                           prick test, specific IgE.  Multiple  logistic
                                           regression analyses examined regional effects.
                                                    Cross-sectional study of children (5-14 yr).
                                                    Survey conducted twice, in 1992-1993 and
                                                    1995-1996; 2335 children surveyed in first
                                                    round, and 2536 in second round. Only 971
                                                    children appeared in both surveys.  The
                                                    frequency of bronchitis, otitus media, frequent
                                                    colds, febrile infections studied.  Because
                                                    changes measured over time in same areas,
                                                    covanate adjustments not necessary.

                                                    The study assessed relationship between TSP
                                                    and airway disease and allergies by parental
                                                    questionnaires in yearly surveys of children (5-
                                                    8 yr) between February and May. The
                                                    questions included pneumonia, bronchitis ever
                                                    diagnosed by physician, number of colds,
                                                    frequent cough, allergic symptoms.
                                                    In all, 19,090 children participated.  Average
                                                    response was 87%. Analyses were conducted
                                                    on 14,144 children for whom information on
                                                    all covanates were available. Variables
                                                    included gender; parent education, heating
                                                    fuel, ETS. Logistic regression used,
                                                    transformed into OR.
Controlling for medical, socio-
demographic, and indoor factors,
children in more polluted area had
circa 50% increase for bronchitic
symptoms and physician-diagnosed
allergies compared to control area
and circa twice the respiratory
symptoms (wheeze, shortness of
breath and cough).  Pulmonary
function tests suggested slightly
increased airway reactivity to cold
for children in polluted area.

PM and SO2 levels both decreased
in the same areas; so results are
confounded.
TSP and SO2 simultaneously
included in the model.  Bronchitis
ever diagnosed showed a significant
association. A decrease in raw
percentage was seen between the
start of the study and the end for
bronchitis. Bronchitis seemed to be
associated only with TSP in spite of
huge differences in mean SO2
levels.
                                                                                                                                  No single pollutant could be separated out
                                                                                                                                  as being responsible for poor respiratory
                                                                                                                                  health.
                                                                                                                         The prevalence of all respiratory
                                                                                                                         symptoms decreased significantly in all
                                                                                                                         three areas over time.
Bronchitis ever diagnosed
TSP per 50 //g/m3
  OR1.63C1(1.37- 1.93)
  Halle (East)           %
      TSP Mg/m3    Bronchitis
1991 102            60.5
1992  73            54.7
1993  62            49.6
1994  52            50.4
1995  46            51.9

-------
g.
O
O
TABLE 6-23 (cont'd).  LONG-TERM PARTICULATE MATTER EXPOSURE:  RESPIRATORY SYMPTOM, LUNG
                                             FUNCTION, AND BIOMARKER EFFECTS

                                                                                                                   Effect estimates as reported by study
                                                                                                                   authors.  Negative coefficients for lung
                                                                                                                   function and ORs greater than 1 for other
                                                                                                                   endpoints suggest effects of PM
         Reference citation, location, duration, type of
         study, sample size, pollutants measured,
         summary of values
Health outcomes measured, analysis design,
covanates included, analysis problems
Results and Comments
Effects of co-pollutants
         Europe (cont'd)
 \*/

 I
 H
 O
 O
 2!
 O
 H
O
         Baldietal. (1999)
         24 areas of seven French towns 1974-1976
         Pollutants: TSP, BS, and SO2> NO4
         3-year average TSP-mean annual values
         ranging 45-243 p
-------
o
tr
NJ
O
O
      TABLE 6-23 (cont'd).  LONG-TERM PARTICULATE MATTER EXPOSURE:  RESPIRATORY SYMPTOM, LUNG
                                                  FUNCTION, AND BIOMARKER EFFECTS
         Reference citation, location, duration, type of
         study, sample size, pollutants measured,
         summary of values
                                          Health outcomes measured, analysis design,
                                          covanates included, analysis problems
                                           Results and Comments
                                           Effects of co-pollutants
                                  Effect estimates as reported by study
                                  authors. Negative coefficients for lung
                                  function and ORs greater than 1 for other
                                  endpomts suggest effects of PM
ON
O
o
2
o
H
O
H
o
HH
H
         Europe (cont'd)

         Turnovska and Kostiranev (1999)
         Dimitrovgrad, Bulgaria, May 1996
         Total suspended paniculate matter (TSPM)
         mean levels were 520 ±161 jug/m3 in 1986
         and 187 ± 9 ^ug/m3 in 1996. S02, H2S, and
         NO2 also measured.
Jedrychowski et al. (1999)
In Krakow, Poland in 1995 and 1997
Spacial distributions for BS and SO2 derived
from network of 17 air monitoring stations.
BS 52.6 Mg/m ± 53.98 in high area and
33.23 ±35.99 in low area.
Jedrychowski and Flak (1998)
In Kracow Poland, in 1991-1995
Daily 24 h concentration of SPM (black
smoke) measured at 17 air monitoring
stations.
High areas had 52.6 //g/m3 mean compared to
low areas at 33.2
Respiratory function of 97 schoolchildren
(mean age 10.4 ± 0.6 yr) measured in May
1996 as a sample of 12% of all four-graders in
Dimitrovgrad. The obtained results were
compared with reference values for Bulgarian
children aged 7 to 14 yr, calculated in the same
laboratory in 1986 and published (Gerginova
etal., 1989; Kostiranevetal., 1994).  Variation
analysis technique were used to treat the data.

Effects on lung function growth studied in
preadolescent children.  Lung function growth
rate measured by gain in FVC and FEV, and
occurrence of slow lung function growth
(SLFG) over the 2 yr period defined as lowest
quintile of the distribution of a given test in
gender group. 1129 children age 9 participated
in first  year and  1001 in follow-up 2 years
later. ATS standard questionnaire and PFT
methods used. Initially univariate descriptive
statistics of pulmonary function indices and
SLFG were established, followed by
multivariate linear regression analyses
including gender, ETS, parental education,
home heating system and mold. SO2 also
analyzed.

Respiratory health survey of 1,129 school
children (aged 9 yr).  Respiratory outcomes
included chronic cough, chronic phlegm,
wheezing, difficulty breathing and asthma.
Multi-variable logistic regression used to
calculate prevalence OR for symptoms
adjusted for potential confounding.
                                                                                     Vital capacity and FEV, were
                                                                                     significantly lower (mean value. =
                                                                                     88.54% and 82.5% respectfully)
                                                                                     comparing values between 1986 and
                                                                                     1996. TSPM pollution had
                                                                                     decreased by 2.74 times to levels
                                                                                     still higher than Bulgarian and
                                                                                     WHO standards.
Statistically significant negative
association between air pollution
level and lung function growth
(FVC and FEV,) over the follow up
in both gender groups. SLFG was
significantly higher in the more
polluted areas only among boys.
In girls there was consistency in the
direction of the effect, but not stat.
significant.  Could not separate BS
and SO2 effects on lung function
growth. Excluding asthma subjects
subsample (size 917) provided
similar results.
The comparison of adjusted effect
estimates revealed chronic phlegm
as unique symptom related neither
to allergy nor to indoor variable but
was associated significantly with
outdoor air pollution category
(APL). No potential confounding
variable had major effect.
Boys
  SLFG (FVC)
     OR = 2.15(CI 1.25 -3.69)
  SLFG (FEV,)
     OR=1.90(CI 1.12-3.25)

Girls
  FVC OR = 1.50 (CI 0.84 - 2.68)
  FEV1 OR=1.39(CI0.78- 2.44)
It was not possible to assess separately the
contribution of the different sources of air
pollutants to the occurrence of respiratory
symptoms.  ETS and household heating
(coal vs. gas vs. central heating) appeared
to be of minimal importance.

-------
M
g.
N>
O
o
     TABLE 6-23 (cont'd).  LONG-TERM PARTICULATE MATTER EXPOSURE:  RESPIRATORY SYMPTOM, LUNG
                                                  FUNCTION, AND BIOMARKER EFFECTS
                                                                                                                       Effect estimates as reported by study
                                                                                                                       authors.  Negative coefficients for lung
                                                                                                                       function and ORs greater than 1 for other
                                                                                                                       endpomts suggest effects of PM
Reference citation, location, duration, type of
study, sample size, pollutants measured,
summary of values
                                                    Health outcomes measured, analysis design,
                                                    covariates included, analysis problems
Results and Comments
Effects of co-pollutants
Tl
H
6
o
z
o
H
/O
G
O
H
          Latin America

          Calderon-Garciduenas et al. (2000)
          Southwest Metropolitan Mexico City
          (SWMMC) winter of 1997 and summer of
          1998.
          Australia

          Lewis etal. (1998)
          Summary measures of PM]0 and SO2
          estimated for each of 10 areas in steel cities of
          New South Wales.
Asia

Wong etal. (1999)
Hong Kong, 1989 to 1991
Sulfate concentrations in respirable particles
fell by 38% after implementing legislation
reducing fuel sulfur levels.
                                          Study of 59 SWMMC children to evaluate
                                          relationship between exposure to ambient
                                          pollutants (O3 and PM1()) and chest x-ray
                                          abnormalities. Fishers exact test used to
                                          determine significance in a 2x2 task between
                                          hyperinflation and exposure to SWMMC
                                          pollutant atmosphere and to control, low-
                                          pollutant city atmosphere.
                                          Cross-sectional survey of children's health and
                                          home environment between Oct 1993 and Dec
                                          1993 evaluated frequency of respiratory
                                          symptoms (night cough, chest colds, wheeze,
                                          and diagnosed asthma). Covariates included
                                          parental education and smoking, unflued gas
                                          heating, indoor cats, age, sex, and maternal
                                          allergy.  Logistic regression analysis used
                                          allowing for clustering by GEE methods.
                                          3405 nonsmoking, women (mean age 36.5 yr;
                                          SD ± 3.0) in a polluted district and a less
                                          polluted district were studied for six
                                          respiratory symptoms via self-completed
                                          questionnaires. Binary latent variable
                                          modeling used.
                                                                                     Bilateral symmetric mild lung
                                                                                     hyperinflation was significantly
                                                                                     associated with exposure to the
                                                                                     SWMMC air pollution mixture
                                                                                     (p>0.0004). This raises concern for
                                                                                     development of chronic disease
                                                                                     outcome in developing lungs.
                                                                                     SO2 was not related to differences in
                                                                                     symptom rates, but adult indoor
                                                                                     smoking was.
                                  Night cough OR 1.34(1.18, 1.53)
                                  Chest colds OR 1.43 (1.12, 1.82)
                                  Wheeze OR 1.13 (0.93, 1.38)
Comparison was by district; no PM
measurements reported. Results
suggest control regulation may have
had some (but not statistically
significant) impact.
 o
 s

-------
O
o
     TABLE 6-23 (cont'd).  LONG-TERM PARTICULATE MATTER EXPOSURE:  RESPIRATORY SYMPTOM, LUNG
     ===^======^==	     FUNCTION, AND BIOMARKER EFFECTS

                                                                                                                       Effect estimates as reported by study
                                                                                                                       authors. Negative coefficients for lung
                                                                                                                       function and ORs greater than 1 for other
                                                                                                                       endpomts suggest effects of PM
          Reference citation, location, duration, type of
          study, sample size, pollutants measured,
          summary of values
Health outcomes measured, analysis design,
covanates included, analysis problems
Results and Comments
Effects of co-pollutants
ON
 H
 6
 o
 2!
 O
 H
O
 G
 O
 H
 tfl
 O
 *J
 O
 H—(
 H
 M
              (cont'd)

          Wang etal. (1999)
          Kaohsiung and Panting, Taiwan
          October 1995 to June 1996
          TSP measured at 11 stations, PMM, at 16
          stations. PM10 annual mean ranged from 19.4
          to 112.81  ,ug/m3 (median = 91.00 ^g/m3)
          TSP ranged from 112.81 to 237.82 ,ug/m3
          (median = 181.00).  CO, NO2, SO2,
          hydrocarbons and O3 also measured.
          Quo etal. (1999)
          Taiwan, October 1955 and May 1996
          PM,0 measured by beta-gauge.
          Also monitoring for SO2, NO2> O3, CO.
Wang etal. (1999)
Chongqumg, China
April to July 1995
Dichot samplers used to measure PM2 s.
Mean PM2 5 level high in both urban
(143 /ug/m3) and suburban (139 //g/m3) area.
SO2 also measured
Relationship between asthma and air pollution
examined in cross-sectional study among
165,173 high school students (11- 16 yr).
Evaluated wheeze, cough and asthma
diagnosed by doctor. Video determined if
student displayed signs of asthma. Only
155,283 students met all requirements for
study analyses and, of these, 117,080 were
covered by air monitoring stations. Multiple
logistic regression analysis used to determine
independent effects of risk factors for asthma
after adjusting for age, gender, ETS, parents
education, area resident, and home incense use.

Study of asthma prevalence and air pollutants.
Survey for respiratory disease and symptoms in
middle-school students age < 13 to > 15 yr.
Total of 1,018,031 (89.3%) students and their
parents responded satisfactorily to the
questionnaire.  Schools located with 2 km of
55 monitoring sites. Logistic regression
analysis conducted, controlling for age,  hx
eczema, parents education.

Study examined relationship between PFT and
air pollution. Pulmonary function testing
performed on 1,075 adults (35 - 60 yr) who
had never smoked and did not use coal stoves
for cooking. Generalized additive model used
to estimate difference, between two areas for
FEV,, FVC, and FEV,/FVC% with adjustment
for confounding factors (gender; age, height,
education, passive smoking, and occupational
exposures).
                                                                                     Asthma significantly related to high
                                                                                     levels of TSP, NO2, CO, O3 and
                                                                                     airborne dust. However PMHI and
                                                                                     SO2 not associated with asthma.
                                                                                     The lifetime prevalence of asthma
                                                                                     was 18.5% and the 1-year
                                                                                     prevalence was 12.5%.
Because of close correlation among
air pollutants, not possible to
separate effects of individual ones.
Factor analysis used to group into
two classes (traffic-related and
stationary fossil fuel-related). No
association found between lifetime
asthma prevalence and nontraffic
related air pollutants (SO2, PM10).

Mean SO2 concentration in the
urban and suburban area highly
statistically significant different
(213 and 103 ^g/m3 respectfully).
PM2 5 difference was small, while
levels high in both areas.  Estimated
effects on FEV1 statistically
different between the two areas.
                                  Adjusted OR

                                  PM,0
                                    1.00(0.96-1.05)

                                  TSP
                                    1.29(1.24-1.34)
Difference between urban and suburban
area excluding occupational exposures:
FEV,
B - 119.79
SE 28.17
t - 4.25
p<0.01
FVC
B - 57.89
SE 30.80
t- 1.88
p < 0.05

-------
2
fa
K)
O
O
TABLE 6-23 (cont'd). LONG-TERM PARTICULATE MATTER EXPOSURE:  RESPIRATORY SYMPTOM, LUNG
                                             FUNCTION, AND BIOMARKER EFFECTS

                                                                                                                  Effect estimates as reported by study
                                                                                                                  authors.  Negative coefficients for lung
                                                                                                                  function and ORs greater than 1 for other
                                                                                                                  endpomts suggest effects of PM
          Reference citation, location, duration, type of
          study, sample size, pollutants measured,
          summary of values
Health outcomes measured, analysis design,
covariates included, analysis problems
                                                                                              Results and Comments
                                                                                              Effects of co-pollutants
K>
£
 Tl
 H
 6
 o
 2
 O
 H
O
 c
 o
          Asia (cont'd)

          Zhang etal. (1999)
          4 areas of 3 Chinese Cities (1985 - 1988)
          TSP levels  ranged from an annual arithmetic
          mean 137 /ug/m3 to 1250 ,ug/m3 using
          gravimetric methods.
          Qianetal. (2000)
          4 China cities
          The 4 year average PM means were 191, 296,
          406, and 1067 Mg/m3.  SO2 and NO2
          measurements were also available.
                                     A pilot study of 4 districts of 3 Chinese cities
                                     in for the years 1985-1988, TSP levels and
                                     respiratory health outcomes studied.  4,108
                                     adults (< 49 yrs)  examined by questionnaires
                                     for couth, phlegm, wheeze, asthma, and
                                     bronchitis. Categorical logistic—regression
                                     model used to calculate odds ratio. SO2 and
                                     NO2 were also examined. Other potential
                                     confounding factors (age, education level,
                                     indoor ventilation, and occupation) examined
                                     in the multiple logistic regression model.

                                     Pilot cross-sectional survey of 2789 elementary
                                     school children in four Chinese communities
                                     chosen for their PM gradient.  Frequency of
                                     respiratory symptoms (cough, phlegm, wheeze,
                                     and diagnosed asthma, bronchitis, or
                                     pneumonia) assessed by questionnaire.
                                     Covariates included parental occupation,
                                     education and smoking.  The analysis used
                                     logistic regression, controlling for age, sex,
                                     parental smoking, use of coal in home, and
                                     home ventilation.
                                           Results suggested that the OR's for
                                           cough, phlegm, persistent cough
                                           and phlegm and wheeze increased
                                           as outdoor TSP concentrations did. .
                                                                                                                                 Wheeze produced largest OR for both
                                                                                                                                 mothers and fathers in all locations.
                                           Results not directly related to
                                           pollution levels, but symptom rates
                                           were highest in highest pollution
                                           area for cough, phlegm,
                                           hospitalization for respiratory
                                           disease, bronchitis, and pneumonia.
                                           No gradient correlating with
                                           pollution levels found for the three
                                           lower exposure communities.
o
>— i
H
m

-------
  1      decreased between surveys as did the prevalence of all respiratory symptoms (including
  2      bronchitis). Also, Kramer et al. (1999) reported a study in six East and West Germany
  3      communities, which found yearly decreasing TSP levels to be related to ever-diagnosed
  4      bronchitis from 1991-1995. Lastly, Jedrychowski et al. (1999) reported an association between
  5      both BS and SO2 levels in various areas of Krakow, Poland, and slowed lung function growth
  6      (FVC and FEV,).
  7
  8      6.3.3.2.3 Summary of Long-Term Particulate Matter Exposure Respiratory Effects
  9           The methodology used in the long-term studies varies much more than the methodology in
 10      the short-term studies.  Some studies reported highly significant results (related to PM) while
 11      others reported no significant results. The cross-sectional studies are often confounded, in part,
 12      by unexplained differences between geographic regions. The studies that looked for a time trend
 13      are also confounded by other conditions that were changing over time.  Probably the most
 14      credible cross-sectional study remains that described by Dockery et al. (1996) and Raizenne et al.
 15      (1996). This study, reported in the previous 1996 PM AQCD, found differences in peak flow
 16      and bronchitis rates associated with fine particle strong acidity. Whereas most studies included
 17      only two to six communities, this study included 24 communities. The effective sample size for
 18      a cross sectional analysis is the number of communities, so that six or fewer communities allow
 19      many fewer degrees of freedom by which to test hypotheses about various  pollutant effects.
 20          Newly available studies since the 1996 PM AQCD, overall, provide evidence consistent
 21      with the findings from the above 24-City Study. Most notably several U.S. and European studies
 22      report associations between PM measures and bronchitis rates and/or lung  function decrements
 23      or slowed lung function growth.  One also provided evidence of PM effects on immune function
 24      in school children, with stronger associations for fine particle indicators than for ambient coarse
25      particles.
26
27
       March 2001                              6-215        DRAFT-DO NOT QUOTE OR CITE

-------
 1     6.4  INTERPRETIVE ASSESSMENT OF EPIDEMIOLOGIC DATABASE
 2          ON HEALTH EFFECTS OF AMBIENT PARTICIPATE MATTER
 3     6.4.1  Introduction
 4          As noted at the outset of this chapter, numerous PM epidemiology studies assessed in the
 5     1996 PM AQCD implicated ambient PM as a likely key contributor to mortality and morbidity
 6     effects observed epidemiologically to be associated with ambient air pollution exposures. Since
 7     preparation of the last previous PM AQCD in 1996, the epidemiologic evidence concerning
 8     ambient PM-related health effects has expanded greatly.  The most important types of additions
 9     to the database beyond that assessed in the  1996 PM AQCD, as evaluated above are:
10     • New multi-city studies on a variety of endpoints providing more precise effects estimates than
11       most smaller-scale individual city studies;
12     • More studies of various health endpoints using ambient PM10 and/or closely related mass
13       concentration indices (e.g., PM13 and PM7), which substantially lessen the need to rely on
14       non-gravimetric indices (e.g., BS or COH);
15     • New studies on a variety  of endpoints for which information on the ambient PM coarse fraction
16       (PM(10_2 5)), the ambient fine-particle fraction (PM2 5), and even ambient ultrafine particle mass
17       measures (PM0 [ and smaller) were observed and/or estimated from site-specific calibrations;
18     • A few new studies in which the relationship of some health endpoints to ambient particle
19       number concentrations were evaluated;
20     • Many new studies which evaluated the sensitivity of estimated PM effects to the inclusion of
21       gaseous co-pollutants in the model;
22     • Preliminary attempts to evaluate the effects of air pollutant combinations or mixtures including
23       PM components, based on empirical combinations (e.g., factor analysis) or source profiles;
24     • Numerous new studies of cardiovascular endpoints, with particular emphasis on assessment of
25       cardiovascular risk factors as well as symptoms;
26     • Additional new studies on asthma and other respiratory conditions potentially exacerbated by
27       PM exposure;
28     • New studies of infants  and children as a potentially susceptible population.
29     The vast majority of the  new PM epidemiology studies, both of short-term and long-term PM
30     exposure, continue to show statistically significant excess mortality risk and/or morbidity

       March 2001                             6-216        DRAFT-DO NOT QUOTE OR CITE

-------
  1      endpoints to be associated with ambient PM indexed by a variety of ambient community
  2      monitoring methods in many U.S. cities and elsewhere.
  3           Several methodological issues, discussed in the 1996 PM AQCD, are still of much
  4      importance in assessing and interpreting the overall PM epidemiology database and its
  5      implications for estimating risks associated with exposure to ambient PM concentrations in the
  6      United States. The fundamental issue essentially subsuming all of the other modeling issues is
  7      the selection of an appropriate statistical model in the absence of any strong prior hypotheses or
  8      information about underlying relationships between health outcomes and ambient PM, other
  9      copollutants, or other important factors such as seasonal variations and/or demographic
 10      characteristics of study populations.  These critical methodological issues are:  (1) potential
 11      confounding of PM effects by co-pollutants (especially major gaseous pollutants such as O3, CO,
 12      NO2, SO2); (2) the attribution of PM effects to specific PM components (e.g., PM,0, PM10_25,
 13      PM2 5, ultrafmes, sulfates, metals, etc.) or source-oriented indicators (motor vehicle emissions,
 14      vegetative burning, etc.); (3) the temporal relationship between exposure and effect (lags,
 15     /mortality displacement, etc.); (4) the general shape of exposure-response relationship(s) between
 16      PM and/or other pollutants and observed health effects (e.g., potential indications of thresholds
 17      for PM effects); and (5) the consequences of measurement error. In addition, the newer multi-
 18      city study results, e.g, the NMMAPS analysis of the 90 largest U.S. cities (Samet et al., 2000a,b)
 19      show evidence of more geographical heterogeneity in the estimated PM risks across cities and
20      regions than had been seen in the studies assessed in the 1996 PM AQCD. Thus, the issue of
21      geographical heterogeneity in PM effects estimates also warrants further evaluation here.
22           Assessing the above issue(s) in relation to the PM epidemiology data base remains quite a
23      challenge. The basic issue is that there are an extremely large number of possible models, any of
24      which may turn out to give the best statistical "fit" of a given set of data, and only some of which
25      can be dismissed a priori as biologically or physically illogical or impossible, except that
26      putative cause clearly cannot follow effect in time. Most of these models are fitted in a stepwise
27      manner, first by removing effects known almost certainly to be present, including general
28      changes in death rates or hospital admissions rates over long time intervals and across season, by
29      day of week, and attributable to weather and climate.  Many of the temporal and weather variable
30      models have been fitted to data using semi-parametric methods such as spline functions or local
31      regression smoothers (loess).  The goodness of fit of these base models has been evaluated by

        March 2001                               6-217        DRAFT-DO NOT QUOTE OR CITE

-------
 1      criteria suitable for generalized linear models with Poisson or hyper-Poisson responses (number
 2      of events) with a log link function, particularly the Akaike Information Criterion (AIC) and the
 3      more conservative Bayes or Schwarz information criterion (BIC), that adjust for the number of
 4      parameters estimated from the data. The Poisson over-dispersion index and the auto-correlation
 5      of residuals are also often used.  Once a best-fitting baseline model is selected, the specification
 6      of variables in the base model is often held fixed while a better model is selected using one or
 7      more PM indices (e.g., fine and coarse) and/or one or more gaseous co-pollutants. In general,
 8      one would expect that the best-fitting models with PM would be models with the largest and
 9      most significant PM indices. If PM effects are confounded with those of other pollutants, then a
10      large positive estimated PM effect might be associated with a non-biological estimated negative
11      effect for one or more other criteria pollutants, as found by some analyses for NO2 in a joint
12      pollutant model (most likely a statistical artifact). Also, if high correlations between PM and one
13      or more gaseous pollutants emitted from a common source (e.g., motor vehicles) exist in a given
14      area, then disentangling their relative individual partial contributions to observed health effects
15      associations becomes very difficult. Unfortunately, there have been very few attempts at broad,
16      systematic investigations of the model selection issue and little reporting in published reports of
17      goodness-of-fit criteria among competing models that provide a better basis by which to better
18      assess or compare models.
19           Given the now extremely large number of published epidemiologic studies of ambient PM
20      associations with health effects in human populations and the considerably wide diversity in
21      applications of even similar statistical approaches (e.g., "time-series analyses" for short-term PM
22      exposure effects), it is neither feasible nor  useful here to try to evaluate the methodological
23      soundness of every individual study. Rather, two feasible approaches are likely to yield useful
24      evaluative information: (1) an overall characterization of evident general commonalities (or
25      notable marked differences) among findings from across the body of studies dealing with
26      particular PM exposure indices and types of health outcomes; and (2) more thorough, critical
27      assessment of key newly published multi-city analyses of PM effects, given that greater scientific
28      weight is likely ascribable to their results than those of smaller sized studies (often of individual
29      cities) yielding presumably less precise effects estimates.
30
31

        March 2001                               6-218        DRAFT-DO NOT QUOTE OR CITE

-------
 1      6.4.2  New Assessments of Confounding
 2           As discussed previously, the issue of potential confounding by weather was extensively
 3      examined in two studies as reviewed in the 1996 PM AQCD, and was considered essentially
 4      resolved. Potential confounding by co-pollutants, however, was nevertheless still suggested by
 5      several studies reviewed in the 1996 AQCD. Therefore, discussion of confounding in this
 6      section is focused on potential confounding among PM and other major gaseous air pollutants as
 7      evaluated in newly available studies.
 8
 9      6.4.2.1 Assessment of Copollutant Confounding
10           Analyses of one city's data by different researchers may produce conflicting results. For
11      example, Moolgavkar and Luebeck (1996) and Samet et al. (1996) or Kelsall et al. (1997),
12      (which presented essentially the same results) analyzed Philadelphia mortality data for nearly the
13      same period (1973-1988 and 1974-1988, respectively), but produced somewhat different results
14      and interpretations. The notable differences in findings in these studies were: (a) NO2 in the
15      Samet et al. 's study was mostly negatively associated (except summer) with mortality, while in
16      the Moolgavkar-Luebeck study, NO2 was mostly positively associated (except winter); and (b) O3
17      in Samet et al.'s study was positively associated with mortality across seasons (weakest in the
18      summer), while  in the Moolgavkar-Luebeck study, O3 was positively associated with mortality
19      only in the summer. The differences may have been due to the difference in the optimum lags
20      chosen for pollutants (in Samet et al., concurrent day levels were used for all the pollutants
21      except CO; whereas, in the Moolgavkar-Luebeck study, one-day lag was used for all pollutants
22      except NO2).  Moolgavkar-Luebeck concluded that "..it is not possible with the  present evidence
23      to show a convincing correlation between particulate air pollution and mortality", while Samet's
24      group concluded "...These analyses confirm the association between TSP and mortality found in
25      previous studies in Philadelphia and the association is robust to consideration of other
26      pollutants".
27           Such discrepancies could, in part, result from instability of regression coefficients due to
28      collinearity of co-pollutants, as well as model specification choice.  The collinearity problem may
29      be further complicated by different seasonal patterns of concentrations for each  pollutant, which
30      also vary from city to city. Thus, evaluation of apparently inconsistent results from one or a few
31      cities analyzed using different model specifications, without quantitative information on city
        March 2001                               6-219        DRAFT-DO NOT QUOTE OR CITE

-------
 1      specific characteristics, is unlikely to yield useful information by which to resolve the issue of
 2      confounding. By analyzing multiple cities' data, a more consistent pattern may emerge, although
 3      differences in approach may still result in inconsistent multi-city results by different researchers.
 4      Several studies have examined the issue of confounding using multi-city analyses.  Basic
 5      descriptions of these studies were provided in earlier text and in Table 6-1; some of their more
 6      salient results regarding confounding by co-pollutants are discussed below.
 7           Samet and co-workers (2000a,b) reported PM,0 RR estimates for PM10-only and multiple
 8      pollutant models that also included O3 as the only gaseous pollutant or O3  and another gaseous
 9      pollutant, in both 20-cities analysis and 90-cities analysis. The effects of adding gaseous
10      pollutants in the model on PM10 coefficients were similar in these two data sets, in that  adding O3
11      did not change PM10 coefficients, but additional inclusions of another gaseous pollutant reduced
12      PM10 coefficients somewhat. Figure 6-10 shows the posterior probability results for the 90-cities
13      analysis. It can be seen that the PM10 coefficient reduced from about 0.47  to 0.35 when another
14      gaseous pollutant was included in the model besides O3. Importantly, however, the posterior
15      probabilities that the overall effects are greater than 0 remain 1.0 in all these models. It should
16      also be noted that the results shown in the figure are for PM10 at lag 1  day (of the 0-, 1-, and
17      2-day lags examined, the 1-day lag was most significant). The lags for the gaseous pollutants
18      included in these models were also apparently 1-day lags. This choice of the same lags seems
19      reasonable, as the air pollution variables are generally highly correlated with no lag. However,
20      using the most significant lags for gaseous pollutants might have produced somewhat different
21      results. That is, even though air pollution variables may be highly correlated, or not, at 0 lag,
22      various health effects possibly due to different pollutants may occur with different lag times.
23           The HEI Health Committee Review Panel commentary on the NMMAPS analyses stated
24      that an important consideration in assessing the validity of the observed PM10 effects is  whether
25      they are due to PM10 itself or due to another air pollutant that is correlated  with PM10. That is, do
26      effects of other pollutants confound the observed PM10 effect?  The NMMAPS investigators took
27      a commonly used approach to address this issue in the mortality analysis:  does the addition of
28      other air pollutant concentrations to the PM,0 regression models result in any substantial change
29      in the  estimated PM10 effect? If the PM10 effect does not change, the other  pollutants presumably
30      have not confounded the observed PM10 effect. The Panel identified a few issues related to
31      possible confounding effects by co-pollutants, but concluded that the probable impact of any of

        March 2001                              6-220        DRAFT-DO NOT QUOTE OR CITE

-------
                    0.0          0.2           0.4
                  % Change in Mortality per 10 pg/m3 Increase in PM10

       Figure 6-10. Marginal posterior distributions for effect of PM10 on total mortality at lag 1
                   with and without control for other pollutants, for the 90 cities.  The numbers
                   in the upper right legend are the posterior probabilities that the overall
                   effects are greater than 0.
       Source: Samet et al. (2000a,b).
1     these was not considered to be sufficiently large to alter the observed PM10 effect. For example,
2     when the investigators controlled for co-pollutants, they assumed the co-pollutants effect in the
3     model to be linear.
4          Another consideration is the impact of limiting assessment of the possible confounding
5     effect to the relevant season for pollutants that have seasonal patterns. This assessment is
6     complicated in these data because the seasonal effect of ozone, for example, is assumed to be
7     somewhat different across the cities. Finally, a pollutant (e.g., sulfate or acid aerosols) for which
8     only inadequate data are available in the AIRS database and which, therefore, could not be
      March 2001
6-221
DRAFT-DO NOT QUOTE OR CITE

-------
 1      analyzed, might be more clearly delineated as responsible for the effects attributed to PM10,
 2      per se.
 3           Even given the above considerations, the HEI Review Panel nevertheless agreed that, in the
 4      NMMAPS 20 cities analysis of potential copollutant confounding, no convincing evidence was
 5      found for PM10 effects on mortality being changed by addition of either O3, SO2, NO2, or CO to
 6      the models, suggesting that none of the other pollutants is responsible for the observed PM10
 7      effects.  Subsequent analyses by the investigators, that appear to use similar statistical techniques,
 8      controlled for gaseous pollutants in the 90 cities and also did not show a confounding copollutant
 9      effects.
10           In the morbidity analysis, based on assessment of the likelihood of confounding by other
11      pollutants in stage 2 of the modeling for 14 of the NMMAPS cities, there was evidence that the
12      PM10 effect on each diagnosis was not confounded, similar to the finding in the mortality analysis
13      (but differences in the approach make it difficult to assess whether morbidity findings are as
14      robust). While the approach used in the morbidity analysis is novel (comparing the PM10
15      regression coefficient with the regression coefficient between PM10 and the co-pollutants), the
16      question arises as to the adequacy of statistical power for performing these analyses.  Power may
17      be low because the regression is fit to only 14 locations, and in some cases 12 locations, and
18      when the regression coefficients between PM10 and the potentially confounding co-pollutants are
19      similar across cities.
20           The HEI commentary further noted that although NMMAPS focuses on the effects of PM10,
21      examination of the independent effects of other pollutants is also warranted. Effects on daily
22      mortality were found for most of the gaseous pollutants (SO2, CO, NO2 )  in the 20 cities,
23      although these effects were generally diminished when the model controlled for PM10 and other
24      pollutants. In contrast, the PM10 effect did not appear to be affected by other pollutants in this
25      model. An effect of each pollutant except ozone on mortality in the 90 cities is shown in the
26      NMMAPS II Report.  A relatively strong effect appears to be present for each of those gaseous
27      pollutants in 90 cities in analyses that assess the effect of each pollutant alone. Thus, findings on
28      independent effects of the gaseous pollutants based on the 20 cities should be viewed as
29      preliminary until a 90 city analysis specifically controlling for PM10 and other pollutants is
30      available.  Evaluation of independent effects of the gaseous pollutants on hospitalizations would
31      also be useful in follow-up analyses.

        March 2001                                6-222        DRAFT-DO NOT QUOTE OR CITE

-------
  1           Schwartz (2000a), in his analysis of 10 U.S. cities' data (see New Multi-City Studies and
  2      Table 6-1 for basic study description), took an approach that is different from the usual
  3      simultaneous inclusion of co-pollutants in the model to  address confounding. He postulated that,
  4      if the PM10 effect is really due to confounding by another pollutant, one would expect a larger
  5      PM10 effect in cities or seasons where PM10 represents more of that other pollutant (i.e., where
  6      the slope of the confounder to PM,0 is larger). This approach relied on the large variability in the
  7      relationship between PM10 and possibly confounding gaseous pollutants across the 10 cities.
  8      Schwartz first illustrated this idea with a simulated example.  In the analysis of the real 10 cities'
  9      data, the PM,0 coefficients obtained from city-specific analyses were regressed on the regression
 10      coefficients relating the gaseous pollutant to PM10 in each city. If the PM)0 effects were due to
 11      confounding  only, according to the model, then such regression would result in zero intercept.
 12      To accommodate greater differences in the gaseous pollutants mean level between the indoor
 13      heating season and the warm season, the city-specific regressions were conducted by season,
 14      producing 20 city-specific coefficients. The results indicated that the resulting intercepts (i.e.,
 15      PM10 effects after controlling for confounders) were not substantially different from that without
 16      the gaseous pollutants (0.57, 0.90, and 0.69, for confounding by SO2, CO, and O3, respectively,
 17      vs. 0.67 percent excess mortality deaths per 10 /ug/m3 increase in PM10).  While this approach
 18      appears to be reasonable, it is  not certain if the data had  sufficient and relevant signals to reflect
 19      actual difference in exposures to  PM10 vs. gaseous co-pollutants across cities and seasons. For
 20      example, a high SO2 to PM10 slope in winter may not be as relevant to a high SO2 to PM10 slope
 21       in summer, because of the lower  air exchange rate in winter. Such air exchange rate would also
 22     vary from city to city, possibly further blurring the relevant exposure picture. Also, the gaseous
 23     pollutant's slope on PM,0 can  be  influenced by error in both PM]0 and the gaseous pollutants.
 24     While such slopes may be more accurately estimated for spatially more uniform pollutants such
 25      as O3 (and fine component of PM in the summer in northeast), for primary pollutants such as CO
 26      and SO2, local source impacts may have contributed to less precision for their slopes on PM10.
 27           In Schwartz's analysis of 10 U.S. cities noted earlier, the new approach was not used to
 28      examine the changes in the gaseous pollutants' mortality effects coefficients. Such an analysis
29      would have been useful in providing an overall assessment of possible confounding among the
30      air pollutants. However, such an  analysis was conducted in Schwartz's analysis (2000b) of
31      Philadelphia data for 1974-1988.  In this analysis, the same approach to test confounding was

        March 2001                                6-223        DRAFT-DO NOT QUOTE OR CITE

-------
 1      applied for both TSP and SO2. Instead of using the variability in PM]0 to gaseous pollutants
 2      relationships across cities, this Philadelphia data analysis used the changing relationship between
 3      TSP and SO2 over the 15-year period. The mortality data were thus analyzed for warm and cold
 4      seasons of each year, yielding 30 regression coefficients for both TSP and SO2.  Regression
 5      coefficients for TSP on SO2, as well as SO2 on TSP, were also obtained for each period. In the
 6      second stage regression, the 30 TSP mortality coefficients were regressed on the regression
 7      coefficients of SO2 on TSP, and vice versa.  In addition, visual range-derived extinction
 8      coefficient (an indicator of fine particles) was analyzed as a confounder for TSP in the same
 9      manner. The results indicated that the RR for SO2 was substantially reduced (from 1.12 to
10      1.02 per 50 ppb SO2 increase) by controlling for TSP, whereas TSP RR was not reduced, but
11      rather increased, by controlling for SO2 (1.09 to 1.21 per 100 /ug/m3 TSP increase).  However, the
12      TSP RR was reduced (1.09 to 1.01) when the extinction coefficient was included in the model.
13      Therefore, the author concluded that the association between air pollution and daily deaths in
14      Philadelphia was due to fine combustion particles.
15           A very different approach to co-pollutant modeling was used by Schwartz in the NMMAPS
16      Part II morbidity analyses, and in a recently published paper on mortality (Schwartz, 2000a).
17      The method attempts to identify total or partial confounding of a nominal causal pollutant such
18      as PM10 with a co-pollutant or other confounder, based on the relationship between a regression
19      of the health effect on the nominal causal pollutant, as a function of the regression coefficient of
20      the nominal causal pollutant against the  designated co-pollutant. If the relationship has zero
21      intercept, then one might infer that the two pollutants are totally confounded, with no direct
22      effect of the causal pollutant on the health endpoint that is not mediated by the co-pollutant.
23      If the relationship has a non-zero intercept, then the causal pollutant and the co-pollutant are
24      partially confounded, with the causal pollutant having a direct effect as well as an effect mediated
25      by the co-pollutant. A non-zero intercept and no relation to the co-pollutant  slope implies that
26      only a direct health effect exists with the causal pollutant, with no confounding by the
27      co-pollutant.
28           Figures 32 and 33 [not shown here] in NMMAPS II (pp. 40-41) appear to  show the
29      expected outcome described above. The vertical axes  on both of these figures show the risk
30      estimates for cardiovascular disease, COPD, or pneumonia in a single-pollutant model in 12 to
31      14 cities, with  a causal pollutant Z = PMI0.  There is no statistically significant relationship

        March 2001                              6-224       DRAFT-DO NOT QUOTE OR CITE

-------
  1      between the estimated PM10 effect on health, and the slopes between Z = PM10 and X = one of
  2      the covariables temperature, relative humidity, SO2, or O3. Visual examination of these figures
  3      suggests that the high-risk estimates for pneumonia vs. the SO2 or O3 slopes in Figure 33 occur in
  4      Colorado Springs, a relatively small city with very little correlation between co-pollutants and
  5      PM. Similarly, the high-risk estimates for COPD vs. the PM-O3 slope in Figure 33 occur in
  6      Boulder, another relatively small Colorado city with very little correlation between co-pollutants
  7      and PM. Absent these three points, there is no relationship between the estimated PM]0 and the
  8      regression slopes between PM10 and one of temperature, relative humidity, SO2 or O3, and
  9      certainly not a linear relationship, which implies only a direct relationship with PM10.
 10           The analogous mortality study (Schwartz, 2000a) does not provide as much detail as the
 11      morbidity study in NMMAPS II, 2000.  Schwartz (p. 566) notes:  "For all three cooccurring
 12      pollutants, the effect size after controlling for confounding was not substantially (or statistically
 13      significantly) different from the baseline result. This is illustrated in Figure 3." Figure 4 [not
 14      shown here] plots each of eighteen city-season pairs, showing little or no relationship. The  lack
 15      of a relationship does not, however, necessarily confirm that there is no  confounding. A more
 16      complete implementation of this intriguing approach would be of interest. Until that time,
 17      however, the potential effects of confounding should be examined by several different
 18      approaches,  included the estimated correlation matrix among all of the estimated regression
 19      coefficients.
 20           In summary, the above results from several multi-city studies using different approaches
 21      suggest that possible confounding influences of gaseous pollutants on PM indices are not
 22      substantial.  However, the interpretation of the relative impact of the gaseous co-pollutants as
 23      putative causative agents requires caution and warrants further, more detailed evaluation.
 24
 25      6.4.2.2  Simulation Analysis of Confounding
 26           Since no single model specification can a priori be designated as "correct" in addressing
 27      confounding effects of co-pollutants, discrepancies in results among studies, even for the same
28      dataset, are expected. While any assessment of relative "adequacy" of these alternative model
29      specifications is difficult with observational data, the implication of "inadequate" model
30      specifications may be studied through simulations using synthetic data in which the "correct"
31      model is known.  Chen et al. (1999) conducted such simulations using a synthetic  data set in

        March 2001                               6-225       DRAFT-DO NOT QUOTE OR  CITE

-------
 1      which the causal variables are known, and the effects of model misspecification were studied in
 2      the presence of two variables (x, and x2), with varying level of correlation, in a Poisson model.
 3      They considered three situations: (1) model under/it, in which mortality was generated with both
 4      x, and x2, but regressed only on x,; (2) model over/it, in which mortality was generated with only
 5      x,, but regressed on both x, and x2; (3) model misfit, in which mortality was generated with either
 6      x, or x2 but regressed on the other variable. They observed that the confounding of covariates in
 7      an overfitted model does not bias the estimated coefficients but does reduce their significance;
 8      and that the effect of model underfit or misfit leads to not only erroneous estimated coefficients
 9      but also to erroneous significance.  Chen et al., based on these observations, suggested that
10      "models which use only one or two air quality variables (such as PM10 and SO2) are probably
11      unreliable, and that models containing several correlated and toxic or potentially toxic air quality
12      variables should also be investigated...". While conceptually useful, this simulation study
13      ignored one factor that is crucial in evaluating the implication of confounding, the relative error.
14      For example, including several correlated pollutants in a regression model may lead to erroneous
15      inferences unless one considers the relative error associated with each of the pollutants. Several
16      simulation studies that considered such relative errors are discussed below.
17
18      6.4.2.3 Alternative Approaches to Deal with Confounding
19           In time-series analyses of the acute effects of PM, the usual approach to deal with gaseous
20      co-pollutants is to treat them as confounders and to simply include them simultaneously in
21      regression models. There has even been a suggestion (as mentioned above) based on a
22      simulation analysis of synthetic data, that "several" correlated pollutants should be included in
23      regression models (Chen et al., 1999). This prevailing approach can not only lead to misleading
24      conclusions in  "identifying" a specific "causal" pollutant (e.g., when pollutants have a varying
25      extent of exposure error), but also ignores the potential combined effects of PM and gaseous
26      co-pollutants (e.g., when PM absorbs SO2 and carries it deeper in the airways, as shown by
27      Amdur and Chen, 1989).
28           Another potential problem of the simultaneous inclusion of PM and gaseous pollutants is
29      that the gaseous pollutant in question may be coming from the same source, or that the PM
30      constituent may be derived from the gaseous pollutants. For example, SO2 can be converted to
31      sulfate, which is a PM constituent. Since a confounder cannot be an intermediate step in the

        March 2001                               6-226       DRAFT-DO NOT QUOTE OR CITE

-------
  1     causal pathway (Rothman and Greenland, 1998), strictly speaking, SO2 does not qualify as a
  2     confounder of PM, except in a situation where the PM is known to be solely of secondary origin
  3     (transported aerosols) and SO2 is solely from local origin.  Furthermore, any reduction in
  4     emission of a gaseous pollutant may also affect the level of PM. In such a case, the inference
  5     drawn from the results of simultaneous regression may be misleading, because the relative risk
  6     for PM is based on the assumption that the covariates could be kept unchanged while the PM
  7     level changes.
  8          Alternative approaches are needed to address the above noted weakness in the general
  9     practice of effect estimation using simple simultaneous regressions.  Several alternative
 10     approaches have been tried in recent years to estimate the effects of air pollution.  The studies
 11     include:  (1) Ozkaynak et al.'s (1996) analysis of Toronto,  CN data; (2) Laden et al.'s (2000)
 12     analysis of the Harvard Six Cities PM2 5 data; (3) Mar et al.'s (2000) analysis of Phoenix, AZ
 13     PM25data;  and, (4) Tsai et al.'s (2000) analysis of 3 New Jersey cities (Newark, Camden, and
 14     Elizabeth) data. These studies, as previously described in this chapter, utilized factor analysis to
 15     identify underlying factors that could be characterized in terms of source types.  This approach
 16     thus greatly lessens or prevents inclusion of correlated individual variables in the regression
 17     model  (depending on the rotation approach used) and also allows source-oriented evaluation of
 18     health  impacts of PM components (as discussed more specifically below in Section 6.4.3).
 19          Factor analysis has been routinely used in the air pollution source apportionment field, but
 20     its application to evaluation of PM health effects is relatively new. It may be a useful alternative
 21     approach for a source-oriented evaluation of the combined  effects of fine particles and gaseous
 22     co-pollutants. The advantages of the factor analysis approach include:  (a) it allows an
 23     examination of association between a health outcome and a group(s) of pollutants that vary
 24     together (due to the same source type); (b) it reduces multi-collinearity in a regression model; and
 25     (c) it may reduce error associated with individual variables. On the other hand, factor analyses
26     can also be  vulnerable to several problems: (a) the factors are sometimes quite sensitive to the
27     inclusion or exclusion of variables from the initial correlation matrix; (b) the minor factors are
28      very sensitive to the number of factors considered in the analysis; and (c) the inclusion of
29      variables  with other sources of variation (measurement error, other artifacts, or physical
30      properties) can have a major impact on the selection of factors.  With respect to the latter
31      problem, there are valid concerns  about studies that include both numerous PM elements

        March  2001                              6-227        DRAFT-DO NOT QUOTE OR CITE

-------
 1      determined by XRF and gaseous pollutants (CO, NO2, O3) in the initial correlation matrix. There
 2      are also additional issues in assessing results from factor analysis studies, including the
 3      "interpretability" of resulting factors and technical issues (such as approach used for rotation of
 4      factors). Thus, there are still some issues that need to be further investigated.
 5
 6      6.4.3  Role of Participate Matter Components
 7           In the 1996 PM AQCD, extensive epidemiologic evidence substantiated very well positive
 8      associations between ambient PM10 concentrations and various health indicators, e.g., mortality,
 9      hospital admissions, respiratory symptoms, pulmonary function decrements, etc..  A somewhat
10      more limited number of studies were then available which substantiated mortality and morbidity
11      associations with various fine particle indicators (e.g., PM2 5, sulfate, H+, etc.); and only one, the
12      Harvard Six Cities analysis by Schwartz et al. (1996a), evaluated relative contributions of the
13      fine PM2 5 versus coarse (PM10.2 5) fraction of PM10, with PM2 5 appearing to be associated more
14      strongly with mortality effects than PM]0.2 5.  Lastly, only a very few studies seemed to be
15      indicative of possible coarse particle effects, e.g., increased asthma risks associated with quite
16      high PM10 concentrations in a few locations where coarse particles strongly dominated the
17      ambient PM10 mix.
18
19      6.4.3.1 Fine-and Coarse-Particle Effects on Mortality
20           A greatly enlarged and still rapidly growing number of new studies published since the
21      1996 PM AQCD provide much new evidence further substantiating ambient PM associations
22      with increased human mortality and morbidity. As indicated in Table 6-1, most newly reported
23      analyses, with few exceptions, continue to show statistically significant associations between
24      short-term (24-h) PM concentrations and increases in daily mortality in many U.S. and Canadian
25      cities, as well as elsewhere.  Also, the reanalyses of Harvard Six City and ACS  study data
26      substantiate the original investigator's findings of long-term PM exposure associations  with
27      increased mortality as well.
28
29      6.4.3.1.1  Effects on Total Mortality
30           The effects estimates from the newly reported studies generally comport well with those
31      derived from the earlier 1996 PM AQCD assessment, which reported risk estimates for excess
        March 2001                               6-228        DRAFT-DO NOT QUOTE OR CITE

-------
  1      total (nonaccidental) deaths associated with short-term PM exposures as generally falling within
  2      the range of ca. 1.5 to 8.5% per 50 /ug/m3 PM10 (24-h) increment and ca. 2.5 to 5.5% increase per
  3      25 //g/m3 PM2 5 (24-h) increment.
  4            Several new PM epidemiology studies which conducted time-series analyses in multiple
  5      cities were noted to be of particular interest, in that they provide evidence of effects across
  6      various geographic locations (using standardized methodologies) and more precise pooled effect
  7      size estimates with narrow confidence bounds reflecting the typically much stronger power of
  8      such multi-city studies over individual-city analyses.  Based on pooled analyses across multiple
  9      cities, the percent total (non-accidental) excess deaths per 50 Aig/m3 PM10 increment were
 10      estimated in different multi-city analyses to be:  (1) 2.3% in the 90 largest U.S. cities;  (2) 3.4% in
 11      10 U.S. cities; (3) 3.5% in the 8 largest Canadian cities; and (4) 2.0% in Western European  cities.
 12            Many new individual-city studies found positive associations (most statistically significant
 13      at p < 0.05) for the PM2 5 fraction, with effect size estimates typically ranging from ca. 2.0 to ca.
 14      8.5% per 25 Aig/m3 PM2 5 for U.S. and Canadian cities. Of the 10 or so new analyses that not
 15      only evaluated PM10 effects but also made an effort to compare fine versus  coarse fraction
 16      contributions to total mortality, only two are multi-city analyses yielding pooled effects
 17      estimates:  (1) the Klemm and Mason (2000) recomputation of Harvard Six Cities data,
 18      confirming the original published findings by Schwartz et al. (1996a); and (2) the Burnett et al.
 19      (2000) study of the 8 largest Canadian cities. Both of these studies found roughly comparable,
 20      statistically significant excess risk estimates for PM2 5, i.e., ca. 3% increased total mortality per
 21      25 //g/m3 PM25 increment.
 22            With  regard to possible coarse particle short-term exposure effects on mortality,  in those
 23      new studies which evaluated PM]0_2 5 effects as well as PM2 5 effects, the coarse particle (PM10_2 5)
 24      fraction was also consistently positively associated with increased total mortality, albeit the
 25      coarse effect size estimates were generally less precise than those for PM2 5  and statistically
 26      significant at p < 0.05 in only a few studies. Still, the overall picture tends to suggest that excess
 27      total mortality risks may well reflect actual coarse particle effects, in at least some locations.
28      This may be most consistently the case in arid areas (e.g., Southern California, the Phoenix area,
29      Mexico City, and Santiago, Chile) and during summer months (perhaps reflecting, in part,
30      stronger contributions of biogenic materials to coarse  fraction PM10_2 5 particles during warmer
31      weather). On the other hand, significant (or nearly significant) elevations in coarse PM-related

        March 2001                               6-229        DRAFT-DO NOT QUOTE OR CITE

-------
 1      total mortality risks elsewhere (e.g., eastern U.S. urban areas in the Harvard Six City Study, the
 2      8 largest Canadian cities, and Detroit, MI) may either reflect (a) typically moderate correlations
 3      there between PM10_2 5 and PM2 5 or, possibly, (b) true PM coarse fraction toxic potency. Excess
 4      total mortality risks associated with short-term (24-h) exposures to coarse fraction particles
 5      capable of depositing in the lower respiratory tract generally fall in the range of 0.5 to 6.0% per
 6      25 Acg/m3 PM10.25 increment for U.S. and Canadian cities.
 7           Three new papers provide particularly interesting new information on relationships between
 8      short-term and coarse particle exposures and total elderly mortality (age 65 and older) using
 9      exposure TEOM data from the EPA ORD NERL monitoring site in Phoenix, AZ. Each used
10      quite different models but each reported statistically significant relationships between mortality
11      and coarse PM, specifically PM,0.2 5, an indicator for the thoracic fraction of coarse-mode PM.
12           Smith et al. (2000) using as the exposure metric a three-day running average performed
13      linear regression of the square root of daily mortality on the long-term trend, meteorological and
14      PM-based variables. Two mortality variables  were used, total (non-accidental) deaths for the city
15      of Phoenix and the same for a larger, regional  area. Using a linear analysis, effects based on
16      coarse PM were statistically significant for both regions, whereas effects based on fine PM
17      (PM2 5) were not. However, when the possibility of a nonlinear response was taken into account,
18      no evidence was  found for a nonlinear effect for coarse PM, but fine PM was found to have a
19      statistically significant effect for concentration thresholds of 20 and 25 yUg/m3. There was no
20      evidence of confounding between fine and coarse PM, suggesting that fine and coarse PM are
21      "essentially separate pollutants having distinct effects".  Smith et al. (2000) also observed a
22      seasonal effect for coarse PM, the effect being statistically significant only during spring and
23      summer.  Based on  a principal component analysis of elemental concentrations, crustal elements
24      are highest in spring and summer and anthropogenic elements lowest, but Smith et al. (2000) felt
25      that the implication that crustal, rather than anthropogenic elements, were responsible for the PM
26      mortality was counterintuitive.
27           Clyde et al. (2000) used a more conventional model, a Poisson regression of log deaths on
28      linear PM variables; but they employed Bayesian model averaging to consider a wide variety of
29      variations in the basic model. They considered three regions, the Phoenix metropolitan area, a
30      small subset of zip code to give a region presumably with uniform PM2 5, and a still smaller zip
31      code region surrounding the monitoring site, thought to be uniform as to PM10 concentrations.

        March 2001                               6-230        DRAFT-DO NOT QUOTE OR CITE

-------
  1      The models considered lags of 0, 1, 2, or 3 days but only for single day PM variables (no running
  2      averages as used by Smith et al., 2000). A PM effect with a reasonable probability was found
  3      only in the uniform PM2 5 region and only for coarse PM.
  4           Mar et al. (2000) used conventional Poisson regression methods and limited their analyses
  5      to the smallest area (called Uniform PM,0 by Clyde et al).  They reported modeling data for lag
  6      days 0 to 4. Coarse PM was marginally significant on lag day 0.  No direct fine particle measures
  7      were statistically significant on day 0.  A regional sulfate factor determined from source
  8      apportionment, however, was statistically significant. No correlations were reported for the
  9      source apportionment factors but the correlation coefficient between sulfur (S) in PM2 5 (as
 10      measured by XRF) with coarse PM was only 0. 13, suggesting separate and distinct effects for
 1 1      regional sulfate and coarse PM.
 12           The above three studies of PM- total mortality relationships in Phoenix tend to suggest a
 13      statistical association of coarse PM with total elderly mortality in addition to and different from
 14      any relationship with fine PM,  fine PM components, or source  factors for fine PM.
 1 5           With regard to long-term PM exposure effects on total (non-accidental) mortality, the
 1 6      newly available evidence from the HEI Reanalyses of Harvard  Six Cities and ACS data (and
 1 7      extensions, thereof), substantiate well associations attributable  to chronic exposures to inhalable
 1 8      thoracic particles (indexed by PM,5 or PM10) and the fine fraction of such particles (indexed by
 1 9      PM2 5 and/or sulfates).  Statistically significant excess risk for total mortality was shown by the
 20      reanalyses to fall in the range of 4-1 8% per 20 //g/m3 PM15/10 increment and 14-28% per
 21       20 /ug/m3 PM2 5 increase, thus suggesting likely stronger associations with fine versus coarse
 22      fraction particles. Significant fine PM associations with total mortality were also found in the
 23       latest reported AHSMOG results for males, but not in females.
 24           Other recent studies on the relation of mortality to particle composition and source (Laden
 25      et al., 2000; Mar et al., 2000; Ozkaynak et al., 1996; Tsai et al., 2000) suggest that particles from
 26      certain sources may have much higher potential for adverse health effects than others, as
 27      delineated by source-oriented evaluations involving factor analyses. Laden et al. (2000)
 28      conducted factor analyses of the elemental composition of PM2 5 for Harvard Six Cities study
29      data for 1 979-1 988. In the analysis  for all six cities combined, the excess risk for daily mortality
30      was estimated to be 3.4% (CI, 1.7 to 5.2) per 10 ,ug/m3 increment in a mobile source factor; 1.1%
3 1      (CI, 0.3 to 2.0) per 10 ^g/m3 for a coal source factor, and -2.3% (CI, -5.8 to 1 .2) per 10
        March 200 1                               6-23 1        DRAFT-DO NOT QUOTE OR CITE

-------
 1      for a crustal factor. There was large variation among the cities and some suggestion of an
 2      association with a fuel oil factor identified by V or Mn, but it was not statistically significant.
 3           Mar et al. (2000) applied factor analysis to evaluate mortality in relation to 1995-1997 fine
 4      particle elemental components and gaseous pollutants (CO, NO2, SO2) in an area of Phoenix, AZ,
 5      close to the air pollution monitors. The PM2 5 constituents included sulfur, Zn, Pb,  soil-corrected
 6      potassium, organic and elemental carbon, and a soil component estimated from oxides of Al, Si,
 7      Ca, Fe, and It.  Based on models fitted using one pollutant at a time, statistically significant
 8      associations were found between total mortality and PMIO, CO (lags 0 and 1), NO2  (lags 0, 1, 3,
 9      4), S (negative), and soil (negative). Statistically significant associations were also found
10      between cardiovascular mortality and CO (lags 0 to 4), NO2 (lags 1 and 4), SO2 (lags 3 and 4),
11      PM2 5 (lags 1,3,4), PM10 (lag 0), PMi0_2 5 (lag 0), and elemental, organic, or total carbon.
12      Cardiovascular mortality was significantly related to a vegetative burning factor (high loadings
13      on organic carbon and soil-corrected potassium), motor vehicle exhaust/resuspended road dust
14      factor (with high loadings on Mn, Fe, Zn, Pb, OC, EC, CO, and NO2), and a regional sulfate
15      factor (with a high loading on S). However, total mortality was negatively associated with a soil
16      factor (high loadings on Al, Fe, Si) and a local SO2 source factor, but was positively associated
17      with  the regional sulfate factor.
18           Tsai et al. (2000) analyzed daily time series of total and cardiorespiratory deaths, using
19      short periods of 1981 -1983 data for Newark, Elizabeth, and Camden, NJ.  In addition to
20      inhalable particle mass (PM15) and fine particle mass  (PM2 5), the study evaluated data on metals
21      Pb, Mn, Fe, Cd, V, Ni, Zn, Cu, and three fractions of extractable organic matter.  Factor analyses
22      were carried out using the metals, CO, and sulfates. The most significant sources or factors
23      identified as predictors of daily mortality were oil burning (targets  V, Ni), Zn and Cd processing,
24      and sulfates. Other factors (dust, motor vehicles targeted by Pb and CO, industrial  Cu or Fe
25      processing) were not significant predictors. In Newark, oil burning sources and sulfates were
26      positive predictors, and Zn/Cd a negative predictor for total mortality. In Camden oil burning
27      and motor vehicle emissions predicted total mortality, but copper showed a marginal negative
28      association. Oil burning, motor vehicle emissions, and sulfates were predictors of
29      cardiorespiratory mortality in Camden. In Elizabeth,  resuspended dust indexed by Fe and Mn
30      showed marginal negative associations with mortality, as did industrial sources traced by Cu.


        March 2001                              6-232        DRAFT-DO NOT QUOTE OR CITE

-------
  1           The set of results from the above factor analyses studies do not yet allow one to identify
  2     with great certainty a clear set of specific high-risk chemical components of PM. Nevertheless,
  3     some commonalities across the studies seem to highlight the likely importance of mobile source
  4     and other fuel combustion emissions (and apparent lesser importance of crustal particles) as
  5     contributing to increased total or cardiorespiratory mortality.
  6
  7     6.4.3.1.2  Effects on Cause-Specific Mortality
  8           Numerous new studies have evaluated PM-related effects on cause-specific mortality.
  9     Most all report positive, often statistically significant (at p < 0.05), short-term (24-h) PM
 10     exposure associations with cardiovascular (CVD)- and respiratory-related deaths.  Cause-specific
 11      effects estimates appear to mainly fall in the range of 3.0 to 7.0% per 25 jug/m3 24-h PM2 5  for
 12     cardiovascular or combined cardiorespiratory mortality and 2.0 to 7.0% per 25 /wg/ni3 24-h PM2 5
 13     for respiratory mortality in U.S. cities.  Effect size estimates for the coarse fraction (PM10.2 5) for
 14     cause-specific mortality generally fall in the range of ca. 3.0 to 8.0% for cardiovascular and ca.
 15     3.0 to 16.0% for respiratory causes per 25 jUg/m3 increase in PM]0_2 5.
 16          Also of particular interest, the above noted study by Mar et al. examined the associations of
 17     a variety of PM indicators with cardiovascular mortality (for age >65),  again in the zip code area
 18     near the Phoenix monitoring site. For this end point, coarse PM was statistically significant on
 19     lag day 0 but not on subsequent lag days. PM2 5 and a number of fine PM indicators were
 20     statistically significant on lag day 1 but not on lag day 0. This suggests a distinct and separate
 21      relationship of PM2 5 and PM10_2 5. As in the case of total mortality, the  only fine PM indicator
 22      found to be statistically significant on lag day 0 was regional sulfate. However, the low
 23      correlation coefficient between S in PM2 5 and PM10.2 5 (r = 0.13) suggests that the two
 24      relationships represent different sets of deaths. Thus, there is some evidence suggesting that the
 25      risk of cardiovascular mortality , as well as that of total mortality, may be statistically associated
 26      with PM10_2 5 and that this relationship may be independent of any relationships with fine particle
 27      indicators.
28
29      6.4.3.2  Fine- and Coarse-Particulate Matter Effects on Morbidity
30           At the time of the 1996 PM AQCD, fine particle morbidity studies were mostly limited to
31      Schwartz et al. (1994), Neas et al. (1995, 1994); Koenig et al. (1993); Dockery et al. (1996); and

        March 2001                               6-233        DRAFT-DO NOT QUOTE OR CITE

-------
 1      Raizenne et al. (1996); and discussion of coarse particles morbidity effects was also limited to
 2      only a few studies (Gordian et al., 1996; Hefflin et al., 1991) which implicated PM10_2 5 a possible
 3      important fraction of PM10. Since the 1996 PM AQCD, several new studies have been published
 4      in which newly available size-fractionated PM data allowed investigation of the effects of both
 5      fine (PM2 5) and coarse (PM10_2 5) particles. Fine (FP) and coarse (CP) particle results are
 6      presented below for studies by morbidity outcome areas, as follows:  cardiovascular disease
 7      (CVD) hospital admissions (HA's), respiratory medical visits and hospital admissions, and
 8      respiratory symptoms and pulmonary function changes.
 9           As discussed in Section 6.3.1 (on cardiovascular effects associated with acute ambient PM
10      exposure), extensive evidence for significant PM,0 effects on cardiovascular-related hospital
11      admissions and visits has recently been provided by several new multi-city studies (Schwartz,
12      1999; Samet et al., 2000a,b; Zanobetti et al., 2000b) that yield pooled estimates of PM-CVD
13      effects  across numerous U.S. cities and regions. These studies found not only significant PM
14      associations, but also associations with other gaseous pollutants as well, thus hinting at likely
15      independent effects of certain gases (O3, CO, NO2, SO2) and/or interactive effects with PM.
16      These and other individual-city studies generally appear to confirm likely excess risk of
17      CVD-related hospital admission for U.S. cities in the range of 3-10% per 50 pig/in3 PM10,
18      especially among the elderly (> 65 yr).
19           In addition to the PM10 studies,  several new U.S. and Canadian studies evaluated fine-mode
20      PM effects on cardiovascular outcomes. Moolgavkar (2000a) reported PM2 5 to be significantly
21      associated with CVD HA for lag 0 and 1 in Los Angeles.  Burnett et al. (1997b) reported that fine
22      particles were significantly associated with CVD HA in a single pollutant model, but not when
23      gases were included in multipollutant models  for the 8 largest Canadian city data.  Stieb et al.
24      (2000)  reported both PM10 and PM2 5  to be associated with CVD emergency department (ED)
25      visits in single pollutant, but not multipollutant models. Similarly, Morgan et al. (1998) reported
26      that PM2 5 measured by nepholonetry was associated with CVD HA for all ages and 65+ yr, but
27      not in the multipollutant model. Tolbert et al. (2000a) reported that coarse particles were
28      significantly associated with dysrhythmias, whereas PM2 5 was not. Other studies (e.g., Liao
29      et al., 1999, Pope et al., 1999b,c) reported associations between increases in PM25 and several
30      measures of decreased heart rate variability, but Gold et al. (1998, 2000) reported a negative
31      association of PM2 5 with heart rate and decreased variability in r-MSSD (one heart rate

        March 2001                                6-234       DRAFT-DO NOT QUOTE OR CITE

-------
  1     variability measure). Overall, these studies appear to implicate fine particles, as well as possibly
  2     some gaseous copollutants, in cardiovascular morbidity, but the relative contributions of fine
  3     particles acting alone or in combination with gases such as O3, CO, NO2 or SO2 remain to be
  4     more clearly delineated and quantified.  The most difficult issue relates to interpretation of
  5     reduced PM effect size and /or statistical significance when copollutants derived from the same
  6     source(s) as PM are included in multipollutant models.
  7           Section 6.3.1 also discussed U.S. and Canadian studies that present analysis of coarse
  8     particles (CP) relationships to CVD outcomes.  Lippmann et al.  (2000) found significant positive
  9     associations of coarse particles (PMIO_2 5) with ischemic heart disease hospital admissions in
 10     Detroit (RR = 1.10, CI 1.026, 1.18). Tolbert et al. (2000a) reported significant positive
 11     associations of heart dysrhythmias with CP (p = 0.04) as well as for elemental carbon (p =
 12     0.004), but these preliminary results must be interpreted with caution until more complete
 13     analyses are carried out and reported. Burnett et al. (1997b) noted that CP was the most robust of
 14     the particle metrics examined to inclusion of gaseous covariates for cardiovascular
 15     hospitalization, but concluded that particle mass and chemistry could not be identified as an
 16     independent risk factor for exacerbation of cardiorespiratory disease in this study. Based on
 17     another Canadian study, Burnett et al. (1999), reported statistically significant associations for
 18     CP in univariate models but not in multipollutant models; but the use of estimated rather than
 19     measured PM exposures limits the interpretation of the PM results reported.
 20          The collective evidence reviewed above, in general, appears to suggest excess risks for
 21     CVD-related hospital admissions of ca. 4.0 to 10% per 25 /ug/m3 PM2 5 or PM10.2 5 increment.
 22          Section 6.3.2 also discussed new studies of effects of short-term PM exposure on the
 23     incidence of respiratory hospital admissions and medical visits.  Several new U.S. and Canadian
 24     studies have yielded particularly interesting results suggestive of roles of both fine and coarse
 25     particles respiratory-related hospital admissions. In an analysis of Detroit  data, Lippmann et al.
 26     (2000) found comparable effect size estimates for PM2 5 and PM10_2 5. That is, the excess risk for
 27     pneumonia hospital admissions (in no copollutant model) was 13% (CI 3.7, 22) per 25 /ug/m3
 28     PM2 5 and 12% (CI 0.8, 24) per 25 ^g/m3 PM10.2 5. Because PM2 5 and PM10.2 5 were not highly
29     correlated, the observed association between coarse particles and health outcomes were possibly
30      not confounded by smaller particles. Despite the greater measurement error associated with
31      PMio-2.5 than with either PM25 and PMIO, this indicator of the coarse particles within the thoracic

        March 2001                               6-235        DRAFT-DO NOT QUOTE OR CITE

-------
 1      fraction was associated with some of the outcome measures. The interesting result is that PMi0.2 5
 2      appeared to be a separate factor from other PM metrics, especially given the effect estimates of
 3      PM10_2 5 with pneumonia hospital admissions, (lag 1; RR = 1.11, 95% CI:  1.006, 1.233). Burnett
 4      et al. (1997b) also reported PM (PM10, PM2 5, and PM10_2 5) associations with respiratory hospital
 5      admissions, even with O3 in the model. Notably, the PM10_2 5 association was significant (RR =
 6      1.13 for 25 yUg/m3; CI = 1.05 - 1.20); and inclusion of ozone still yielded a significant coarse
 7      mass RR = 1.11 (CI = 1.04 -1.19). Moolgavkar et al. (2000) showed the most consistent
 8      association for PM10 across lags (0-4d), while PM2 5 yielded the strongest positive PM metric
 9      association at lag 3 days.  Also, Moolgavkar (2000a) reported that, in Los Angeles, both PM10
10      and PM2 5 yielded both positive and negative associations at different lags for single pollutant
11      models but not in two pollutant models.  Delfino et al. (1997) reported that both PM2 5 and PMIO
12      are positively associated with ED visits for respiratory disease. Morgan et al. (1998) reported
13      that PM2 5 estimated from nephelometry yielded a PM2 5 association with COPD HA for 1-hr max
14      PM that was more positive than 24-h average PM2 5.
15           Some new studies appear to substantiate PM associations with asthma-related hospital
16      admissions. For example, Norris et al (1999) reported associations of emergency department
17      visits for asthma in children with both PM2 5 and PM10.2 5. Two other studies presented uniquely
18      different analyses of hospital admissions in the Seattle, Washington area. Sheppard et al. (1999)
19      studied relationships between PM metrics that included PM10.2 5 and non-elderly adult hospital
20      admissions for asthma in the greater Seattle area and reported significant relative rates for PMto,
21      PM25andPM10,25(lagged 1 day). For PMI025, the relative risk was 1.04 (95% CI 1.01, 1.07).
22      In a different analysis, Lumley and Heagerty (1999) examined PM( and PM1(M in the King
23      County, WA (Seattle) area during the same time period but for hospital admissions for overall
24      respiratory disease.  Since only a significant hospital admission association was found with PM, 0
25      and not PM10.,, a dominant role by sub-micron particles in PM2 5 - asthma HA association was
26      suggested, but this may not be an appropriate conclusion based on several differences between
27      the study analysis methods and differences between asthma versus respiratory outcome  measures
28      used in the two Seattle studies.
29           Several other studies (Chen et al. 2000; Choudhury et al., 1997; Moolgavlar 2000a;
30      Lippsett et al., 1997) report results for areas (e.g., Reno-Sparks, NV; Anchorage, AK; Phoenix,
31      AZ; Santa Clara, CA) where coarse particles tend to constituent a large fraction of PM,0 but no

        March 2001                               6-236        DRAFT-DO NOT QUOTE OR CITE

-------
  1     measures of PMIO_25 were available.  These studies showing significant PM10 effects on
  2     respiratory hospital admissions provide additional data suggestive of likely coarse particle effects
  3     on respiratory morbidity.
  4          Thus, although PM10 mass has most often been implicated as the PM pollution index
  5     affecting respiratory hospital admissions, the overall collection of new studies reviewed in
  6     Section 6.3.2 appear to suggest relative roles for both fine and coarse PM mass fractions, such as
  7     PM2 5 and PM10.2 5.
  8          Section 6.3.3 assessed relationships between PM exposure on lung function and respiratory
  9     symptoms.  While most data examine PM10 effects, several studies also examined fine and coarse
 10     fraction effects.  Schwartz and Neas (2000) report that cough was the only response in which
 11     coarse particles appeared to provide an independent contribution to explaining the increased
 12     incidence.  The correlation between CM and PM25 was moderate (0.41). Coarse particles had
 13     little association with evening peak flow. Tiittanen et al. (1999) also reported a significant effect
 14     of PM10.2 5 for cough. Thus, cough may be an appropriate outcome related to coarse particle
 15     effects.  However, the limited data base suggests that further study  is appropriate. The report by
 16     Zhang, et al. (2000) of an association between coarse particles and  the indicator "runny nose" is
 17     noted also.
 18          For respiratory symptoms and PFT changes, several new asthma studies report relationships
 19     with ambient PM measures. The peak flow analyses results for asthmatics tend to show small
 20     decrements for both PM10 and PM2 5. Several studies included PM2 5 and PM10 independently in
 21      their analyses of peak flow.  Of these, Gold et al. (1999), Naeher et al. (1999), Tiittanen et al.
 22     (1999), Pekkanen et al. (1997), and Romieu et al. (1996) all found comparable results  for PM2 5
 23     and PMI0. The study of Peters et al. (1997b) found slightly larger effects for PM2 5.  The study of
 24     Schwartz and Neas (2000) found larger effects for PM2 5 than for coarse mode particles.  Three
 25      studies included both PMIO and PM2 5 in their analyses of respiratory symptoms. The studies of
 26      Peters et al. (1997b) and Tiittanen et al. (1999) found similar effects for the two PM measures.
 27      Only the Romieu et al. (1996) study found slightly larger effects for PM2 5.
28           For non-asthmatics, several studies evaluated PM2 5 effects.  Naeher et al.  (1999) reported
29      similar AM PEF decrements for both PM2 5 and PM10.  Neas et al. (1996) reported a
30      nonsignificant negative association for PEF and PM2 „ and Neas et  al. (1999) also reported
31      negative but nonsignificant PEF results.  Schwartz and Neas (2000) reported a significantly PM

        March 2001                              6-237        DRAFT-DO NOT QUOTE OR CITE

-------
  1      PEF association with PM2 5, and Tiittanen et al. (1999) also reported negative but nonsignificant
  2      association for PEF and PM2 5. Gold et al. (1999) reported significantly PEF results.  Schwartz
  3      and Neas (2000) reported significant PM2 5 effects relative to lower respiratory symptoms.
  4      Tiittanen et al. (1999) showed significant effects for cough and PM2 5 for a 4-day average.
  5           Another study, Peters et al. (1997b) in Erfurt in 1992 is unique for two reasons: (1) they
  6      studied the size distribution in the range 0.01  to 2.5 ^m and (2) examined the number of
  7      particles. They report that the health effects of 5 day means of the number count (NC) for
  8      ultrafine particles were larger than those related to the mass of the fine particles. For NC 0.01 -
  9      0.1, cough was significant for the same day and the five day mean.
10           In a chronic respiratory disease study of 22-24 North American communities evaluated in
11      the 1996 PM AQCD, Raizenne et al. (1996) found PM2, to be related to a statistically significant
12      FVC deficit of-3.21% (-4.98, -1.41). Dockery et al. (1996) also reported PM2, associations
13      with increased bronchitis; odds ratio = 1.50 (95% CI = 0.91, 2.47).
14           The above new studies offer much more information than was available in 1996. Effects
15      were noted for several morbidity endpoints: cardiovascular hospital admissions, respiratory
16      hospital admissions and cough.  Still insufficient data exists from these relatively limited studies
17      to allow strong conclusions at this time as to which size-related ambient PM components may be
18      most strongly related to one or another morbidity endpoints.  Very preliminarily, however,  fine
19      particles appear to be more strongly implicated in cardiovascular outcomes than are coarse ones,
20      whereas both seem to impact respiratory endpoints.
21
22      6.4.4  The Question of Lags
23           In most of the past air pollution health effects time-series studies, after the basic model (the
24      best model with  weather and seasonal cycles as covariates) was developed, several pollution lags
25      (usually 0 to 3 or 4 days) were individually introduced and the most significant lag(s) chosen for
26      the RR calculation. While this practice may bias the chance of finding a significant association,
27      without a firm biological reason to establish a fixed pre-determined lag, it appears reasonable.
28      Due to likely individual variability in response to air pollution, the apparent lags of effects
29      observed for aggregated population counts are expected to be "distributed" (i.e., symmetric or
30      skewed bell-shape).  The "most significant lag" in such distributed lags is also expected to
31      fluctuate statistically. The "vote-counting" of the most significant lags reported in the past
        March 2001                               6-238       DRAFT-DO NOT QUOTE OR CITE

-------
  1     PM-mortality studies shows that 0 and 1 day lags are, in that order, the most frequently reported
  2     "optimal" lags, but such estimates may be biased because these lags are also likely the most
  3     frequently examined ones.  Thus, a more systematic approach across different data sets was
  4     needed to investigate this issue.
  5          The recent Samet et al. (2000b) analysis of the 90 largest U.S. cities provides particularly
  6     useful  information on this matter. Figure 6-11 depicts Samet et al.'s overall pooled results,
  7     showing the posterior distribution of PM10 effects for the 90 cities for lag 0, 1, and 2 days.  It can
  8     be seen that the effect size estimate for lag  Iday is about twice that for lag 0 or lag 2 days, though
  9     their distributions overlap.  However, a careful examination of Figures 6 and 7 in NMMAPS I
 10     suggests that the maximum PM10 effect may occur in different cities with somewhat different lag
 11     relationships. In terms of the magnitude of the estimated PMIO effects, Table 6-24, based on
 12     NMMAPS I Figure 7 (posterior bivariate distribution for each county; PM10 effect adjusted for
 13     O3), suggests that somewhat different patterns may apply in different locations. These data
 14     suggest that while lag 1 effects are typically the largest, there may be some situations in which
 15     lag 0 or lag 2 effects are larger.
 16          The NMMAPS mortality and morbidity analyses, and another HEI-sponsored study on PM
 17     components (Lippmann et al., 2000) illustrate three different ways to deal with temporal
 18     structure: (1) assume all sites have the same lag, e.g.,  1 day, for a given effect; (2) use the lag or
 19     moving average giving the largest or most significant effect; and (3) use a flexible distributed lag
 20     model, with parameters adjusted to each site.
 21          The NMMAPS mortality analyses used the first approach.  This approach introduces a
 22     consistent response model across all locations. However, since the cardiovascular, respiratory, or
 23     other causes of acute mortality usually associated with PM are not at all specific, there is  little
 24     a priori reason to believe that they must have the same relation to current or previous PM
 25     exposures at different sites.  The imposed consistency in lag that maximizes the aggregate effect
 26     of lag 1 across all cities, in Figure 15-18 and 24 of NMMAPS II, may obscure important regional
 27     or local differences for lags other than 1 day.
28          The NMMAPS morbidity studies evaluate  0- and 1-day lags, the moving average of 0 and
29     1-day lags, polynomial distributed lag models, and unrestricted distributed lag models. The
30     first-stage models for each city in the study were fitted for each city, with no restriction as to a
31      consistent model across all cities, and combined  across all 14 cities in the second stage as shown

        March 2001                                6-239       DRAFT-DO NOT QUOTE OR CITE

-------
__
....
— —
LAGO
LAG1
LAG 2
                                                                              0.99
                                                                              1.00
                                                                              0.98
              I
             -0.2
 I
0.0
                  0.2         0.4         0.6         0.8         1.0
% Change in Mortality per 10 |jg/m3 Increase in PM10
      Figure 6-11.  Marginal posterior distribution for effects of PM10 on all cause mortality at
                   lag 0,1, and 2 for the 90 cities. From Samet et al. (2000a,b). The numbers in
                   the upper right legend are posterior probabilities that overall effects are
                   greater than 0.
1     in Table 14 and Figure 23 of NMMAPS II.  A comparison of the data tabulated in the NMMAPS
2     Report Appendices shows large differences across cities in the apparent magnitude of the PM,0
3     effect, depending on how the PM concentration data over the preceding few days are used.
4          The approach used in Lippmann et al. (2000) and many other studies is to use the model
5     that maximizes some global model goodness-of-fit criterion.  This leads to selection of different
6     models at different sites, as might be expected. However, the best-fitting model (for lags, for
7     example) is often the model with the largest or most significant PM10 coefficient.  All models for
8     the pollutant(s) of interest are usually compared among themselves only after a preliminary
9     baseline model has been fitted. The baseline model takes into account most of the other
      March 2001
                     6-240
                                        DRAFT-DO NOT QUOTE OR CITE

-------
         TABLE 6-24. COMPARISON
           ANALYSES FOR 0,1, AND
OF PM10 EFFECT SIZES ESTIMATED BY NMMAPS
2 DAY LAGS FOR THE 20 LARGEST U.S. CITIES
        County
     Ordered PMm effect sizes
                                                      10
        Los Angeles
        New York
        Chicago
        Dallas/Fort Worth
        Houston
        San Diego
        Santa Ana /Anaheim
        Phoenix
        Detroit
        Miami
        Philadelphia
        Seattle
        San Jose
        Cleveland
        San Bernardino
        Pittsburgh
        Oakland
        San Antonio
        Riverside
     Lag 0 < lag 1 « lag 2
     Lag 0 = lag 1 » lag 2
     Unreadable
     Lag 0> lag l,lag 1< lag 2
     Lag 0< lag l,lag 1 > lag 2
     Lag 0 = lag 1 > lag 2
     Lag 0 > lag 1 > lag 2
     Lag 0 = lag 1 < lag 2
     Lag 0< lag l,lag 1 > lag 2
     Lag 0 < lag 1 = lag 2
     Lag 0< lag l,lag 1 > lag 2
     Lag 0< lag 1, lag 1 > lag 2
     Lag 0 > lag 1 = lag 2
     Lag 0> lag l,lag 1 < lag 2
     Lag 0 > lag 1 = lag 2
     Lag 0< lag l,lag 1 > lag 2
     Lag 0 < lag 1 = lag 2
     Lag 0 = lag 1 < lag 2
     Lag 0< lag l,lag 1 > lag 2
1     variables with which PM10 could be plausibly associated, so that the remaining variation in
2     morbidity or mortality that can be explained by including PM,0 indicators with different temporal
3     structures is nearly "orthogonal" or independent of the baseline model.  The restriction to the
4     same lag day at all sites certainly increases the precision of that estimate, but possibly at the cost
5     of obscuring different relationships between time of exposure and health effect at other sites.
      March 2001
        6-241
DRAFT-DO NOT QUOTE OR CITE

-------
 1           An additional complication in assessing the shape of a distributed lag is that the apparent
 2      spread of the distributed lag may depend on the pattern of persistence of air pollution (i.e.,
 3      episodes may persist for a few days), which may vary from city to city and from pollutant to
 4      pollutant. If this is the case, fixing the lag across cities or across pollutants may not be ideal, and
 5      may tend to obscure important nuances of lag structures that may provide important clues to
 6      possible different lags between PM exposures and different cause-specific effects.
 7           Thus, it is possible that the extent of lag and its spread may vary depending on the cause of
 8      death. For example, Rossi et al. (1999) report that, in their analysis of TSP-cause specific
 9      mortality in Milan, Italy, the lags varied for different cause of death (i.e., same day for respiratory
10      infections and heart failure; 3-4 days for myocardial infarction  and COPD).  Thus, the lag for
11      total mortality may exhibit mixed lags (weighted by the frequency of deaths in each cause).
12      Another example was reported for a recent Mexico City study (Borja-Aburto et al., 1998), in
13      which they found significant PM2 5-total mortality associations  for same day and 4-day lag, but
14      not for the intervening 2 to 3 days (percent increases per 25 Aig/m3 were 3.38, -4.00, 1.03, 1.08,
15      3.43, 2.49, for 0 through 5 day lags, respectively). The authors state: "This phenomenon is
16      consistent with both a harvesting of highly susceptible persons on the day of exposure to high
17      pollution levels and a lagged increase in mortality due to  delayed effects of reduction of
18      pulmonary defenses, cardiovascular complications, or other homeostatic changes among
19      less-compromised individuals". It  is interesting to note that Wichmann et al. (2000) also
20      reported that the most predictive single day effects on mortality for mass concentrations of
21      0.01-2.5 /2 particles were either immediate (0-1 d lag) or delayed (4-5 d lag) for their data from
22      Erfurt, Germany.
23           It should also be noted that if one chooses the most significant single lag day only, and if
24      more than one lag day shows positive (significant or otherwise) associations with mortality, then
25      reporting a RR for only one lag would also underestimate the pollution effects.  Schwartz
26      (2000b) investigated this issue, using the 10 U.S. cities data where daily PM10 values were
27      available for 1986-1993.  Daily total (non-accidental) deaths of persons 65 years of age and older
28      were analyzed. For each city, a GAM Poisson model adjusting for temperature, dewpoint,
29      barometric pressure, day-of-week, season, and time was fitted.  Effects of distributed lag were
30      examined using four models:  1-day mean at lag 0 day; 2-day mean at lag 0 and 1  day; second-
31      degree distributed lag model using lags 0 through 5 days; unconstrained distributed lag model

        March 2001                               6-242        DRAFT-DO NOT QUOTE OR CITE

-------
  1     using lags 0 through 5 days. The inverse variance weighted averages of the ten cities' estimates
  2     were used to combine results.  The results indicated that the effect size estimates for the
  3     quadratic distributed model and unconstrained distributed lag model were similar.  Both
  4     distributed lag models resulted in substantially larger effect size estimates (7.25% and 6.62%,
  5     respectively, as percent excess total death per 50 /wg/m3 increase in PM10) than the single day lag
  6     (3.29%) and moderately larger effect size estimates than the two-day average models (5.36%).
  7     Samet et al. (2000a,b) also applied 7- and 14-day unconstrained distributed lag models to
  8     Chicago, Minneapolis/St. Paul, and Pittsburgh data, and reported that the sum of the 7-day
  9     distributed lag coefficients was greater than the estimates based on a single day's value, but the
 10     14-day estimate was substantially lower than the 7-day estimate in Chicago and Minneapolis/
 11     St. Paul.  Thus, it is possible that the usual RR estimate using one lag day may underestimate PM
 12     effects.
 13
 14     6.4.5  New Assessments of Mortality Displacement
 15          There have been a few studies that investigated the question of "harvesting", a phenomenon
 16     in which a deficit in mortality occurs following days with (pollution-caused) elevated mortality,
 17     due to depletion of the susceptible population pool. This issue is very important in interpreting
 18     the public health implication of the reported short-term PM mortality effects.  The 1996 PM
 19     AQCD discussed suggestive evidence observed by Spix et al. (1993) during a period when air
 20     pollution levels were relatively high. Recent studies, however, generally typically used data from
 21      areas with lower, non-episodic pollution levels.
 22          Schwartz (2000c) separated time-series air pollution, weather, and mortality data from
 23      Boston, MA, into three components:  (1) seasonal and longer fluctuations; (2) "intermediate"
 24      fluctuations; (3) "short-term" fluctuations. By varying the cut-off between the intermediate and
 25      short term, evidence of harvesting was sought. The idea is, for example, if the extent of
 26      harvesting were a matter of a few days, associations between weekly average values of mortality
 27      and air pollution (controlling for seasonal cycles) would not be seen. For COPD, Schwartz
 28      (2000c) reported evidence indicating that most of the mortality was only displaced by a few
29      weeks; for pneumonia, heart attacks, and all  cause mortality, the effect size increased as longer
30      time scales were included.  The percent increase in deaths associated with a 25 ^g/m3 increase in
31      PM2 5 increased from 5.3% (95%CI:  6.8, 9.0) to 9.64% (95%CI: 8.2, 11.1).
        March 2001                               6-243       DRAFT-DO NOT QUOTE OR CITE

-------
 1           Schwartz and Zanobetti (2000) used the same approach described above to analyze a larger
 2      data set from Chicago, IL for 1988-1993.  Total (non-accidental), in-hospital, out-of-hospital
 3      deaths, as well as heart disease, COPD, and pneumonia elderly hospital admissions were
 4      analyzed to investigate possible PM10"harvesting" effects. GAM Poisson models adjusting for
 5      temperature, relative humidity, day-of-week, and season were applied in baseline models using
 6      the average of the same day and previous day's PM,0. Seasonal and trend decomposition
 7      techniques called STL were applied to the health outcome and exposure data to decompose them
 8      into different time-scales (i.e., short-term to long-term), excluding long seasonal cycles (120 day
 9      window). The associations were examined with smoothing windows of 15, 30, 45, and 60 days.
10      The effect size estimate for deaths outside hospital was larger than for deaths inside hospital.
11      All cause mortality showed an increase in effect size at longer time scales. The effect size for
12      deaths outside hospital increases more steeply with increasing time scale than that for inside
13      hospital  deaths.
14           Zanobetti et al. (2000a) used GAM distributed lag models to help quantify mortality
15      displacement in Milan, Italy,  1980-1989. Non-accidental total deaths were regressed on smooth
16      functions of TSP distributed over the same day and the previous 45 days using penalized splines
17      for the smooth terms and seasonal cycles, temperature, humidity, day-of-week, holidays, and
18      influenza epidemics.  The mortality displacement was modeled as the initial positive increase,
19      negative rebound (due to depletion), followed by another positive coefficients period, and the
20      sum of the three phases were considered as the  total cumulative effect.  TSP was positively
21      associated with mortality up to 13 days, followed by nearly zero coefficients between 14 and
22      20 days, and then followed by smaller but positive coefficients up to the 45th day (maximum
23      examined).  The sum of these coefficients was over three times larger than that for the single-day
24      estimate.
25           Zeger et al. (1999) first illustrated, through simulation, the implication of harvesting for PM
26      regression coefficients (i.e., mortality relative risk) as observed in frequency domain.  Three
27      levels of harvesting, 3 days, 30 days, and 300 days were simulated. As  expected, the shorter the
28      harvesting, the larger the PM coefficient in the  higher frequency range.  However, in the real data
29      from Philadelphia, the regression coefficients increased toward the lower frequency range,
30      suggesting that the extent of harvesting, if it exists, is not in the short-term range. Zeger
31      suggested that "harvesting-resistant" regression coefficients could be obtained by excluding the

        March 2001                               6-244        DRAFT-DO NOT QUOTE OR CITE

-------
  1      coefficients in the very high frequency range (to eliminate short-term harvesting) and in the very
  2      low frequency range (to eliminate seasonal confounding). Since the observed frequency domain
  3      coefficients in the very high frequency range were smaller than those in the mid frequency range,
  4      eliminating the "short-term harvesting" effects would only increase the average of those
  5      coefficients in the rest of the frequency range.
  6           Frequency domain analyses are rarely performed in air pollution health effects studies,
  7      except perhaps the spectra analysis (variance decomposition by frequency) to identify seasonal
  8      cycles.  Examinations of the correlation by frequency (coherence) and the regression coefficients
  9      by frequency (gain) may be useful in evaluating the potentially frequency-dependent
 10      relationships among multiple time series.  A few past examples in air pollution health effects
 11      studies include:  (1) Shumway et al.'s (1983) analysis of London mortality analysis, in which
 12      they observed that significant coherence occurred beyond two week periodicity (they interpreted
 13      this as "pollution has to persist to affect mortality"); (2) Shumway et al.'s (1988) analysis of
 14      Los Angeles mortality data, in which they also found larger coherence in the lower frequency;
 15      (3) Ito's (1990) analysis of London mortality data in which he observed relatively constant gain
 16      (regression coefficient) for pollutants across the frequency range,  except the annual cycle. These
 17      results also suggest that associations and effect size, at least, are not concentrated in the very high
 18      frequency range.
 19           Schwartz (2000c), Zanobetti et al. (2000a), and Zeger et al.'s (1999) results all suggest that
 20     the extent of harvesting, if any, is not a matter of only a few days. Other past studies that used
 21      frequency domain analyses are also at least qualitatively in agreement with the evidence against
 22     the short-term only harvesting. Since very long wave cycles (> 6 months) need to be controlled
 23      in time-series analyses to avoid seasonal confounding, the extent of harvesting beyond 6 months
 24      periodicity is not possible in time-series study design. While these studies suggest that observed
 25      short-term associations  are not simply due to short-term harvesting, more data are needed to
 26      obtain quantitative estimates of the extent of prematurity of deaths.
 27
28      6.4.6  New Assessment of Threshold in Concentration-Response
29             Relationships
30           In the 1996 PM AQCD, the limitations of identifying 'threshold' in the concentration-
31      response relationships in observational studies were discussed including the low data density in

        March 2001                               6-245       DRAFT-DO NOT QUOTE OR CITE

-------
 1      the lower PM concentration range, the small number of quantile indicators often used, and the
 2      possible influence of measurement error.  Also, a threshold for a population, as opposed to a
 3      threshold for an individual, has some conceptual issues that need to be noted.  For example,
 4      Schwartz (1999) discussed that, since individual thresholds would vary from person to person
 5      due to individual differences in genetic level susceptibility and pre-existing disease conditions, it
 6      would be almost mathematically impossible for a threshold to exist in the population. The
 7      person-to-person difference in the relationship between personal exposure and the concentration
 8      observed at a monitor would also add to the variability.  Because one cannot directly measure but
 9      can only compute or estimate a population threshold, it would be difficult to interpret an
10      observed threshold, if any, biologically. Despite these issues, several studies have attempted to
11      address the question of threshold by analyzing large databases, or by conducting simulations.
12           Cakmak et al. (1999) investigated methods to detect and estimate threshold levels in time
13      series studies. Based on the realistic range of error observed from actual Toronto pollution data
14      (average site-to-site correlation:  0.90 for O3; 0.76 for COH; 0.69 for TSP; 0.59 for SO2; 0.58  for
15      NO2; and 0.44 for CO), pollution levels were generated with multiplicative error for six levels of
16      exposure error (1.0, 0.9, 0.8, 0.72, 0.6, 0.4, site-to-site correlation). Mortality series were
17      generated with three PMIO threshold levels (12.8 /^g/m3, 24.6 /^g/m3, and 34.4 Mg/m3). LOESS
18      with a 60% span was used to observe the exposure-response curves for these 18 combinations of
19      exposure-response relationships with error. A parameter threshold model was also fit using non-
20      linear least squares.  Graphical presentations indicate that LOESS adequately detects threshold
21      under no error, but the thresholds were "smoothed out" under the extreme error scenario. Use of
22      a parametric threshold model was adequate to give "nearly unbiased" estimates of threshold
23      concentrations even under the conditions of extreme measurement error, but the uncertainty in
24      the threshold estimates increased with the degree of error. They concluded, "if threshold exists,
25      it is highly likely that standard statistical analysis can detect it".
26           Schwartz and Zanobetti (2000) investigated the presence of threshold by simulation and
27      actual data analysis of 10 U.S. cities: New Haven, CT; Pittsburgh, PA; Birmingham, AL;
28      Detroit, MI; Canton, OH; Chicago, IL; Minneapolis-St. Paul, MN; Colorado Springs, CO;
29      Spokane, WA; and Seattle, WA, where daily PM10 were available for years 1986-1993. First, a
30      simulation was conducted to show that the combining smoothed curves across cities (the authors
31      called this approach "meta-smoothing") could produce unbiased exposure-response curves.

        March 2001                               6-246        DRAFT-DO NOT QUOTE OR CITE

-------
  1     Three hypothetical curves (linear, piecewise linear, and logarithmic curves) were used to generate
  2     mortality series in the 10 cities, and GAM Poisson models were used to estimate respective
  3     exposure-response curves. Effects  of measurement errors were also simulated. In the analysis of
  4     actual 10 cities data, GAM Poisson models were fitted, adjusting for temperature, dewpoint, and
  5     barometric pressure, and day-of-week. Smooth function of PM10 with the same span (0.7) was
  6     used in each of the cities. The predicted values of the log relative risks were computed for
  7     2 Mg/m3 increments between 5.5 /ug/m3 and 69.5 /ug/m3 of PM10 levels. Then, the predicted
  8     values were combined across cities using inverse-variance weighting.  The simulation results
  9     indicated that the "meta-smoothing" approach did not bias the underlying relationships for the
 10     linear and threshold models, but did result in a slight downward bias for the logarithmic model.
 11     Measurement error (additive or multiplicative) in the simulations did not cause upward bias in
 12     the relationship below threshold. The threshold detection in the simulation was not very
 13     sensitive to the choice of span in smoothing. In the analysis of real data from 10 cities, the
 14     combined curve did not show evidence of a threshold in the PMi0-mortality associations.
 15           The Smith et al.  (2000) study  of associations between daily total mortality and PM2 5 and
 16     PM,0_25 in Phoenix, AZ (during 1995-1997) also investigated the possibility of a threshold.
 17     In the linear model, the authors found that mortality was significantly associated with PM10_2 5,
 18     but not with PM2 5. In modeling possible thresholds, they applied: (1)  a piecewise linear model
 19     in which several possible thresholds were specified; and (2) a B-spline (spline with cubic
 20     polynomials) model with 4 knots. Using the piecewise model, there was no indication that there
 21      was a threshold for PM10_2 5.  However, for PM2 5, the piecewise model resulted in suggestive
 22     evidence for a threshold, around 20 to 25 /ug/m3. The B-spline results also showed no evidence
 23     of threshold for PM10_2 5, but for PM2 5,  a non-linear curve showed a change in the slope around
 24     20 /xtg/m3. A further Bayesian analysis for threshold selection suggested a clear peak in the
 25      posterior density around 22 /ug/m3.  These results, if they in fact reflect reality,  make it difficult to
 26      evaluate the relative roles of different PM components (in this case, PM2 5 vs. PM,0_2 5).
27      However, the concentration-response curve for PM2 5 presented in this publication suggests more
28      of a U- or V-shaped relationship than the usual "hockey stick" relationship. Such a relationship
29      is, unlike temperature-mortality relationship, difficult to interpret biologically.  Because the
30      sample size of this data (=3 years) is relatively small,  further investigation of this issue using
31      similar methods but a larger data set is warranted.

        March 2001                               6-247        DRAFT-DO NOT QUOTE OR CITE

-------
 1           Daniels et al. (2000) examined the presence of threshold using the largest 20 U.S. cities for
 2      1987-1994. The authors compared three log-linear GAM regression models: (1) using a linear
 3      PM10 term; (2) using a cubic spline of PM10 with knots at 30 and 60 ^g/m3 (corresponding
 4      approximately to 25 and 75 percentile of the distribution); and, (3) using a threshold model with
 5      a grid search in the range between 5 and 200 /wg/m3 with 5 //g/m3 increment. The covariates
 6      included in these models are similar to those used by the same research group previously (Kelsall
 7      et al., 1997; Samet et al., 2000a,b), including the smoothing function of time, temperature and
 8      dewpoint, and day-of-week indicators. Total, cardiorespiratory, and other mortality series were
 9      analyzed. These models were fit for each city separately, and for model (1) and (2), the
10      combined estimates across cities were obtained by using inverse variance weighting if there was
11      no heterogeneity across cities, or by using a two-level hierarchical model  if there  was
12      heterogeneity. The best fit among the models, within each city and over all cities, were also
13      determined using the Akaike's Information Criterion (AIC). The results using the spline model
14      showed that, for total and cardiorespiratory mortality, the spline curves were roughly linear,
15      consistent with the lack of a threshold. For mortality from other causes, however, the curve did
16      not increase until PM10 concentrations exceeded 50 A^g/m3. While the test of heterogeneity
17      indicated that there was considerable heterogeneity in these curves across cities, the shapes of the
18      curves were similar across cities, with no indication of one city unduly influencing the overall
19      estimate of the curves. The hypothesis of linearity was examined by comparing the AIC values
20      across models.  The results suggested that the linear model was preferred  over the spline and the
21      threshold models.  Thus, these results suggest that linear models without a threshold may well be
22      appropriate for estimating the effects of PM,0 on the types of mortality of main interest.
23
24      6.4.7 New Theoretical Assessments of Consequences  of Measurement Error
25           Since the 1996 PM AQCD, there have been some advances in conceptual framework
26      development to investigate the effects of measurement error on PM health effects estimated in
27      time-series studies.  Several new studies evaluated the extent of bias caused by measurement
28      errors under a number of scenarios with varying extent of error variance and covariance structure
29      between co-pollutants.
30           Zidek et al. (1996) investigated, through simulation, the joint effects of multi-collinearity
31      and measurement error in Poisson regression model, with two covariates with varying extent of
        March 2001                              6-248       DRAFT-DO NOT QUOTE OR CITE

-------
  1     relative errors and correlation. Their error model was of classical error form (W=X+U, where W
  2     and X are surrogate and true measurements, respectively, and the error U is normally distributed).
  3     The results illustrated the transfer of effects from the "causal" variable to the confounder.
  4     However, for the confounder to have larger coefficients than the true predictor, the correlation
  5     between the two covariates had to be large (r = 0.9), with moderate error (a > 0.5) for the true
  6     predictor, and no error for the confounder in their scenarios. The transfer-of-causality effect was
  7     mitigated when the confounder also became subject to error. Another interesting finding that
  8     Zidek et al, reported is the behavior of the standard errors of these  coefficients: when the
  9     correlation between the covariates was high (r = 0.9) and both covariates had no error, the
 10     standard errors for both coefficients were inflated by factor of 2; however, this phenomenon
 11     disappeared when the confounder had error.  Thus, multi-collinearity influences the significance
 12     of the coefficient of the causal variable only when the confounder is accurately measured.
 13           Zeger et al. (2000) also conducted a mathematical analysis of PM mortality effects in
 14     ordinary least square model (OLS) with the classical error model, under varying extent of error
 15     variance and correlation between two predictor variables. The error described here was
 16     analytical error (e.g., discrepancy between the co-located monitors). In general, they found that
 17     positive regression coefficients are only attenuated, but null predictors (zero coefficient) or weak
 18     predictors are only able to appear stronger than true positive predictors under unusual conditions:
 19     (1) true predictors must have very large positive or negative  correlation (i.e., ]r| > 0.9); (2)
 20     measurement error must be substantial (i.e., error variance ~ signal variance); and (3)
 21     measurement errors must have a large negative correlation.  They concluded that estimated FP
 22     health effects are likely underestimated, although the magnitude of bias due to the analytical
 23     measurement error is not very large.
 24          Zeger et al. (1999) illustrated the implication of the classical error model and the Berkson
25     error model (i.e., X = W + U) in the  context of time-series study design.  Their simulation of the
26     classical error model with two predictors, with various combinations of error variance and
27     correlation between the predictors/error terms, showed results similar to  those reported by Zidek
28     et al. (1996).  Most notably, for the transfer of the effects of one variable to the other (i.e., error-
29     induced confounding) to be large, the two predictors or their errors need to be substantially
30     correlated. Also, for the spurious association of a null predictor to  be more significant than the


        March 2001                                6-249       DRAFT-DO NOT QUOTE OR CITE

-------
 1      true predictor, their measurement errors have to be extremely negatively correlated—a condition
 2      not yet demonstrated as occurring in actual air pollution data sets.
 3           Zeger et al. also laid out a comprehensive framework for evaluating the effects of exposure
 4      measurement error on estimates of air pollution mortality relative risks in time-series studies.
 5      The error, the difference between personal exposure and the central station's measurement of
 6      ambient concentration was decomposed into three components:  (1) the error due to having
 7      aggregate rather than individual exposure; (2) the difference between the average personal
 8      exposure and the true ambient concentration level; and, (3) the difference between the true and
 9      measured ambient concentration level. By aggregating individual risks to obtain expected
10      number of deaths, they showed that the first component of error (the aggregate rather than
11      individual) is a Berkson error, and, therefore is not a significant contributor to bias in the
12      estimated risk. The second error component is a classical error and can introduce bias if there are
13      short-term associations between indoor source contributions and ambient concentration levels.
14      Recent analysis, however, both using experimental data (Mage et al., 1999; Wilson et al., 2000)
15      and theoretical interpretations and models (Ott et al., 2000) indicate that there is no relationship
16      between the ambient concentration and the nonambient components of personal exposure to PM.
17      However, a bias can arise due to the difference between the personal exposure to ambient PM
18      (indoors plus outdoors) and the  ambient concentration. The third error component is the
19      difference between the true and the measured ambient concentration. According to Zeger et al.
20      the final term is largely of the Berkson type  if the average of the available monitors is an
21      unbiased estimate of the true spatially averaged ambient level.
22           Using this framework, Zeger et al. (2000) then used PTEAM Riverside, CA data to
23      estimate the second error component and its influence on estimated risks. The correlation
24      coefficient between the error (the average population PM10 total exposure minus the ambient
25      PM10 concentration) and the ambient PM10 concentration was estimated to be -0.63. Since this
26      correlation is negative, the flz (the estimated value of the pollution-mortality relative risk in the
27      regression of mortality on z,, the daily ambient concentration) will tend to underestimate the
                   A
28      coefficient fix that would be obtained in the regression of mortality on xt, the daily average total
29      personal exposure, in a single-pollutant analysis.  Zeger et al. (2000) then proceed to assess the
30      size of the bias that will result from this exposure misclassification, using daily ambient
31      concentration, zr As shown in Equation 9, the daily average total personal exposure, xt, can be

        March 2001                               6-250       DRAFT-DO NOT QUOTE OR CITE

-------
  1      separated into a variable component, 0, zp dependent on the daily ambient concentration, z,, and
  2      a constant component, 00, independent of the ambient concentration.
  3
  4                                       xt = 90 + G.z, + et                                 [9]
  5      where £t is an error term.
  6
  7           If the nonambient component of the total personal exposure is independent of the ambient
  8      concentration, as appears to be the case, Equation 9 from Zeger et al. (2000) becomes the
  9      regression analysis equation familiar to exposure analysts (Dockery and Spengler, 1981; Ott
 10      et al., 2000; Wilson et al., 2000). In this case, 00 gives the average nonambient component of the
 11      total personal exposure and 0, gives the ratio of the ambient component of personal exposure to
 12      the ambient concentration. (The ambient component of personal exposure includes exposure to
 13      ambient PM while outdoors and, while indoors, exposure to ambient PM that has infiltrated
 14      indoors.) In this well-known approach to adjust for exposure measurement error, called
                                                                                  i-   yv   ^
 15      regression calibration (Carroll et al., 1995), the estimate of j3x has the simple form f}x = Pz/9\.
 16      Thus, for the regression calibration, the value of fix (based on the total personal exposure) does
 17      not depend on the total personal exposure but is given by /?z, based on the ambient concentration,
 18      times 015 the ratio of the ambient component of personal exposure to the ambient concentration.
 19      A regression analysis of the PTEAM data gave an estimate 0, = 0.60.
20           Zeger et al. (2000) use Equation 9, with 60 = 59.95 and 0, = 0.60, estimated from the
21      PTEAM data, to simulate values of daily average personal exposure, x*t, from the ambient
22      concentrations, zt, for PM10 in Riverside, CA, 1987-1994. They then compare the mean of the
                  vv
23      simulated J3X s, obtained by the series of log-linear regressions of mortality on the simulated x*t,
24      with the normal  approximation of the likelihood function for the coefficient Pz from the
25      log-linear regression of mortality directly on zt.  The resulting /3>z / (3X = 0.59, is very close to
26      0, = 0.60.  Dominici et al.  (2000) provide a more complete analysis of the bias in /3Z as an
27      estimate of f3x using the PTEAM Study and four other data sets and a more complete statistical
                                                          ^             /\
28      model.  Their findings were qualitatively similar in that flx was close to j9z/0,.  Thus, it appears
       March 2001                              6-251        DRAFT-DO NOT QUOTE OR CITE

-------
  1      that the bias is very close to 0, which depends not on the total personal exposure but only on the
  2      ratio of the ambient component of personal exposure to the ambient concentration.
  3           Zeger et al. (2000), in the analyses described above, also suggested that the error due to the
  4      difference between the average personal exposure and the ambient level (the second error type
  5      described above) is likely the largest source of bias in estimated relative risk. This suggestion at
  6      least partly comes from the comparison of PTEAM data and site-to-site correlation (the third type
  7      of error described above) for PM10 and O3 in 8 US cities. While PM10 and O3 both showed
  8      relatively high site-to-site correlation (=0.6-0.9), a similar extent of site-to-site correlation for
  9      other pollutants is not necessarily expected.  Ito et al. (1998) estimated site-to-site correlations
10      (after adjusted for seasonal cycles) for PM10, O3, SO2, NO2, CO, temperature, dewpoint
11      temperature, and relative humidity, using multiple stations' data from seven central and eastern
12      states (IL, IN, MI, OH, PA, WV, WI), and found that, in a geographic scale of less 100 miles,
13      these variables could be categorized into three groups in terms of the extent of correlation:
14      weather variables (r > 0.9); O3, PM10, NO2 (r: 0.6 - 0.8); CO and SO2 (r < 0.5).  These results
15      suggest that the contribution from the third component of error, as described in Zeger et al.
16      (2000), would vary among pollution and weather variables. Furthermore, the contribution from
17      the second component of error would also vary among pollutants; i.e., the ratio of ambient
18      exposure to ambient concentration, called the attenuation coefficient, is expected to be different
19      for each pollutant.  Some of the ongoing studies are expected to shed some light on this issue.
20      However, more information is needed on attenuation coefficients for a variety of pollutants.
21           With regard to the PM exposure,  longitudinal studies (Wallace, 2000; Mage et al., 1999),
22      show reasonably good correlation (r = 0.6 to 0.9) between ambient PM concentrations and
23      average population PM  exposure, lending support for the use of ambient data as a surrogate for
24      personal exposure to ambient PM in time-series mortality or morbidity studies.  Furthermore,
25      fine particles are expected to show even better site-to-site correlation than PM10. Wilson and Sun
26      (1997)  examined site-to-site correlation of PM10, PM2 5, and PM10.2 5 in Philadelphia and
27      St. Louis, and found that site-to-site correlations were high (r = 0.9) for PM2 5 but low for PM10_2 5
28      (r ~  0.4),  indicating that fine particles have  smaller errors in representing community-wide
29      exposures.  This finding supports Lipfert and Wyzga's (1997) speculation that the stronger
30      mortality associations for fine particles than coarse particles found in the Schwartz et al. (1996a)
31      study may be due in part to larger measurement error for coarse particles.

        March 2001                              6-252        DRAFT-DO NOT QUOTE OR CITE

-------
  1           However, as Lipfert and Wyzga (1997) suggested, the issue is not whether the fine particle
  2     association with mortality is a "false positive", but rather, whether the weaker mortality
  3     association with coarse particles is a "false negative". Carrothers and Evans (2000) also
  4     investigated the joint effects of correlation and relative error, but they specifically addressed the
  5     issue of fine (FP) vs. coarse particle (CP) effect, by assuming three levels of relative toxicity of
  6     fine vs. coarse particles (ppp / PCP =1,3, and 10) and, then, evaluating the bias, (B = {E[PF] /
  7     E[PC.]} / (PF/ Pc}, as a function of FP-CP correlation and relative error associated with FP and
  8     CP. Their results indicate: (1) if the FP and CP have the same toxicity, there is no bias (i.e.,
  9     B=l) as long as FP and CP are measured with equal precision, but, if, for example, FP is
 10     measured more precisely than CP, then FP will appear to be more toxic than CP (i.e., B > 1);
 11     (2) when FP is more toxic than CP (i.e., Ppp / PCP = 3 and 10), however, the equal precision of FP
 12     and CP results in downward bias of FP (B < 1), implying a relative overestimation of the less
 13     toxic CP.  That is, to achieve non-bias, FP must be measured more precisely than CP, even more
 14     so as the correlation between FP and CP increases. They also applied this model to real data
 15     from the Harvard Six Cities Study, in particular, the data from Boston and Knoxville. Estimation
 16     of spatial variability for Boston was based on external data and a range of spatial variability for
 17     Knoxville (since there was no spatial data available for this city). For Boston, where the
 18     estimated  FP-CP correlation was low (r = 0.28), estimated error was smaller for FP than for CP
 19     (0.85 vs. 0.65, as correlation between true vs. error-added series), and the observed FP to CP
 20     coefficient ratio was high (11), the calculated FP to CP coefficient ratio was even larger (26)-thus
 21      providing  evidence against the hypothesis that FP is absorbing some of the coefficient of CP.
 22     For Knoxville, where FP-CP correlation was moderate (0.54), the error for FP was smaller than
 23     for CP (0.9 vs. 0.75), and the observed FP to CP coefficient ratio was  1.4, the calculated true FP
 24     to CP coefficient ratio was smaller (0.9) than the observed value, indicating that the coefficient
 25      was overestimated for the better-measured FP, while the  coefficient was underestimated for the
 26      worse-measured CP.  Since the amount (and the direction) of bias depended on several variables
27      (i.e., correlation between FP and CP; the relative error for FP and CP; and, the underlying true
28      ratio of the FP toxicity to CP toxicity), the authors concluded "...for instance, it is inadequate to
29      state that differences in measurement error among fine and coarse particles will lead to false
30      negative findings for coarse particles".


        March 2001                               6-253        DRAFT-DO NOT QUOTE  OR CITE

-------
 1           Fung and Krewski (1999) conducted a simulation study of measurement error adjustment
 2      methods for Poisson models, using scenarios similar to those used in the simulation studies that
 3      investigated implication of joint effects of correlated covariates with measurement error. The
 4      measurement error adjustment methods employed were the Regression Calibration (RCAL)
 5      method (Carroll et al., 1995) and the Simulation Extrapolation (SIMEX) method (Cook and
 6      Stefanski, 1994). Briefly, RCAL algorithm consists of: (1) estimation of the regression of X on
 7      W (observed version of X, with error) and Z (covariate without error); (2) replacement of X by
 8      its estimate from (1), and conducting the standard analysis (i.e., regression); and (3) adjustment
 9      of the resulting standard error of coefficient to account for the calibration modeling.  SIMEX
10      algorithm consists of: (1) addition of successively larger amount of error to the original data;
11      (2) obtaining naive regression coefficients for each of the error added data sets; and, (3) back
12      extrapolation of the obtained coefficients to the error-free case using a quadratic or other
13      function.  Fung and Krewski examined the cases for: (1) Px = 0.25; Pz = 0.25; (2) Px = 0.0;
14      pz = 0.25; (3) px = 0.25; pz = O.O., all with varying level of correlation (-0.8 to 0.8) with and
15      without classical additive error, and also considering Berkson type error.  The behaviors of naive
16      estimates were essentially similar to other simulation studies. In most cases with the classical
17      error, RCAL performed better than SIMEX (which performed comparably when X-Z correlation
18      was small), recovering underlying coefficients. In the presence of Berkson type error, however,
19      even RCAL did not recover the underlying coefficients when X-Z correlation was large (> 0.5).
20      This is the first study to examine the performance of available error adjustment methods that can
21      be applied to time-series Poisson regression. The authors recommend RCAL over SIMEX.
22      Possible reasons why RCAL performed better  than SIMEX in these scenarios were not discussed,
23      nor are they clear from the information given in the publication.  There has not been a study to
24      apply these error adjustment methods in real time-series health effects studies.  These
25      methodologies require either replicate measurements or some knowledge on the nature of error
26      (i.e., distributional properties, correlation, etc.). Since the information regarding the nature of
27      error is still being collected at this time, it may take some time before applications of these
28      methods become practical.
29           Another issue that measurement error may affect is the detection of threshold in time-series
30      studies. Lipfert and Wyzga (1996) suggested that measurement error may obscure the true shape
31      of the exposure-response curve, and that such error could make the exposure-response curve to

        March 2001                              6-254       DRAFT-DO NOT  QUOTE OR CITE

-------
  1      appear linear even when a threshold may exist.  However, based on a simulation with realistic
  2      range of exposure error (due to site-to-site correlation), Cakmak et al. (1999) illustrated that the
  3      modern smoothing approach, LOESS, can adequately detect threshold levels (12.8 ,ug/m3,
  4      24.6 Aig/m3, and 34.4 /^g/m3) even with the presence of exposure error (see also Section 6.4.6
  5      above).
  6           Other issues related to exposure error that have not been investigated include potential
  7      differential error among subpopulations.  If the exposure errors are different between susceptible
  8      population groups (e.g., people with COPD) and the rest of the population, the estimation of bias
  9      may need to take such differences into account.  Also, the exposure errors may vary from season
 10      to season, due to seasonal differences in the use of indoor emission sources and air exchange
 11      rates due to air conditioning and heating. This may possibly explain reported season-specific
 12      effects of PM and other pollutants.  Such season-specific contributions of errors from indoor and
 13      outdoor sources are also expected to be different from pollutant to pollutant.
 14           In summary, the studies that examined joint effects of correlation and error suggest that PM
 15      effects are likely underestimated, and that spurious PM effects (i.e., qualitative bias such as
 16      change in the sign of coefficient) due to transferring of effects from other covariates require
 17      extreme conditions and are, therefore, unlikely.  Also, one simulation study suggests that, under
 18      the likely range of error for PM, it is unlikely that a threshold is ignored by common smoothing
 19      methods. More data are needed to examine the exposure errors for other pollutants, since their
 20      relative error contributions will influence their relative significance in relative risk estimates.
 21
 22      6.4.8  New Assessment of Methodological Issues
 23      6.4.8.1  Time Series Model Specification
 24           Methodological issues in time-series analyses of air pollution-mortality association were
25      discussed extensively in the 1996 PM AQCD. Since then, increasing numbers of researchers
26      have been utilizing essentially the same Poisson regression approach: (1) model seasonal cycles
27      and other temporal trends using smoothing functions of time; (2) model weather effects using
28      smoothing functions of temperature, humidity, and/or their interaction at various lags; (3) after
29      adjustment for these confounding factors, enter various lags (and averaging periods) of air
30      pollutant, and report results for all the lags, and/or report results for the lags that resulted in the
31      highest significance; (4) repeat (3) with other pollutants in the model; (5) conduct sensitivity
        March 2001                               6-255        DRAFT-DO NOT QUOTE OR CITE

-------
 1      analyses using alternative weather model specifications.  Seasonal cycles and weather effects are
 2      often modeled using Generalized Additive Models (GAM).  As the modeling of temporal trends
 3      became more efficient using the GAM models, it became clearer that the residual over-dispersion
 4      and autocorrelation could be essentially eliminated.  Also, more researchers appear to rely on
 5      Akaike's Information Criteria (AIC) or on the more conservative Bayes (Schwarz, 1978)
 6      Information Criterion (BIC) to choose between models when epidemiological reasoning does not
 7      favor one over the other. While these techniques do not necessarily eliminate inadequate model
 8      specifications, they do help "standardize" the approaches that researchers can take, reducing the
 9      inconsistency in model specification among studies.
10           A few remaining inconsistencies in approach among studies include: (1) choice of the
11      range of lags and averaging periods of pollution included; (2) smoothing spans used for modeling
12      temporal trends and weather effects; (3) the increment used to calculate relative risks; and,
13      (4) choice to detrend pollution variables.  The choice of lag can lead to inconsistent results even
14      for the  same data. The choice of the combination of lags multiplies as the number of
15      co-pollutants in the model increases. In the case of temperature effects, it has been repeatedly
16      observed that the heat effects tend to be immediate (0 or 1 day lag), while cold effects tend to lag
17      longer (2 to 4 days). For pollutants, however, reported lags are less  consistent. The smoothing
18      span for temporal trends can be determined based on epidemiological reasons (i.e., eliminate
19      influenza epidemics), but the span for weather effects may be determined through data
20      exploration.  Using the inter-quartile range for all the co-pollutants may be problematic when
21      co-pollutants have inconsistent distributional characteristics.  While these issues may appear
22      rather minor, in practice, they may make substantial differences in reported effects and
23      interpretations.
24
25      6.4.8.2  Case-Crossover Study Design
26          Navidi et al. (1999) proposed the use of "bi-directional" controls in applying the case-
27      crossover design to study acute effects of air pollution. In the original case-crossover studies in
28      which risk factors were behavior-related (e.g., coffee consumption), the control period was
29      chosen prior to the case period (i.e., retrospective uni-directional) because choosing the control
30      period after the event would interfere with behavioral modification associated with the risk
31      factor, possibly resulting in bias. In the case of environmental exposures such as ambient air

        March 2001                               6-256        DRAFT-DO NOT QUOTE OR CITE

-------
  1     pollution, however, the event is unlikely to modify future exposure. Furthermore, in the case of
  2     observational air pollution study, the bi-directional control periods would be necessary to avoid
  3     confounding due to temporal trends in both events (e.g., influenza-related mortality or morbidity)
  4     and exposure (natural seasonal trends). Navidi conducted simulations to illustrate that the
  5     relative risk estimates are resistant to confounding by time-trend.
  6          Bateson and Schwartz (1999) also conducted a simulation study to compare five case-
  7     crossover control sampling strategies including the matched pair, a symmetric bi-directional, a
  8     total history approach, and the two approaches that Navidi proposed. The symmetric
  9     bi-directional approach using 1-week lag estimated the true relative risks correctly in the
 10     presence of confounding seasonal trends, whereas the other four approaches failed to control for
 11     the confounding trends. They concluded that the bi-directional case-crossover design could
 12     control for confounding by design, though it is not as efficient as Poisson time-series analysis.
 13          There have been several studies that applied the case-crossover design to analyze air
 14     pollution - mortality associations, as described below.
 15          Neas et. al. (1999) analyzed Philadelphia TSP data for 1973-1980. Total, age over 65,
 16     cancer, and cardiovascular deaths were analyzed for their association with TSP.  A conditional
 17     logistic regression analysis with a case-crossover design was conducted using the control periods
 18     of 7, 14, and 21 days before and after the case period. Other covariates included temperature on
 19     the previous day, dewpoint on the same day, an indicator for hot days (> 80°F), an indicator for
 20     humid days (dewpoint > 66°F), and interaction between the same-day temperature and winter
 21     season. In each set of the six control periods, TSP was associated with total mortality. A model
 22     with four symmetric reference periods 7 and 14 days around the case period produced a similar
 23     result. A model with  only two symmetric reference periods of 7 days around the case produced a
 24     larger estimate.  A larger effect was seen for deaths in persons > 65 years of age and for deaths
25     due to pneumonia and to cardiovascular disease.  Thus, this study basically confirmed the
26     original findings by Schwartz and Dockery (1992) for this city.
27          Sunyer et al. (2000)  analyzed Barcelona, Spain BS data for 1990-1995.  Those who were
28     over age 35 and had sought emergency room services for COPD exacerbation between 1985 and
29     1989, and had died during 1990-1995 were included in analysis. Total, respiratory, and
30     cardiovascular deaths  were analyzed using a conditional logistic regression analysis with a case-
31      crossover design, adjusting for temperature, relative humidity, and influenza epidemics.

        March 2001                               6-257       DRAFT-DO NOT QUOTE OR CITE

-------
 1      Bi-directional control period at 7 days was used. The average of the same and previous 2 days
 2      was used for pollution exposure period. Data were also stratified by potential effect modifiers
 3      (e.g., age, gender, severity of ER visits, number of ER visits, etc.) and were analyzed. BS levels
 4      were associated with all cause deaths. The association was stronger for respiratory causes.  Older
 5      women, patients admitted to intensive care units, and patients with a higher rate of ER visits were
 6      at greater risk of deaths associated with BS.
 7           Lee and Schwartz (1999) analyzed data from Seoul, Korea for 1991 -1995. Total deaths
 8      were analyzed for their association with TSP, SO2, and O3.  A conditional logistic regression
 9      analysis with a case-crossover design was conducted. Three-day moving average values (current
10      and two past days) of TSP and SO2, and 1-hr max O3 were analyzed separately. The control
11      periods are 7 and 14 days before and/or after the case period.  Both unidirectional and
12      bi-directional controls (7 or 7 and 14  days) were examined, resulting in six sets of control
13      selection schemes.  Other covariates included temperature and relative humidity.  Among the six
14      control periods, the two unidirectional retrospective control schemes resulted in odds ratios less
15      than 1; the two unidirectional prospective control schemes resulted in larger odds ratios (e.g.,
16      1.4 for 50 ppb increase in SO2); and bi-directional control schemes resulted in odds ratios
17      between those for uni-directional schemes.  SO2 was more significantly associated with mortality
18      than TSP. These results suggested that risk estimates were rather sensitive to the choice of the
19      control periods.
20           These analyses suggest that the  overall findings are not very sensitive to these analytic
21      choices; thus we can have more confidence in the mortality results. The sensitivity analyses are
22      not as extensive for examining the PM10 effect on morbidity, and the investigators used a
23      different time window across the 14 cities to control for temporal effects. Future analyses of
24      both the mortality and morbidity data might include a seasonally stratified analysis (given the
25      seasonal variability in pollutant concentrations, outcome measures, and potential confounding
26      factors). Loss of statistical power due to the shorter periods of observation in any season should
27      be only a minor issue, at least in the mortality data  set.
28
29      6.4.9  Heterogeneity of Particulate Matter Effects Estimates
30           Approximately 35 then-available acute PM exposure community epidemiologic studies
31      were assessed in the 1996 PM  AQCD as collectively demonstrating increased risks of mortality
        March 2001                               6-258        DRAFT-DO NOT QUOTE OR CITE

-------
  1     being associated with short-term (24-h) PM exposures indexed by various ambient PM
  2     measurement indices (e.g., PM10, PM2 5, BS, COH, sulfates, etc.) in many different cities in the
  3     United States and internationally.  Much homogeneity appeared to exist across various
  4     geographic locations, with many studies suggesting, for example, increased relative risk (RR)
  5     estimates for total nonaccidental mortality on the order of 1.025 to 1.05 (or 2.5 to 5.0% excess
  6     deaths) per 50 yUg/m3 increase in 24-h PM10, with statistically significant results extending more
  7     broadly in the range of 1.5 to 8.0%. The elderly >65 yrs. old and those with preexisting
  8     cardiopulmonary conditions had somewhat higher excess risks. One study, the Harvard Six City
  9     Study, also provided estimates of increased RR for total mortality falling in the range of 1.02 to
 10     1.056 (2.0 to 5.6% excess deaths) per 25 ywg/m3 24-h PM2 5 increment.
 11          Now, more than 70 new time-series PM-mortality studies assessed earlier in this chapter
 12     provide extensive additional evidence which, qualitatively, largely substantiates significant
 13     ambient PM-mortality relationships, again based on 24-h exposures indexed by a wide variety of
 14     PM metrics in many different cities of the United States, in Canada, in Mexico, and elsewhere (in
 15     South America, Europe, Asia, etc.). The newly available effect size estimates from such studies
 16     are reasonably consistent with the ranges derived from the earlier studies reviewed in the 1996
 17     PM AQCD. For example, newly estimated PM,0 effects generally fall in the range of 1.0 to 8.0%
 18     excess deaths per 50 Aig/m3 PM10 increment in 24-h concentration; whereas new PM2 5 excess
 19     estimates for short-term exposures generally fall in the range of 2 to 8% per 25 Mg/m3 increment
 20     in 24-h PM2 5 concentration.
 21           However, somewhat greater spatial heterogeneity appears to exist across newly reported
 22     study results, both with regard to PM-mortality and morbidity effects. The newly apparent
 23      heterogeneity of findings across  locations is perhaps most notable in relation to reports based on
 24     multiple-city studies in which investigators used the same analytical strategies and models
 25      adjusted for the same or similar co-pollutants and meteorological conditions, raising the
 26      possibility of different findings reflecting real location-specific differences in exposure-response
27      relationships rather than potential differences in models used, pollutants measured and included
28      in the models, etc.  Some examples of newly reported and well-conducted multiple-city studies
29      include: the NMMAPS analyses  of mortality and morbidity in 20 and 90 U.S. cities (Samet et al.,
30      2000a,b; Dominici et al., 2000); the Schwartz (2000b,c) analyses of 10 U.S. cities; the study of
31      eight largest Canadian cities (Burnett et al., 2000); the study of hospital admissions in eight U.S.

        March 2001                               6-259        DRAFT-DO NOT QUOTE OR CITE

-------
  1      counties (Schwartz, 1999); and the APHEA studies of mortality and morbidity in several
  2      European cities (Katsouyanni et al., 1997; Zmirou et al., 1998). The recently completed large
  3      NMMAPS studies of morbidity and mortality in U.S. cities add especially useful and important
  4      information about potential U.S. within- and between-region heterogeneity.
  5
  6      6.4.9.1 Evaluation of Heterogeneity of Participate Matter Mortality Effect Estimates
  7           In all of the U.S. multi-city analyses, the heterogeneity in the PM estimates across cities
  8      was not explained by city-specific characteristics in the 2nd stage model. The heterogeneity of
  9      effects estimates across cities in the multi-city analyses may be due to chance alone, to
10      mis-specification of covariate effects in small cities, or to real differences from location to
11      location in effects of different location-specific ambient PM mixes, for which no mechanistic
12      explanations are yet known. Or, the apparent heterogeneity may simply reflect imprecise PM
13      effect estimates derived from smaller-sized analyses of less extensive available air pollution data
14      or numbers of deaths in some cities tending to obscure more precise effects estimates from
15      larger-size analyses for other locations, which tend to be consistently more positive and
16      statistically significant.
17           Some of these possibilities can be evaluated by using data from the NMMAPS study
18      (Samet et al., 2000b).  Data in Figure 6-1 were optically scanned and digitized, producing
19      reasonably accurate estimates by comparison with the 20 largest U.S. cities in their Table A-2.
20      The cities were divided among  7 regions, and excess risk with 95% confidence intervals plotted
21      against the total number of effective observations, measured by the number of days of PM10 data
22      times the mean number of daily deaths in the community. This provides a useful measure of the
23      weight that might be assigned to the results, since the uncertainty of the RR estimate based on a
24      Poisson mean is roughly inversely proportional to this product.  That is, the expected pattern
25      typically shows less spread of estimated excess risk with increasing death-days of data.  A more
26      refined weight index would also include the spread in the distribution of PM concentrations.  The
27      results are plotted in Figure 6-12 for all cities and Figure 6-13 for each of the 7 regions.
28           Figure 6-12 for all cities suggests some relationship between precision of the effects
29      estimates and study weight, overall. That is, the more the mortality-days observations, the
30      narrower the 95% confidence intervals and the more precise the effects estimates (with nearly all
31

        March 2001                               6-260        DRAFT-DO NOT QUOTE OR CITE

-------
                                             All  Cities
            6      7      8      9     10     11     12     13
            Natural  Log  of Mortality -  Days
Figure 6-12. The EPA-derived plot showing relationship of PMi0 total mortality effects
          estimates and 95% confidence intervals for all cities in the Samet et al.
          (2000a,b) NMMAPS 90-cities analyses in relation to study size (i.e., the
          natural logarithm of numbers of deaths times days of PM observations). Note
          generally narrower confidence intervals for more homogeneously positive
          effects estimates as study size increases beyond about the log 9 value (i.e.,
          beyond about 8,000 deaths-days of observation). The dashed line depicts the
          overall nationwide effect estimate (grand mean) of approximately 0.5% per
          10 (tg/m3 PM10.
March 2001
6-261
DRAFT-DO NOT QUOTE OR CITE

-------
    5.4
Q.
1
0. 00
3?
»" -1.8
ct
to
U)
1
.

-'






I 1
Northwest
r I ~"Tr T t
I IF J
i i



•

-


- J
•
j.
•
Industrial Midwest
iflllTT '
• •• | ij-fTi"I""3r
1 •LIi
                                                                              Northeast
                                                                                   I.
       7      8      9     10     11  7   8   9   10  11   12  13  6   7   8   9   10  11  12
         Natural Log of Mortality - Days      Natura| Log Qf Morta|jty. Dgys      Natura| Log Qf Morta|i(y. Dgys
o «•**
0.

o>
a 1.8
o
o> n n
£
to
(A
R
X
"-1 C 4
Southern California


'II I
11
-
[II
1
-



















-

- .
i i i i i
Other Midwest



















i i





'
7
-




i i














-

J-





i i i
Southeast
r

: •
_





i_


_

-LI


• • ,
n
••

I]
1

"
-JH
tl .




i i
       8      9     10     11     12 6   7   8   9   10  11   12  7      8     9     10    11
        Natural Log of Mortality - Days      Natura| Log of Morta|ity. Days      Natura| Log of Morta|ity. Days
    5.4
 :>
 Q.
  3.
 O
 Of.
 to
1.8 -

-
-
1 1






Southwest


1 1
i-i
i i

     '8.0    8.5    9.0    9.5    10.0
         Natural Log of Mortality - Days
Figure 6-13. The EPA-derived plots showing relationships of PMi0-mortality (total,
             nonaccidental) effects estimates and 95% confidence intervals to study size
             (defined as in Figure 6-10) for cities broken out by regions as per the
             NMMAPS regional analyses of Samet et al. (2000a,b). Dashed line on each
             plate depicts overall nationwide effect estimate (grand mean) of
             approximately 0.5% per 10 /ug/m3 PM,
                                                 MO'
March 2001
                                      6-262
DRAFT-DO NOT QUOTE OR CITE

-------
  1      these for cities with > log 9 mortality-days being positive and many statistically significant at
  2      p <0.05).
  3           The Figure 6-13 depiction for each of the 7 regions is also informative. In the Northeast,
  4      there is considerable homogeneity (not heterogeneity) of effect size for larger study-size cities,
  5      even with moderately wide confidence intervals for those with log mortality-days = 8 to 9, and all
  6      clearly exceed the overall nationwide grand mean indicated by the dashed line. On the other
  7      hand, the smaller study-size Northeast cities (with much wider confidence intervals at log < 8)
  8      show much greater heterogeneity of effects estimates and less precision.  Also, most of the
  9      estimates for larger study-size (log > 9) cities in the industrial midwest are positive and several
 10      statistically significant, so that an overall significant regional risk is plausible there as well.
 11      There may even be some tendency for relatively large risks for some cities with small study sizes
 12      and wide confidence intervals in the industrial midwest, and further investigation of that would
 13      be of interest. The plot for Southern California in Figure 6-13 clearly shows a rather consistent
 14      estimate of effect size and width of the confidence intervals across cities of varying study-size.
 15      All risk estimates are positive and most are significant at p s  0.05 or nearly so for the Southern
 16      California cities.  For Northwestern cities plotted in Figure 6-13, the value for Oakland, CA (at
 17      ca. log 9.5) is notable (it being very positive and significant), whereas many but not all of the
 18      other cities have positive effect estimates not too far off the nationwide grand mean, but with
 19      sufficiently wide  confidence intervals so as not to be statistically significant at p < 0.05. The
 20      Southwestern cities (except for 2 cities), too, mostly appear to have effect sizes near the
 21       nationwide mean, but with confidence intervals too wide to be significant at p < 0.05. The
 22      "Other" (non-industrial or "Upper", as per NMMAPS) Midwest cities and the Southeastern cities
 23       in Figure 6-13 show more heterogeneity, although most of the larger study size cities (log > 9.0)
 24       tend to be positive and not far off the nationwide mean (even though not significant at p < 0.05).
 25       Given the wide range of effects estimates and confidence intervals seen for Southeastern cities,
 26      further splitting of the region might be informative.
27           In fact, closer reexamination of results for each of the regions may reveal interesting new
28      insights into what factors may account for any apparent  disparities among the cities within a
29      given region or across regions. Several possibilities readily come to mind. First, cursory
30      inspection of the mean PM10 levels shown for each city in Appendix 6A-2 suggests that many of
31      the cities showing low effects  estimates and wide confidence intervals tend to be among those

        March 2001                                6-263        DRAFT-DO NOT QUOTE OR  CITE

-------
 1      having the lowest mean PM10 levels and, therefore, likely the smallest range of PM10 values
 2      across which to distinguish any PM-related effect, if present. It may also be possible that those
 3      areas with higher PM2 5 proportions of PM10 mass (i.e., larger percentages of fine particles) may
 4      show higher effects estimates (e.g., in Northeastern cities) than those with higher coarse-mode
 5      fractions (e.g., as would be more typical of Southwestern cities).  Also, more industrialized cities
 6      with greater fine-particle emissions from coal combustion (e.g., in the industrial Midwest) and/or
 7      those with high fine-particle emissions from heavy motor vehicle emissions (e.g., typical of
 8      Southern California cities) may show larger PM10 effects estimates than other cities. Lastly, the
 9      extent of air-conditioning use may also account for some of the differences, with greater use in
10      many Southeastern and Southwestern cities perhaps decreasing actual human exposure to
11      ambient particles present versus higher personal exposure to ambient PM (including indoors) in
12      those areas where less air-conditioning is used (e.g., the Northeast and industrial Midwest).
13
14      6.4.9.2  Comparison of Spatial Relationships in the NMMAPS and Cohort
15              Reanalyses Studies
16           Both the NMMAPS and HEI Cohort  Reanalyses studies had a sufficiently large number of
17      U.S. cities to allow considerable resolution of regional PM effects within the "lower 48" states,
18      but an attempt was made to take this approach to a much more detailed level in the Cohort
19      Reanalysis studies than in NMMAPS. There were:  88 cities with PM10 effect  size estimates in
20      NMMAPS; 50 cities with PM25 and 151 cities with sulfates in the original Pope et al. (1995)
21      ACS analyses and in the HEI reanalyses using the original data; and 63 cities with PM2 5 data and
22      144 cities with sulfate data in the additional analyses done by the HEI Cohort Reanalysis team.
23      The relatively large number of data points utilized in the HEI reanalyses effort and additional
24      analyses allowed estimation of surfaces for elevated long-term concentrations of PM2 5, sulfates,
25      and SO2 with resolution on a scale of a few tens to hundreds of kilometers.
26          The patterns for PM2 5 and sulfates are similar, but not identical, hi particular, the modeled
27      PM2 5 surface (Krewski et al., 2000; Figure 18) has peak levels around Chicago - Gary, in the
28      eastern Kentucky - Cleveland region, and around Birmingham AL, with elevated but lower PM2 5
29      almost everywhere east of the Mississippi,  as well as southern California. This is similar to the
30      modeled sulfate surface (Krewski et al., 2000; Figure 16), with the absence of a peak in
31      Birmingham and an emerging sulfate peak in Atlanta.  The only area with markedly elevated SO2

        March 2001                               6-264       DRAFT-DO NOT QUOTE OR CITE

-------
  1     concentrations is the Cleveland - Pittsburgh region. A preliminary evaluation is that secondary
  2     sulfates in particles derived from local SO2 are more likely to be important in the industrial
  3     midwest, south from the Chicago - Gary region into Ohio, northeastern Kentucky, West Virginia,
  4     and southwest Pennsylvania, possibly related to combustion of high-sulfur fuels.
  5          The overlay of mortality with air pollution patterns is also of much interest.  The spatial
  6     overlay of long-term PM2 5 and mortality (Krewski et al., 2000; Figure 21) is highest from
  7     southern Ohio to northeastern Kentucky/West Virginia, but also includes a significant association
  8     over most of the industrial midwest from Illinois to the eastern non-coastal parts of North
  9     Carolina, Virginia, Pennsylvania, and New York. This is reflected, in diminished form, by the
 10     sulfates and SO2 maps (Krewski et al., 2000; Figures 19 and 20), where there appears to be a
 11     somewhat tighter focus of elevated risk in the upper Ohio River Valley area.  This suggests that,
 12     while SO2 may be an important precursor of sulfates  in this region, there may also be some other
 13     (non-sulfur) contributors to associations between PM2 5 and long-term mortality, embracing a
 14     wide area of the Northcentral Midwest and non-coastal Mid-Atlantic region.
 15          It should be noticed that, while a variety of spatial modeling  approaches were discussed in
 16     the NMMAPS methodology report (NMMAPS Part I, pp. 66-71 [Samet et al., 2000aj),  the
 17     primary spatial analyses in the  90-city study (NMMAPS, Part II [Samet et al., 2000b]) were
 18     based on a simpler seven-region breakdown of the contiguous 48 states. The 20-city results
 19     reported for the spatial model in NMMAPS I show a much smaller posterior probability of a
 20     PM10 excess risk of short-term mortality, with a spatial posterior probability vs.  a non-spatial
 21     probability of a PM10 effect of 0.89 vs. 0.98 at lag 0, of 0.92 vs. 0.99 at lag 1, and of 0.85 vs. 0.97
 22     at lag 2. The evidence that PM10 is associated with an excess short-term mortality risk is still
 23     moderately strong with a spatial model, but less strong than with a non-spatial model.
 24          Even so, there is a considerable degree of coherence between the short-term and long-term
 25     mortality findings of the two studies, with strong evidence of a modest but significant short-term
26     PMIO effect and a large long-term fine particle (PM2 5 in general or sulfate) effect in the industrial
27     Midwest. The short-term PM10 effects are large in the Northeast and in Southern California
28      (though less certain there), whereas long-term PM2 5 effects seem to be moderate to high in these
29      areas as well.  This may tend to suggest that at least some of the more notable PM10 effects  found
30      in the NMMAPS regional analyses may coincide with the presence of higher proportions of fine
31      versus coarse particles in the PM10 mix.

        March 2001                              6-265        DRAFT-DO NOT QUOTE  OR CITE

-------
 1           The apparently substantial differences in PM10 and/or PM2 5 effect sizes across different
 2      regions should not be attributed merely to possible variations in measurement error or other
 3      statistical artifact(s).  Some of these differences may reflect: real regional differences in particle
 4      composition or co-pollutant mix; differences in relative human exposures to ambient particles or
 5      other gaseous pollutants; sociodemographic differences (e.g., percent of infants or elderly in
 6      regional population); or other important, as of yet unidentified PM effect modifiers.
 7
 8
 9      6.5 KEY FINDINGS AND CONCLUSIONS DERIVED FROM
10          PARTICULATE MATTER EPIDEMIOLOGY STUDIES
11           It is not possible to assign any absolute measure of certainty to conclusions based on the
12      findings of the epidemiology studies discussed in this chapter. However, these observational
13      study findings would be further enhanced by supportive findings of causal studies from other
14      scientific disciplines (dosimetry, toxicology, etc.), as discussed in Chapters 7 to 9.  The most
15      salient conclusions derived from the PM epidemiology studies include:
16      (1)   A very large and sufficiently convincing body of epidemiology evidence substantiates
17           strong associations between short- and long-term ambient PM10 exposures (inferred from
18           stationary air monitor measures) and mortality/morbidity effects to conclude that PM10 (or
19           one or more PM10 components) is a probable contributory cause of human health effects.
20      (2)   It is likely that there is meaningful heterogeneity in the city-specific excess risk estimates
21           for the relationships between short-term ambient PM10 concentrations and acute health
22           effects.  The reasons for such variation in effects estimates are not well understood at this
23           time, but do not negate ambient PM's likely causative contribution to observed PM-
24           mortality and/or morbidity associations in many locations.
25      (3)   A smaller (but growing) body of epidemiology evidence is sufficiently indicative of
26           associations between short- and long-term ambient PM2 5 exposures (inferred from
27           stationary air monitor measures) and health effects to conclude that PM2 5 (or one or more
28           PM2 5 components) is a probable contributing cause of observed PM-associated health
29           effects.  Some new epidemiology findings also suggest that health effects are associated
30           with mass or number concentrations of ultrafine (nuclei-mode) particles, but not necessarily
31           more so than for other ambient fine PM components.

        March 2001                               6-266        DRAFT-DO NOT QUOTE OR CITE

-------
   1      (4)   An even smaller body of evidence exists which appears to support an association between
  2           short-term ambient coarse-fraction (PM10_2 5) exposures (inferred from stationary air
  3           monitor measures) and short-term health effects in epidemiology studies. This suggests
  4           that PM10_2 5, or some constituent component(s) of PM10_2 5, may be a contributory cause of
  5           health effects in some locations. Reasons for differences among findings on coarse-particle
  6           health effects reported for different cities are still poorly understood, but several of the
  7           locations where significant PM10.2 5 effects have been observed (Phoenix, Mexico City,
  8           Santiago) tend  to be in drier climates and may have contributions to observed effects due to
  9           higher levels of organic particles from biogenic processes (endotoxins, molds, etc.) during
 10           warm months.  Other studies suggest that coarse  fraction (PM10.2 5) particles of crustal
 11           origin are unlikely to exert notable health effects under most ambient exposure conditions.
 12      (5)   Long-term PM exposure durations, on the order of months to years, as well as  on the order
 13           of a few days, are likely associated with serious human health effects (indexed by mortality,
 14           hospital admissions/medical visits, etc.). More chronic PM exposures, on the order of
 15           years or decades, appear to be associated with life shortening beyond that accounted for by
 16           the simple accumulation of the more acute effects of short-term PM exposures (on the order
 17           of a few days).  While the few studies of this relationship were generally well conducted,
 18           notable uncertainties remain regarding the meaning, magnitude, and mechanisms for more
 19           chronic health effects of long-term PM exposures. New findings of associations between
 20           ambient PM exposures (indexed by various measures) during early pregnancy and/or early
 21            post-natally and slowed fetal growth or infant mortality, respectively, suggest potentially
 22           much larger life-shortening impacts of PM than previously estimated.
 23      (6)   Considerable coherence exists among effect size estimates for ambient PM health effects.
 24            For example, results derived from several multi-city studies, based on pooled analyses of
 25           data combined across multiple cities (thought to yield the most precise effect size
 26           estimates), show the percent excess total (non-accidental) deaths estimated per 50 /ug/m3
 27           increase in 24-h PM10 to be:  2.3% in the 90 largest U.S. cities (4.5% in the Northeast U.S.
 28           region); 3.4% in 10 U.S. cities;  3.5% in the 8 largest Canadian cities; and 2.0% in western
 29           European cities  (using PM10 = TSP*0.55).  These combined estimates are consistent with
30           the range of PM,0 estimates previously reported in the 1996 PM AQCD. These and excess
31           risk estimates from many other individual-city studies, generally falling in the range of ca.

        March 2001                               6-267        DRAFT-DO NOT QUOTE OR CITE

-------
 1           1.5 to 8.0% per 50 /ug/m3 24-h PM10 increment, also comport well with numerous new
 2           studies confirming increased cause-specific cardiovascular- and respiratory-related
 3           mortality. They are also coherent with larger effect sizes reported for cardiovascular (in the
 4           range of ca. 3.0 to 10.0% per 50 Aig/m3 24-h PM10 increment) and respiratory (in the range
 5           of ca. 5 to 25% per 50 /wg/m3 24-h PM10) hospital admissions/visits, as would be expected
 6           for these morbidity endpoints versus those for PM10-related mortality.
 7      (7)   Several independent panel studies (but not all) that evaluated temporal associations
 8           between PM exposures and measures  of heart beat rhythm in elderly subjects provide
 9           generally consistent indications of decreased heart rate variability (HRV) being associated
10           with ambient PM exposure (decreased HRV being an indicator of increased risk for serious
11           cardiovascular outcomes, e.g., heart attacks). Other studies point toward changes in blood
12           characteristics (e.g., C-reactive protein levels) related to increased risk of ischemic heart
13           disease also being associated with ambient PM exposures. However, these heart rhythm
14           and blood characteristics findings should currently be viewed as providing only limited or
15           preliminary support for PM-related cardiovascular effects.
16      (8)   Notable new evidence now exists which substantiates positive associations between
17           ambient PM concentrations and  increased respiratory-related hospital admissions,
18           emergency department, and other medical visits, particularly in relation to PM10 levels.
19           Of much interest are new, but limited, findings tending to implicate not only fine particle
20           components but also coarse (e.g., PMIO_2 5) particles as likely contributing to exacerbation of
21           asthma conditions.  Also of much interest are emerging new findings indicative of likely
22           increased occurrence of chronic  bronchitis in association with (especially chronic) PM
23           exposure.
24      (9)   One major methodological issue affecting epidemiology studies of both short-term and
25           long-term PM exposure effects is that ambient PM of varying size ranges is typically found
26           in association with other air pollutants, including gaseous criteria pollutants (e.g, O3, NO2,
27           SO2, CO), air toxics, and/or bioaerosols. Available statistical methods for assessing
28           potential confounding arising from these associations may not yet be fully adequate.  The
29           inclusion of multiple pollutants often  produces statistically unstable  estimates.  Omission of
30           other pollutants may incorrectly attribute their independent effects to PM. Much progress
31           in sorting out relative contributions of ambient PM components versus other copollutants is

        March 2001                               6-268       DRAFT-DO NOT QUOTE OR CITE

-------
  1          nevertheless being made and, overall, tends to substantiate that observed PM effects are at
  2          least partly due to ambient PM acting alone or in the presence of other covarying gaseous
  3          pollutants.
  4     (10) It is likely that differences in observed health effects will be found to depend as much on
  5          site-specific differences in chemical and physical composition characteristics of ambient
  6          particles as  on differences in PM mass concentration. For example, the Utah Valley study
  7          (Dockery et al., 1999; Pope et al., 1991,  1999b) showed that PMIO particles, known to be
  8          richer in metals during exposure periods while the steel mill was operating, were more
  9          highly associated with adverse health effects than was PM10 during the PM exposure
 10          reduction while the steel mill was closed. In contrast, PM10 or PM2 5 was relatively higher
 11          in crustal particles during windblown dust episodes in Spokane and in three central Utah
 12          sites than at other times, but was not  associated with higher total mortality. These
 13          differences require more research that may become more feasible as the PM2 5 sampling
 14          network produces air quality data  related to speciated samples.
 15     (11) The above reasons suggest it is inadvisable to pool epidemiology studies at different
 16          locations, different time periods, with different population sub-groups, or different health
 17          endpoints, without assessing the consequences of these differences. Published multi-city
 18          analyses using common data bases, measurement devices, and analytical strategies such as
 19          those carried out in the APHEA and NMMAPS studies are likely to be useful after careful
 20          evaluation.  Pooled analyses of more  diverse collections of independent studies of different
 21          cities, using varying methodology and/or data quality or representativeness, are likely less
 22          credible and should not, in general, be used without careful assessment of their underlying
 23          scientific comparability.
 24     (12) It may be possible that different PM components may produce effects which appear at
25          different lags or that different preexisting conditions may lead to different delays between
26          exposure and effect. Thus, although maximum effect sizes for PM effects have often been
27          reported for 0-1 day lags, evidence is  also beginning to suggest that more consideration
28          should be given to lags of several days.  Also, if it is considered that all health effects
29          occurring at different lag days are  all real effects, so that the risks for each lag day should
30          be additive, then higher overall risks may exist than implied by maximum estimates for any
31           particular single or two-day lags.

        March 2001                               6-269       DRAFT-DO NOT QUOTE OR CITE

-------
 1      (13)  Certain classes of ambient particles may be distinctly less toxic than others and may not
 2           exert human health effects at typical ambient exposure concentrations or only under special
 3           circumstances. For example, particles of crustal origin may be relatively non-toxic under
 4           most circumstances compared to those of combustion origin. However, crustal particles
 5           contaminated with pesticides or herbicides (as may occur in agricultural situations) or with
 6           emissions from vehicles, smelters, or other industrial operations may be sufficiently toxic
 7           to cause human health effects under some exposure conditions.  More research is needed to
 8           identify conditions under which one or another class of particles cause little or no adverse
 9           health effects, as well as conditions under which particles cause  notable effects.
10      (14)  Certain epidemiology evidence suggests that reducing ambient PM,0 concentrations may
11           reduce a variety of health effects on a time scale from a few days to a few months.  This has
12           been found in epidemiology studies of "natural experiments" such as in the Utah Valley,
13           and by supporting toxicology studies using the particles from ambient community sampling
14           filters from the Utah Valley. Recent studies in Germany and in the Czech Republic also
15           support a hypothesis that reductions in air pollution are associated with reductions in the
16           incidence of adverse health effects, but these studies cannot unambiguously attribute
17           improved health to reduced PM alone.
18      (15)  Adverse health effects in children are emerging as a more important area of concern than in
19           the  1996 PM AQCD. Unfortunately, relatively little is known about the relationship of PM
20           to the most serious health endpoints (low birth weight, preterm birth, neonatal and infant
21           mortality, emergency hospital admissions and mortality in older children).  Also, little is yet
22           known about involvement of PM exposure in the progression from less serious childhood
23           conditions, such as asthma and respiratory symptoms, to more serious disease endpoints
24           later in life. This is an important health issue because childhood illness or death may cost a
25           very large number of productive life-years. Lastly, new epidemiologic studies of ambient
26           PM associations with increased non-hospital medical visits (physician visits) and asthma
27           effects suggest likely much larger health impacts and costs to society due to ambient PM
28           than just those indexed by mortality and/or hospital admissions/visits.
29
        March 2001                               6-270       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. (1995a) Long-term
  6             ambient concentrations of particulates and oxidants and development of chronic disease in a cohort of
  7             nonsmoking California residents. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on
  8             particulate air pollution and human mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol.
  9             7: 19-34.
 10       Abbey, D. E.; Ostro, B. E.; Fraser, G.; Vancuren, T.; Burchette, R. J. (1995b) Estimating fine particulates less that
 11             2.5 microns in aerodynamic diameter (PM2 5) from airport visibility data in California. J. Exposure Anal.
 12             Environ. Epidemiol. 5:  161-180.
 13       Abbey, D. E.; Burchette, R. J.; Knutsen, S. F.; McDonnell, W. F.; Lebowitz, M. D.; Enright, P. L. (1998) Long-term
 14             particulate and other air pollutants and lung function in nonsmokers. Am. J. Respir. Crit. Care Med.
 15             158:289-298.
 16       Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. J.; Knutsen, S. F.; Beeson, W. L.; Yang, J. X.  (1999)
 17             Long-term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir. Crit.
 18             Care Med. 159:373-382.
 19       Ackermann-Liebrich, U.; Leuenberger, P.; Schwartz, J.; Schindler, C.; Monn, C.; Bolognini, B.; Bongard, J. P.;
 20             Brandli, O.; Domenighetti, G.; Elsasser, S.; Gnze,  L.; Karrer, W.; Keller, R.; Keller-Wossidlo, H.;
 21             Kiinzli, N.; Martin, B. W.; Medici, T. C.;  Perruchoud, A. P.; Schoni, M. H.; Tschopp, J. M.; Villiger, B.;
 22             Wiithrich, B.; Zellweger, J. P.; Zemp, E. (1997) Lung function and long term exposure to air pollutants in
 23             Switzerland. Am.  J. Respir. Crit. Care Med. 155: 122-129.
 24       Adams, P. F.; Hendershot, G. E.; Marano, M. A. (1999) Current estimates from the National Health Interview
 25             Survey, 1996. Hyattsville, MD: U.S. Department of Health and Human Services, Public Health Service,
 26             National Center for Health Statistics; publication no. 99-1528. (Vital and health statistics: v. 10, no. 200).
 27             Available: http://www.cdc.gov/nchs/products/pubs/pubd/senes/srlO/pre-200/pre-200.htm [12 March, 2001].
 28       Agocs, M. M.; White, M. C.; Ursicz, G.; Olson,  D. R.; Vamos, A. (1997) A longitudinal study of ambient air
 29             pollutants and the lung peak expiratory flow rates among asthmatic children in Hungary. Int. J. Epidemiol.
 30             26: 1272-1280.
 31       Alberdi Odriozola, J. C.; Diaz Jimenez, J.; Montero Rubio, J. C.; Miron Perez, I. J.; Pajares Ortiz, M. S.; Ribera
 32             Rodrigues, P. (1998) Air pollution and mortality in Madrid, Spain: a time-series analysis. Int. Arch. Occup.
 33             Environ. Health 71: 543-549.
 34       Amdur, M. O.; Chen, L. C. (1989) Furnace-generated acid aerosols: speciation and pulmonary effects.
 35             In: Symposium on the health effects of acid aerosols; October 1987; Research Triangle Park, NC.
 36             Environ. Health Perspect.  79: 147-150.
 37       Anderson, H. R.; Ponce de Leon, A.; Bland, J. M.;  Bower, J. S.; Strachan, D. P. (1996) Air pollution and daily
 38             mortality in London: 1987-92. Br. Med. J. 312: 665-669.
 39       Anderson, H. R.; Spix, C.; Medina, S.; Schouten, J. P.; Castellsague, J.; Rossi, G.; Zmirou, D.; Touloumi, G.;
 40             Wojtyniak, B.; Ponka, A.; Bacharova, L.;  Schwartz, J.; Katsouyanni,  K. (1997) Air pollution and daily
 41             admissions for chronic obstructive pulmonary disease in 6 European cities: results  from the APHEA  project.
 42             Eur. Respir. J. 10: 1064-1071.
 43       Anderson, H. R.; Ponce de Leon, A.; Bland, J.  M.; Bower, J. S.;  Emberlin, J.; Strachen, D. P. (1998) Air pollution,
 44            pollens, and daily admissions for asthma in London 1987-92. Thorax 53: 842-848.
 45       Atkinson, R. W.; Anderson, H. R.; Strachan, D. P.; Bland, J. M.; Bremner, S. A.; Ponce de Leon, A. (1999a)
 46             Short-term associations  between outdoor air pollution and visits to accident and emergency departments in
 47            London for respiratory complaints. Eur. Respir. J. 13: 257-265.
48      Atkinson, R. W.; Bremner, S. A.; Anderson, H. R.;  Strachan, D. P.; Bland, J. M.; Ponce de Leon, A. (1999b)
49            Short-term associations  between emergency hospital admissions for respiratory and cardiovascular disease
 50            and outdoor air pollution in London. Arch. Environ. Health 54: 398-411.
 51       Awasthi, S.; Glick, H. A.; Fletcher, R. H.; Ahmed, N. (1996) Ambient air pollution & respiratory symptoms
 52            complex in preschool children. Indian J. Med. Res. 104: 257-262.
53       Bailey, D. L. R.; Clayton, P. (1982) The measurement of suspended particle  and total carbon concentrations in the
54             atmosphere using standard smoke shade methods. Atmos.  Environ. 16: 2683-2690.

         March 2001                                    6-271         DRAFT-DO NOT QUOTE OR CITE

-------
  1       Baldi, I.; Tessier, J. F.; Kauffmann, F.; Jacqmin-Gadda, H.; Nejjari, C.; Salamon, R. (1999) Prevalence of asthma
  2             and mean levels of air pollution: results from the French PAARC survey. Eur. Respir. J. 14: 132-138.
  3       Bateson, T. F.; Schwartz, J. (1999) Control for seasonal variation and time trend in case-crossover studies of acute
  4             effects of environmental exposures. Epidemiology 10: 539-544.
  5       Beeson, W. L.; Abbey, D. E.; Knutsen, S. F. (1998) Long-term concentrations of ambient air pollutants and incident
  6             lung cancer in California adults: results from the AHSMOG study. Environ. Health Perspect. 106: 813-823.
  7       Berglund, D. J.; Abbey, D. E.; Lebowitz, M. D.; Knutsen, S. F.; McDonnell, W. F. (1999) Respiratory symptoms
  8             and pulmonary function in an elderly nonsmoking population. Chest 115: 49-59.
  9       Beyer, U.; Franke, K.; Cyrys, J.;  Peters, A.; Heinrich, J.; Wichmann, H. E.; Brunekreef, B. (1998) Air pollution and
10             respiratory health of children: the PEACE panel study in Hettstedt and Zerbst, Eastern Germany. Eur. Respir.
11             Rev. 8:61-69.
12       Bobak, M.; Leon, D. A. (1992) Air pollution and infant mortality in the Czech Republic, 1986-1988.
13             Lancet (8826): 1010-1014.
14       Bobak, M.; Leon, D. (1998)  Air pollution and infant mortality: the effects are specific for respiratory causes in
15             postneonatal period. Epidemiology 9: S58.
16       Bobak, M.; Leon, D. A. (1999) Pregnancy outcomes and outdoor air pollution: an ecological study in districts of the
17             Czech Republic 1986-8. Occup. Environ. Med. 56: 539-543.
18       Bobak, M.; Roberts, A.  (1997) Heterogeneity of air pollution effects is related to average temperature [letter].
19             Br. Med. J. 315:  1161.
20       Boezen, M.; Schouten, J.; Rijcken, B.; Vonk, J.; Gemtsen, J.; Van Der Zee, S.; Hoek, G.; Brunekreef, B.;
21             Postma, D. (1998) Peak expiratory flow variability, bronchial responsiveness, and susceptibility to ambient
22             air pollution in adults. Am. J. Respir. Crit. Care Med. 158: 1848-1854.
23       Boezen, H. M.; Van Der Zee, S.  C.; Postma, D. S.; Vonk, J. M.; Gerritsen, J.; Hoek, G.; Brunekreef, B.;
24             Rijcken, B.; Schouten, J. P. (1999) Effects of ambient  air pollution on upper and lower respiratory symptoms
25             and peak expiratory flow in children. Lancet 353: 874-878.
26       Borja-Aburto, V. H.; Loomis, D. P.; Bangdiwala, S. I.; Shy, C. M.; Rascon-Pacheco, R. A. (1997) Ozone,
27             suspended particulates, and daily mortality in Mexico City. Am. J. Epidemiol. 145: 258-268.
28       Borja-Aburto, V. H.; Castillejos, M.; Gold, D. R.; Bierzwinski, S.; Loomis, D. (1998) Mortality and ambient fine
29             particles in southwest Mexico City, 1993-1995. Environ. Health Perspect. 106: 849-855.
30       Braga, A. L. F.; Conceicao, G. M. S.; Pereira, L. A. A.; Kishi, H. S.; Pereira, J. C. R.; Andrade, M. F.; Goncalves,
31             F. L. T.; Saldiva, P. H. N.; Latorre, M. R. D. O. (1999) Air pollution and pediatric respiratory hospital
32             admissions in Sao Paulo, Brazil. J. Environ. Med. 1: 95-102.
33       Braga, A. L. F.; Zanobetti, A.; Schwartz, J. (2000)  Do respiratory epidemics confound the association between air
34             pollution and daily deaths? Eur. Respir. J. 16: 723-728.
35       Braun-Fahrlander, C.; Vuille, J. C.; Sennhauser, F. H.; Neu, U.; Kiinzle, T.; Grize, L.; Gassner, M.; Minder, C.;
36             Schindler, C.; Varonier, H. S.; Wiithnch, B.; SCARPOL team. (1997) Respiratory health and long-term
37             exposure to air pollutants  in Swiss schoolchildren. Am. J. Respir. Crit. Care Med. 155: 1042-1049.
38       Bremner, S. A.; Anderson, H. R.; Atkinson, R. W.; McMichael, A. J.; Strachan, D. P.; Bland, J. M.; Bower, J. S.
39             (1999) Short term associations between outdoor air pollution and mortality in London 1992-4. Occup.
40             Environ. Med. 56: 237-244.
41       Brook, J. R.; Wiebe, A. H.; Woodhouse, S. A.; Audette, C. V.; Dann, T. F.; Callaghan, S.; Piechowski, M.;
42             Dabek-Zlotorzynska, E.; Dloughy, J. F. (1997) Temporal and spatial relationships in fine particle strong
43             acidity, sulphate, PM10, and PM25 across multiple Canadian locations. Atmos. Environ. 31: 4223-4236.
44       Brunekreef, B. (1997) Air pollution and life expectancy: is there a relation? Occup.  Environ. Med. 54: 781-784.
45       Brunekreef, B.; Janssen, N. A. H.; Van Vliet, P. H. N.; Aarts, F. J. H. (2000) Paniculate matter concentrations in
46             relation to degree of urbanization and proximity to highways  in The Netherlands. Presented at:  PM2000:
47             particulate matter and health—the scientific  basis for regulatory decision-making, specialty conference &
48             exhibition; January; Charleston, SC. Pittsburgh, PA: Air & Waste Management Association.
49       Burnett, R. T.; Dales, R. E.; Raizenne, M. E.; Krewski, D.; Summers, P. W.; Roberts,  G. R.; Raad-Young, M.;
50             Dann, T.; Brook, J. (1994) Effects of low ambient levels of ozone and sulfates on the frequency of
51             respiratory admissions to Ontario hospitals. Environ. Res. 65: 172-194.
52       Burnett, R. T.; Dales, R.; Krewski, D.; Vincent, R.; Dann, T.;  Brook, J. R.  (1995) Associations between ambient
53             particulate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. Am. J. Epidemiol.
54             142: 15-22.
         March 2001                                    6-272        DRAFT-DO NOT QUOTE OR CITE

-------
  1       Burnett, R. T.; Dales, R. E.; Brook, J. R.; Raizenne, M. E.; Krewski, D. (1997a) Association between ambient
  2             carbon monoxide levels and hospitalizations for congestive heart failure in the elderly in 10 Canadian cities.
  3             Epidemiology 8: 162-167.
  4       Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997b) The role of paniculate size and chemistry in the
  5             association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
  6             Environ. Health Perspect. 105: 614-620.
  7       Burnett, R. T.; Brook, J. R.; Yung, W. T.; Dales, R. E.; Krewski, D. (1997c) Association between ozone and
  8             hospitalization for respiratory diseases in 16 Canadian cities. Environ. Res. 72: 24-31.
  9       Burnett, R. T.; Cakmak, S.; Brook, J. R. (1998a) The effect of the urban ambient air pollution mix on daily mortality
 10             rates in 11 Canadian cities. Can. J. Public Health 89: 152-156.
 11       Burnett, R. T.; Cakmak, S.; Raizenne, M. E.; Stieb, D.; Vincent, R.; Krewski, D.; Brook, J. R.; Philips, O.;
 12             Ozkaynak, H. (1998b) The association between ambient carbon monoxide levels and daily mortality in
 13             Toronto, Canada. J. Air Waste Manage. Assoc. 48: 689-700.
 14       Burnett, R. T.; Smith-Doiron, M.; Stieb, D.; Cakmak, S.; Brook, J. R. (1999) Effects of particulate and gaseous air
 15             pollution on cardiorespiratory hospitalizations. Arch. Environ. Health 54:130-139.
 16       Burnett, R. T.; Brook, J., Dann, T.; Delocla, C.; Philips, O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
 17             (2000) Association between particulate- and gas-phase components of urban air pollution and daily mortality
 18             in eight Canadian cities. In: Grant, L. D., ed. PM2000: particulate matter and health. Inhalation Toxicol.
 19             12(suppl.4): 15-39.
 20       Cakmak, S.; Burnett, R. T.; Krewski, D. (1999) Methods for detecting and estimating population threshold
 21             concentrations for air pollution-related mortality with exposure measurement error. Risk Anal. 19: 487-496.
 22       Calderon-Garciduenas, L.; Mora-Tiscareno, A.; Chung, C. J.; Valencia, G.;  Fordham, L. A.; Garcia, R.; Osnaya, N.;
 23             Romero, L.; Acufia, H.; Villarreal-Calderon, A.; Devlin, R. B.; Koren, H. S. (2000) Exposure to air pollution
 24             is associated with lung hyperinflation in healthy children and adolescents in southwest Mexico City: a pilot
 25             study.  Inhalation Toxicol. 12: 537-561.
 26       Carroll, R. J.; Ruppert, D.; Stefanski, L. A. (1995) Measurement error in nonlinear models. London, United
 27             Kingdom: Chapman & Hall. (Cox, D. R.; Hinkley, D. V.; Keiding, N.; Reid, N.; Rubin, D. B.; Silverman,
 28             B. W., eds. Monographs on statistics and applied probability: v. 63).
 29       Carrothers, T. J.; Evans, J. S. (2000) Assessing the impact of differential measurement error on estimates of fine
 30             particle mortality. J. Air Waste Manage. Assoc. 50: 65-74.
 31       Cassino, C.; Ito, K.; Bader, I.; Ciotoli, C.; Thurston, G.; Reibman, J. (1999) Cigarette smoking and ozone-associated
 32             emergency department use for asthma by adults in New York City. Am. J. Respir. Crit. Care Med.
 33             159: 1773-1779.
 34       Castillejos, M.; Borja-Aburto, V. H.; Dockery, D. W.;  Gold, D. R.; Loomis, D. (2000) Airborne coarse particles and
 35             mortality. In: Inhalation Toxicology: proceedings of the third colloquium on particulate air pollution and
 36             human health; June,  1999; Durham, NC. Inhalation Toxicology 12(suppl. 1): 61-72.
 37       Checkoway, H.; Levy, D.; Sheppard, L.; Kaufman, J.; Koenig, J.; Siscovick, D. (2000) A case-crossover analysis of
 38             fine particulate matter air pollution and out-of-hospital sudden cardiac arrest. Cambridge, MA: Health Effects
 39             Institute; research report 99. Available: http://www.healtheffects.org/pubs-recent.htm [19 March, 2001].
 40       Chen, C.; Chock, D. P.; Winkler, S. L. (1999) A simulation study of confounding in generalized linear models for
 41              air pollution epidemiology. Environ. Health Perspect. 107: 217-222.
 42       Chen, L.; Yang, W.; Jennison, B. L.; Omaye, S. T. (2000) Air particulate pollution and hospital admissions for
 43             chronic obstructive pulmonary disease in Reno, Nevada. Inhalation Toxicol. 12: 281-298.
44       Chew, F. T.; Goh, D. Y. T.; Ooi, B. C.; Saharom, R.; Hui, J. K. S.; Lee, B. W. (1999) Association of ambient
45             air-pollution levels with acute asthma exacerbation among children in Singapore. Allergy (Copenhagen)
46             54:320-329.
47        Chock, D. P.; Winkler, S.; Chen, C. (2000) A study of the association between daily mortality and ambient air
48             pollutant concentrations in Pittsburgh, Pennsylvania. J. Air Waste Manage. Assoc. 50: 1481-1500.
49        Choudhury, A. H.; Gordian, M. E.; Morris,  S. S. (1997) Associations between respiratory illness and PM,0 air
50             pollution. Arch. Environ. Health 52:  113-117.
51       Cifuentes, L. A.; Vega, J.; Kopfer, K.; Lave, L. B. (2000) Effect of the fine fraction of particulate matter versus the
52             coarse  mass and other pollutants on daily mortality in Santiago, Chile. J. Air Waste Manage. Assoc.
53             50: 1287-1298.
54
55


         March 2001                                    6-273         DRAFT-DO NOT QUOTE OR CITE

-------
  1       Clench-Aas, J.; Bartonova, A.; Skj0nsberg, O. H.; Leegaard, J.; Hagen, L. O.; Giasver, P.; Moseng, J.; Roemer, W.
  2             (1998) Air pollution and respiratory health of children: the PEACE study in Oslo, Norway. Eur. Respir. Rev.
  3             8:36-43.
  4       Clyde, M. A.; Guttorp, P.; Sullivan, E. (2000) Effects of ambient fine and coarse particles on mortality in Phoenix,
  5             Arizona. J. Exposure Anal. Environ. Epidemiol.: submitted.
  6       Cook, J. R.; Stefanski, L. A. (1994) Simulation-extrapolation estimation in parametric measurement error models.
  7             J. Am. Stat. Assoc. 89: 1314-1328.
  8       Coutant,  R. W. (1977) Effect of environmental variables on collection of atmospheric sulfate. Environ. Sci.
  9             Technol. 11:873-878.
10       Cropper, M. L.; Simon, N. B.; Alberini, A.; Arora, S.; Sharma, P. K. (1997) The health benefits of air pollution
11             control in  Delhi. Am. J. Agric. Econ. 79: 1625-1629.
12       Cuijpers, C. E. J.; Swaen, G. M. H.; Wesseling, G.; Wouters, E. F. M. (1994) Acute respiratory effects of summer
13             smog in primary school children. Toxicol. Lett. 72: 227-235.
14       Dab, W.; Medina, S.; Quenel, P.; Le Moullec, Y.; Le Tertre, A.; Thelot, B.; Monteil, C.;  Lameloise, P.; Pirard, P.;
15             Momas, I.; Ferry, R.; Festy, B. (1996) Short term respiratory health effects of ambient air pollution: results of
16             the APHEA project in Paris. In: St Leger, S., ed. The APHEA project. Short term  effects of air pollution on
17             health: a European approach using epidemiological time series data. J. Epidemiol. Community Health
18             50(suppl. 1):S42-S46.
19       Damia, A. D.; Fabregas, M. L.; Tordera, M. P.; Torrero, L. C. (1999) Effects of air pollution and weather conditions
20             on asthma exacerbation. Respiration 66: 52-58.
21       Daniels, M.; Dominici, F.; Samet, J. M.; Zeger, S. L. (2000) Estimating particulate matter-mortality dose-response
22             curves and threshold levels: an analysis of daily time-series for the 20 largest  US cities. Am. J. Epidemiol.
23             152:397-406.
24       Delfino, R. J.; Coate, B. D.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H.; Koutrakis, P. (1996) Daily asthma severity in
25             relation to personal ozone exposure and outdoor fungal spores. Am. J. Respir. Crit. Care Med. 154: 633-641.
26       Delfino, R. J.; Murphy-Moulton, A. M.;  Burnett, R. T.; Brook, J. R.; Becklake,  M. R. (1997) Effects of air pollution
27             on emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Crit. Care Med.
28             155:568-576.
29       Delfino, R. J.; Murphy-Moulton, A. M.;  Becklake, M. R. (1998) Emergency room visits for respiratory illnesses
30             among the elderly in Montreal: association with low level ozone exposure. Environ. Res. 76: 67-77.
31       Dejmek, J.; Selevan, S. G.; Benes, L; Solansky, I.;  Sram, R. J. (1999) Fetal growth and maternal exposure to
32             particulate matter during pregnancy. Environ. Health Perspect. 107: 475-480.
33       Diaz, J.; Garcia, R.; Ribera, P.; Alberdi, J. C.; Hernandez, E.; Pajares, M. S.; Otero,  A. (1999) Modeling of air
34             pollution and its relationship with mortality and morbidity in Madrid,  Spain. Int. Arch. Occup. Environ.
35             Health 72: 366-376.
36       Dockery, D. W.;  Spengler, J. D. (1981) Personal exposure to respirable particulates and sulfates. J. Air Pollut.
37             Control Assoc. 31: 153-159.
38       Dockery, D. W.;  Spengler, J. D.; Neas, L. M.; Speizer, F. E.; Ferris, B. G., Jr.; Ware, J. H.; Brunekreef, B. (1989)
39             An epidemiologic study of respiratory health status and indicators of indoor air pollution from combustion
40             sources. In: Harper, J. P., ed. Combustion processes and the quality of the indoor environment:  transactions
41             of an international specialty conference; September 1988; Niagara Falls,  NY.  Pittsburgh, PA: Air & Waste
42             Management Association; pp. 262-271. (A&WMA transactions series: TR-15).
43       Dockery, D. W.;  Schwartz, J.; Spengler, J. D. (1992) Air pollution and daily  mortality: associations with particulates
44             and acid aerosols.  Environ. Res. 59: 362-373.
45       Dockery, D. W.;  Pope, C. A, III; Xu, X.; Spengler, J. D.; Ware, J. H.; Fay, M. E.; Ferris, B. G., Jr.; Speizer, F. E.
46             (1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med.
47             329: 1753-1759.
48       Dockery, D. W.;  Cunningham, J.;  Damokosh, A. I.; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
49             Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
50             symptoms. Environ.  Health Perspect. 104: 500-505.
51       Dockery, D. W.;  Pope, C. A., Ill; Kanner, R. E.; Villegas, G. M.; Schwartz, J. (1999) Daily changes in oxygen
52             saturation and pulse rate associated with particulate air pollution and barometric pressure. Cambridge, MA:
53             Health Effects Institute; research report no. 83.
54       Dominici, F.; Zeger, S. L.; Samet, J. (2000) A measurement error model for time-series studies of air pollution and
55             mortality.  Biostatistics 1: 157-175.

         March 2001                                     6-274         DRAFT-DO NOT QUOTE OR CITE

-------
  1       Fairley, D. (1990) The relationship of daily mortality to suspended participates in Santa Clara county, 1980-86.
  2             Environ. Health Perspect. 89: 159-168.
  3       Fairley, D. (1999) Daily mortality and air pollution in Santa Clara County, California: 1989-1996. Environ. Health
  4             Perspect. 107:637-641.
  5       Fanucchi, M. V.; Plopper, C. G. (1997) Pulmonary developmental responses to toxicants. In: Roth, R. A., ed.
  6             Toxicology of the respiratory system, v. 8 of comprehensive toxicology. New York, NY: Pergamon, p.
  7             203-220.
  8       Fanucchi, M. V.; Wong, V. J.; Hinds, D.; Tarkington, B. K.; Van Winkle, L. S.; Evans, M. J.; Plopper, C. G. (2000)
  9             Repeated episodes of exposure to ozone alters postnatal development of distal conducting airways in infant
 10             rhesus monkeys. Am. J. Respir. Crit. Care Med. 161: A615.
 11       Forsberg, B.; Segerstedt, B.; Stjernberg, N.; Roemer, W. (1998) Air pollution and respiratory health of children:
 12             the PEACE panel study in Umea, Sweden. Eur. Respir. Rev. 8: 12-19.
 13       Frischer, T.; Studmcka, M.; Gartner, C.; Tauber, E.; Horak, F.; Veiter, A.; Spengler, J,;  Kiihr, J.; Urbanek, R.
 14             (1999) Lung function growth and ambient ozone: a three-year population study in school children. Am. J.
 15             Respir. Crit. Care Med. 160:  390-396.
 16       Fung, K. Y.; Krewski, D. (1999) On measurement error adjustment methods in Poisson regression. Environmetrics
 17             10:213-224.
 18       Gamble, J. L. (1998) Effects of ambient air pollution on daily mortality: a time series analysis of Dallas, Texas,
 19             1990-1994. Presented at: 91st annual meeting and exhibition of the Air & Waste Management Association;
 20             June; San Diego, CA. Pittsburgh, PA: Air & Waste Management Association; paper no. 98-MP26.03.
 21       Garcia-Aymerich, J.; Tobias, A.; Anto, J. M.; Sunyer, J. (2000) Air pollution and mortality in a cohort of patients
 22             with chronic obstructive pulmonary disease: a time series analysis. J. Epidemiol. Community Health
 23             54:73-74.
 24       Garty, B. Z.; Kosman, E.; Ganor, E.; Berger, V.; Garty, L.; Wietzen, T.; Waisman, Y.; Mimouni, M.; Waisel, Y.
 25             (1998) Emergency room visits of asthmatic  children, relation to air pollution, weather, and airborne allergens.
 26             Ann. Allergy Asthma Immunol. 81: 563-570.
 27       Gauderman, W. J.; Mcconnell, R.; Gilliland, F.; London, S.; Thomas, D.; Avol, E.; Vora, H.; Berhane, K.;
 28             Rappaport, E. B.; Lurmann, F.; Margolis, H. G.; Peters, J. (2000) Association between air pollution and lung
 29             function growth in southern California children. Am. J. Respir. Crit. Care Med. 162: 1383-1390.
 30       Gerginova, M.;  Kostianev,  S.; Ivanova, M. (1989)  Reference values of dynamic pulmonary functional indices in
 31             boys between 7 and  14 years of age. Pediatria 28: 52-58.
 32       Gielen, M. H.; Van Der Zee, S. C.; Van Wijnen, J. H.; Van Steen, C. J.; Brunekreef, B.  (1997) Acute effects of
 33             summer air pollution on respiratory health of asthmatic children. Am. J. Respir. Crit. Care Med.
 34             155:2105-2108.
 35       Gilmour, P. S.; Brown, D. M.; Lindsay, T. G.; Beswick, P. H.; MacNee, W.; Donaldson, K. (1996) Adverse health
 36             effects of PM,0 particles: involvement of iron in generation of hydroxyl radical. Occup. Environ. Med.
 37             53: 817-822.
 38       Gold, A.; Litonius, J.; Schwartz, M.; Verrier, R.; Milstein, A.; Larson, E.; Lovett, B. (1998) Cardiovascular
 39            vulnerability to particulate pollution. Presented at: 1998 International conference [American Thoracic
 40             Society]; April; Chicago, IL. Am. J. Respir. Crit. Care Med. 157: A261.
 41        Gold, D. R.; Damokosh, A. L; Pope, C. A., Ill; Dockery, D. W.; McDonnell, W. F.; Serrano, P.; Retama, A.;
 42            Castillejos, M. (1999) Particulate and ozone pollutant effects on the respiratory function of children in
 43             southwest Mexico City. Epidemiology 10: 8-16.
 44        Gold, D. R.; Litonjua, A.; Schwartz, J.; Lovett, E.;  Larson, A.; Nearing, B.; Allen, G.; Verrier, M.; Cherry, R.;
 45             Verrier, R. (2000) Ambient pollution and heart rate variability. Circulation 101: 1267-1273.
 46       Goldberg, M. S.; Bailar, J. C., Ill; Burnett, R. T.; Brook, J. R.; Tamblyn, R.; Bonvalot, Y.; Ernst, P.; Flegel, K. M.;
47             Singh, R. K.; Valois, M.-F. (2000) Identifying subgroups of the general population that may be susceptible to
48             short-term increases in particulate air pollution: a time-series study in Montreal, Quebec. Cambridge, MA:
49             Health Effects Institute; research report 97. Available: http://www.healtheffects.org/pubs-research.htm
 50             [15 February, 2001].
 51       Gordian, M. E.;  Ozkaynak,  H.; Xue, J.; Morris, S. S.; Spengler, J. D. (1996) Particulate air pollution and respiratory
 52             disease in Anchorage, Alaska. Environ.  Health Perspect. 104: 290-297.
53       Gouveia, N.; Fletcher, T. (2000) Respiratory diseases in children and outdoor air pollution in Sao Paulo, Brazil:
54             a time series analysis. Occup. Environ. Med. 57: 477-483.
         March 2001                                    6-275         DRAFT-DO NOT QUOTE OR CITE

-------
  1       Grievink, L.; Van der Zee, S. C.; Hoek, G.; Boezen, H. M.; Van't Veer, P.; Brunekreef, B. (1999) Modulation of the
  2             acute respiratory effects of winter air pollution by serum and dietary antioxidants: a panel study. Eur. Respir.
  3             J. 13: 1439-1446.
  4       Giintzel, O.; Bollag, U.; Helfenstein, U. (1996) Asthma and exacerbation of chronic bronchitis: sentinel and
  5             environmental data in a time series analysis. Zentralbl. Hyg. Umweltmed. 198: 383-393.
  6       Guo, Y. L.; Lin, Y.-C.; Sung, F.-C.; Huang, S.-L.; Ko, Y.-C.; Lai, J.-S.; Su, H.-J.; Shaw, C.-K.; Lin, R.-S.; Dockery,
  7             D. W. (1999) Climate, traffic-related air pollutants, and asthma prevalence in middle-school children in
  8             Taiwan. Environ. Health Perspect. 107: 1001 -1006.
  9       Gwynn, R. C.; Burnett, R. T.; Thurston, G. D. (2000)  A time-series analysis of acidic paniculate matter and daily
10             mortality and morbidity in the Buffalo, New York, region. Environ.  Health Perspect.  108: 125-133.
11       Hagen, J. A.; Nafstad, P.; Skrondal, A.; Bjorkly, S.; Magnus, P. (2000) Associations between outdoor air pollutants
12             and hospitalization for respiratory diseases. Epidemiology 11:  136-140.
13       Hajat, S.; Haines, A.; Goubet, S. A.; Atkinson, R. W.; Anderson, H. R. (1999) Association of air pollution with
14             daily GP consultations for asthma and other lower respiratory conditions in London. Thorax 54: 597-605.
15       Haluszka, J.; Pisiewicz, K.; Miczynski, J.; Roemer, W.; Tomalak, W. (1998) Air pollution and respiratory health in
16             children: the PEACE panel study in Krakow, Poland. Eur. Respir. Rev. 8: 94-100.
17       Harre, E. S. M.; Price, P. D.; Ayrey, R. B.; Toop, L. J.; Martin, I. R.; Town, G. I. (1997) Respiratory effects of air
18             pollution in chronic obstructive pulmonary disease: a three month prospective study. Thorax  52: 1040-1044.
19       Harris, R. J.  (1975) A primer of multivariate statistics. New York, NY: Academic Press; pp.  155-224.
20       Hefflin, B. J.; Jalaludin, B.; McClure, E.; Cobb, N.; Johnson, C. A.; Jecha, L.; Etzel, R. A. (1994) Surveillance for
21             dust storms and respiratory diseases in Washington State, 1991. Arch. Environ. Health 49: 170-174.
22       Heinrich, J.; Hoelscher, B.; Wjst, M.; Ritz, B.; Cyrys, J.; Wichmann, H.-E. (1999) Respiratory diseases and
23             allergies in two  polluted areas in East Germany. Environ. Health Perspect.  107: 53-62.
24       Heinrich, J.; Hoelscher, B.; Wichmann, H. E. (2000) Decline of ambient air pollution and respiratory symptoms in
25             children. Am. J. Respir. Crit. Care Med. 161: 1930-1936.
26       Hiltermann, T. J. N.; de Bruijne, C. R.; Stolk, J.; Zwinderman, A. H.; Spieksma, F. Th. M.; Roemer, W.;
27             Steerenberg, P. A.; Fischer,  P. H.; van Bree, L.; Hiemstra, P. S. (1997) Effects of photochemical air pollution
28             and allergen exposure on upper respiratory tract inflammation in asthmatics. Am. J. Respir. Crit. Care Med.
29             156: 1765-1772.
30       Hiltermann, T. J. N.; Stolk, J.; Van der Zee, S. C.; Brunekreef, B.; De Bruijne, C. R.; Fischer, P. H.; Ameling,
31             C. B.; Sterk, P. J.; Hiemstra, P. S.; Van Bree, L. (1998) Asthma severity and susceptibility to air pollution.
32             Eur. Respir. J. 11:686-693.
33       Hoek, G.; Brunekreef,  B. (1994) Effects of low-level winter air pollution concentrations on respiratory health of
34             Dutch children. Environ. Res. 64: 136-150.
35       Hoek, G.; Schwartz, J. D.; Groot, B.; Eilers, P. (1997) Effects of ambient particulate matter and ozone on daily
36             mortality in Rotterdam, the Netherlands. Arch.  Environ. Health 52: 455-463.
37       Hoek, G.; Dockery, D. W.; Pope, A.; Neas, L.; Roemer, W.; Brunekreef, B. (1998) Association between PM10 and
38             decrements in peak expiratory flow rates in children: reanalysis of data from five panel studies. Eur. Respir.
39             J. 11:  1307-1311.
40       Hoek, G.; Brunekreef,  B.; Verhoeff, A.; van, Wijnen,  J.; Fischer, P. (2000) Daily mortality and air pollution in the
41             Netherlands. J. Air Waste Manage. Assoc. 50:  1380-1389.
42       Hodgkin, J. E.; Abbey, D. E.; Euler, G. L.; Magie, A. R. (1984) COPD prevalence in nonsmokers in high and low
43             photochemical air pollution  areas. Chest 86: 830-838.
44       Hong, Y.-C.; Leem, J.-H.; Ha, E.-H.; Christiani, D. C. (1999) PM10 exposure, gaseous pollutants, and daily
45             mortality in Inchon, South Korea. Environ. Health Perspect. 107: 873-878.
46       Ito, K. (1990) An examination of the role of aerosol acidity in historical London, England daily mortality
47             [dissertation]. Syracuse, NY: New York University. Available  from: University Microfilms International,
48             Ann Arbor, MI; AAD91-13012.
49       Ito, K.; Burnett, R. T.;  Lippmann, M. (1998) Particulate matter components and respiratory hospital admissions in
50             the elderly in Detroit, Ml. In: Abstracts: 1998 International conference; April; Chicago, IL, American
51             Thoracic Society. Am. J. Respir. Crit. Care Med. 157(3 suppl.): A879.
52       Jacobs, J.; Kreutzer, R.; Smith, D. (1997) Rice burning and asthma hospitalizations, Butte County, California,
53             1983-1992. Environ. Health Perspect. 105: 980-985.
54       Jamason, P. F.; Kalkstein, L. S.; Gergen, P. J. (1997) A synoptic evaluation of asthma hospital admissions in
55             New York City. Am. J. Respir. Crit. Care Med. 156: 1781-1788.


         March 2001                                     6-276         DRAFT-DO NOT QUOTE OR CITE

-------
   1       Jedrychowski, W.; Flak, E. (1998) Separate and combined effects of the outdoor and indoor air quality on chronic
   2             respiratory symptoms adjusted for allergy among preadolescent children. Int. J. Occup. Med. Environ. Health
   3             11:19-35.
   4       Jedrychowski, W.; Flak, E.; Mroz, E. (1999) The adverse effect of low levels of ambient air pollutants on lung
   5             function growth in preadolescent children. Environ. Health Perspect. 107: 669-674.
   6       Kalandidi, A.; Gratziou, C.; Katsouyanni, K.; Manalis, N.; Tzala, L.; Pantazopoulou, A.; Efthimiou, M.;
   7             Roussos, C.; Roemer, W. (1998) Air pollution and respiratory health of children: the PEACE panel study in
   8             Athens, Greece. Eur. Respir. Rev. 8: 117-124.
   9       Katsouyanni, K.; Touloumi, G. (1998) Causes of regional differences in air pollution effects are being studied
 10             further [letter]. Br. Med. J. 316: 1982.
 11       Katsouyanni, K.; Schwartz, J.; Spix, C.; Touloumi, G.; Zmirou, D.; Zanobetti, A.; Wojtyniak, B.; Vonk, J. M.;
 12             Tobias, A.; Ponka, A.; Medina, S.; Bacharova, L.; Andersen, H. R. (1996) Short term effects of air pollution
 13             on health: a European approach using epidemiology time series data: the APHEA protocol. In:  St Leger, S.,
 14             ed. The APHEA project. Short term effects of air pollution on health: a European approach using
 15             epidemiological time series data. J. Epidemiol. Community Health 50(suppl. 1): S12-S18.
 16       Katsouyanni, K.; Touloumi, G.; Spix, C.; Schwartz, J.; Balducci, F.; Medina, S.; Rossi, G.; Wojtyniak, B.;
 17             Sunyer, J.; Bacharova, L.; Schouten, J. P.; Ponka, A.; Anderson,  H. R. (1997) Short term effects of ambient
 18             sulphur dioxide and particulate matter on mortality in 12 European cities: results from time series data from
 19             the APHEA project. Br. Med. J. 314: 1658-1663.
 20       Keles, N.; Ilicali, O. C.; Deger, K. (1999) Impact of air pollution on prevalence of rhinitis in Istanbul.  Arch.
 21             Environ. Health 54: 48-51.
 22       Kelsall, J. E.; Samet, J. M.; Zeger, S. L.; Xu, J. (1997) Air pollution and mortality in Philadelphia, 1974-1988.
 23             Am. J. Epidemiol. 146: 750-762.
 24       Klemm, R. J.; Mason, R. M., Jr. (2000) Aerosol research and inhalation epidemiological study (ARIES): air quality
 25             and daily mortality statistical modeling—interim results. J. Air. Waste Manage. Assoc. 50: 1433-1439.
 26       Klemm, R. J.; Mason, R. M., Jr.; Heilig, C. M.; Neas, L. M.; Dockery, D. W. (2000) Is daily mortality associated
 27             specifically with fine particles? Data reconstruction and replication of analyses. J. Air Waste Manage. Assoc.
 28             50: 1215-1222.
 29       Koenig, J. Q,; Larson, T. V.; Hanley, Q. S.; Rebolledo, V.; Dumler, K.; Checkoway, H.; Wang, S.-Z.;  Lin, D.;
 30             Pierson, W. E. (1993) Pulmonary function changes in children associated with fine particulate matter.
 31             Environ. Res. 63: 26-38.
 32       Kontos, A. S.; Fassois, S. D.; Deli, M. F. (1999) Short-term effects of air pollution on childhood respiratory illness
 33             in Piraeus, Greece, 1987-1992: nonparametric stochastic dynamic analysis. Environ. Res. 81: 275-296.
 34       Korrick, S. A.; Neas, L. M.; Dockery, D. W.; Gold, D. R.; Allen, G. A.;  Hill, L. B.; Kimball, K. D.; Rosner, B. A.;
 35             Speizer, F. E. (1998) Effects of ozone and other pollutants on the pulmonary function of adult hikers.
 36             Environ. Health Perspect. 106: 93-99.
 37       Kostianev, S.; Gerginova, M.; Ivanova, M. (1994) Reference values of lung function parameters in Bulgarian girls
 38             aged 7 to  14 years. Pediatria 33: 30-33.
 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 Teplice, Czech Republic. Eur. Respir.  Rev.
 41             8:70-77.
 42       Kotesovec, F.; Skorkovsky, J.; Brynda, J.; Peters, A.; Heinrich, J. (2000) Daily mortality and air pollution in
 43            northern Bohemia; different effects for men and women. Cent. Eur.  J. Public Health 8: 120-127.
44       Kramer, U.; Behrendt, H.; Dolgner, R.; Ranft, U.; Ring, J.; Wilier, H.; Schlipkoter, H.-W.  (1999) Airway diseases
45            and allergies in East and West German children  during the first 5 years after reunification: time  trends  and
46            the impact of sulphur dioxide and total suspended particles. Int. J. Epidemiol. 28: 865-873.
47       Kramer, U.; Koch, T.; Ranft, U.; Ring, J.; Behrendt, H. (2000) Traffic-related air pollution is associated with  atopy
48            in children living in urban areas. Epidemiology 11: 64-70.
49       Krewski, D.; Burnett, R. T.; Goldberg, M. S.; Hoover, K.; Siemiatycki, J.; Jerrett, M.; Abrahamowicz,  M.; White,
50            W. H. (2000) Reanallysis of the Harvard Six Cities study and the American Cancer Society study of
51            particulate air pollution and mortality. A special  report of the Institute's Particle Epidemiology Reanalysis
52            Project. Cambridge, MA: Health Effects Institute.
53      Kiinzli, N.; Tager, I. B. (1997) The semi-individual study in air pollution epidemiology: a valid design as compared
54            to ecologic studies. Environ. Health Perspect.  105: 1078-1083.
         March 2001                                     6-277         DRAFT-DO NOT QUOTE OR CITE

-------
  1       Kiinzli, N.; Ackermann-Liebrich, U.; Brandli, O.; Tschopp, J. M.; Schindler, C.; Leuenberger, P.; SAPALDIA
  2             Team. (2000) Clinically "small" effects of air pollution on FVC have a large public health impact. Eur.
  3             Respir. J. 15: 131-136.
  4       Laden, F.; Neas, L. M.; Dockery, D. W.; Schwartz, J. (2000) Association of fine particulate matter from different
  5             sources with daily mortality in six U.S. cities. Environ. Health Perspect. 108: 941-947.
  6       Lave, L. B.; Seskin, E. P. (1977) Air pollution and human health. Baltimore, MD: The Johns Hopkins University
  7             Press.
  8       Lebowitz, M. D.; Collins, L.; Holberg, C. J. (1987) Time series analyses of respiratory responses to indoor and
  9             outdoor environmental phenomena. Environ. Res. 43:  332-341.
10       Lee, J.-T.; Schwartz, J. (1999) Reanalysis of the effects of air pollution on daily mortality in Seoul, Korea:
11             a case-crossover design. Environ. Health Perspect. 107: 633-636.
12       Lee, J.-T.; Shy, C. M. (1999) Respiratory function as measured by peak expiratory flow rate and PM,0:
13             six communities study. J. Exposure Anal. Environ. Epidemiol. 9: 293-299.
14       Lee, R. E., Jr.; Caldwell, J. S.; Morgan, G. B. (1972) The evaluation of methods for measuring suspended
15             particulates in air. Atmos. Environ. 6: 593-622.
16       Lee, J.-T.; Shin, D.; Chung, Y. (1999) Air pollution and daily mortality in Seoul and Ulsan, Korea. Environ. Health
17             Perspect. 107: 149-154.
18       Leonard), G. S.; Houthuijs, D.; Steerenberg, P. A.; Fletcher, T.; Armstrong, B.; Antova, T. (2000) Immune
19             biomarkers in relation to exposure to particulate matter: a cross-sectional survey in 17 cities of central
20             Europe. In: Grant, L. D., ed. PM2000: particulate matter and health. Inhalation Toxicol. 12(suppl. 4): 1-14.
21       Levy, D.  (1998) Fine particulate air pollution and out-of-hospital mortality in King County, Washington. In: Vostal,
22             J. J., ed. Health effects of particulate matter in ambient air. Proceedings of an international conference; 1997;
23             Prague, Czech Republic. Pittsburgh, PA: Air & Waste Management Association; pp. 262-271. (A&WMA
24             publication VIP-80).
25       Lewis, P. R.; Hensley, M. J.; Wlodarczyk, J.; Toneguzzi, R. C.; Westley-Wise, V. J.; Dunn, T.; Calvert, D. (1998)
26             Outdoor air pollution and children's respiratory symptoms in the steel cities of New South Wales. Med. J.
27             Aust. 169: 459-463.
28       Liao, D.;  Creason, J.; Shy, C.; Williams, R.; Watts, R.; Zweidinger, R. (1999) Daily variation of particulate air
29             pollution and poor cardiac autonomic control in the  elderly. Environ. Health Perspect. 107: 521-525.
30       Lin, C. A.; Martins, M. A.; Farhat, S. C. L.; Pope, C. A., Ill;  Conceicao, G. M. S.; Anastacio, V. M.; Hatanaka, M.;
31             Andrade, W. C.; Hamaue, W. R.; Bohm, G. M.; Saldiva, P. H. N. (1999) Air pollution and respiratory illness
32             of children in Sao Paulo, Brazil. Paediatr. Perinat. Epidemiol. 13: 475-488.
33       Linn, W.  S.; Shamoo, D. A.; Anderson, K. R.; Peng, R.-C.; Avol, E. L.; Hackney, J. D.; Gong, H., Jr. (1996)
34             Short-term air pollution exposures and responses in  Los Angeles area schoolchildren. J. Exposure Anal.
35             Environ. Epidemiol. 6: 449-472.
36       Linn, W.  S.; Szlachcic, Y.;  Gong,  H., Jr.; Kinney, P. L.; Berhane, K. T. (2000) Air pollution and daily hospital
37             admissions in metropolitan Los Angeles. Environ. Health Perspect.  108: 427-434.
38       Lipfert, F. W.; Wyzga, R. E. (1996)  The effects of exposure error on environmental epidemiology. In: Lee, J.;
39             Phalen, R., eds. Proceedings of the 2nd colloquium on particulate air pollution and human health; May;
40             Park City, UT; pp. 4-295 - 4-302.
41       Lipfert, F. W.; Wyzga, R. E. (1997)  Air pollution and mortality: the implications of uncertainties in regression
42             modeling and exposure measurement. J. Air Waste Manage. Assoc. 47: 517-523.
43       Lipfert, F. W.; Morris, S. C.; Wyzga, R. E. (2000a) Daily mortality in the Philadelphia metropolitan area and
44             size-classified particulate matter. J. Air Waste Manage. Assoc.: 1501-1513.
45       Lipfert, F. W.; Perry, H. M., Jr.; Miller, J. P.; Baty, J. D.; Wyzga, R. E.; Carmody, S. E. (2000b) The Washington
46             University-EPRI veterans' cohort mortality study: preliminary results. In: Grant, L. D., ed. PM2000:
47             particulate matter and health.  Inhalation Toxicol.  12(suppl. 4): 41-73.
48       Lipfert, F. W.; Zhang, J.; Wyzga, R. E. (2000c) Infant mortality and air pollution: a comprehensive analysis of U.S.
49             data for 1990. J. Air Waste Manage. Assoc. 50: 1350-1366.
50       Lippmann, M.; Thurston, G. D. (1996) Sulfate concentrations as an indicator of ambient particulate matter air
51             pollution for health risk evaluations. J. Exposure Anal. Environ. Epidemiol. 6: 123-146.
52       Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of particulate matter components with daily
53             mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report no.
54             95.
         March 2001                                     6-278        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       Liu, L.-J. S.; Olson, M. P., Ill; Allen, G. A.; Koutrakis, P.; McDonnell, W. F.; Gerrity, T. R. (1994) Evaluation of
   4             the Harvard  ozone passive sampler on human subjects indoors. Environ. Sci. Technol. 28: 915-923.
   5       Long, W.; Tate, R.  B.; Neuman, M.; Manfreda, J.; Becker, A. B.; Anthonisen, N. R. (1998) Respiratory symptoms
   6             in  a susceptible population due to burning of agricultural residue. Chest 113: 351-357.
   7       Loomis, D.; Castillejos, M.; Gold, D. R.; McDonnell, W.; Borja-Aburto, V. H. (1999) Air pollution and infant
   8             mortality in Mexico City. Epidemiology 10: 118-123.
   9       Lumley, T.; Heagerty, P. (1999) Weighted empirical adaptive variance estimators for correlated data regression.
 10             J. R. Stat. Soc. B 61 (part 2): 459-477.
 11       Mage, D.; Wilson, W.; Hasselblad, V.; Grant, L. (1999) Assessment of human exposure to ambient paniculate
 12             matter. J. Air Waste Manage. Assoc. 49: 174-185.
 13       Mar, T. F.; Morris, G. A.; Koenig, J. Q.;  Larson, T. V. (2000) Associations between air pollution and mortality in
 14             Phoenix, 1995-1997. Environ. Health Perspect. 108: 347-353.
 15       Marcus, A. H.; Chapman, R. (1998) Estimating the health effects of fine particles from epidemiology studies:
 16             how serious  are problems of measurement error, correlation, and confounding? In: Chow, J.; Koutrakis, P.,
 17             eds. 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       McConnell, R.; Berhane, K.; Gilliland, F.; London, S. J.; Vora, H.; Avol, E.; Gauderman, W. J.; Margolis, H. G.;
 20             Lurmann, F.; Thomas, D. C.; Peters, J. M. (1999) Air pollution and bronchitic symptoms in southern
 21             California children with asthma. Environ. Health Perspect. 107: 757-760.
 22       McDonnell, W. F.;  Nishino-Ishikawa, N.; Petersen, F. F.;  Chen, L. H.; Abbey, D. E.  (2000) Relationships of
 23             mortality with the fine and coarse  fractions of long-term ambient PM10 concentrations in nonsmokers.
 24             J. Exposure Anal. Environ. Epidemiol. 10: 427-436.
 25       McGregor, G. R.; Walters, S.; Wordley, J. (1999) Daily hospital respiratory admissions and winter air mass types,
 26             Birmingham, UK. Int. J. Biometeorol. 43: 21-30.
 27       Medina, S.; Le Tertre, A.; Quenel, P.; Le Moullec, Y.; Lameloise, P.; Guzzo, J. C.; Festy, B.; Ferry, R.; Dab, W.
 28             (1997) Air pollution and doctors' house calls: results from the ERPURS system for monitoring the effects of
 29             air pollution  on public health in greater Paris, France, 1991-1995.  Environ. Res. 75: 73-84.
 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       Moolgavkar, S. H. (2000a) Air Pollution  and Mortality in Three U.S. Counties. Environ. Health perspective
 35             108:777-784.
 36       Moolgavkar, S. H. (2000b) Air pollution  and hospital admissions for diseases of the circulatory system in three U.S.
 37            metropolitan  areas. J. Air Waste Manage Assoc. 50: 1199-1206.
 38       Moolgavkar, S. H. (2000c) Air pollution and hospital admissions for chronic obstructive pulmonary disease in three
 39            metropolitan  areas in the United States. In: Grant, L. D., ed. PM2000: particulate matter and health.
 40            Inhalation Toxicol. 12(suppl. 4): 75-90.
 41       Moolgavkar, S. H.; Luebeck, E. G. (1996) A critical review of the evidence on particulate air pollution and
 42            mortality. Epidemiology 7: 420-428.
 43       Moolgavkar, S. H.; Luebeck, E. G.; Anderson, E. L. (1997) Air pollution and hospital admissions for respiratory
 44            causes in Minneapolis-St. Paul and Birmingham. Epidemiology 8:  364-370.
 45       Moolgavkar, S. H.; Hazelton, W.; Luebeck, G.; Levy, D.; Sheppard,  L. (2000) Air pollution, pollens, and
46            admissions for chronic respiratory  disease in King County, Washington. In: Inhalation toxicology:
47            proceedings of the third colloquium on particulate air pollution and human health; June, 1999; Durham, NC.
48            Inhalation Toxicology 12(suppl. 1): 157-171.
49      Morgan, G.; Corbett, S.; Wlodarczyk, J.; Lewis, P. (1998) Air pollution and daily mortality in Sydney, Australia,
 50            1989 through 1993. Am. J. Public Health 88: 759-764.
51      Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
52            evidence of an increased effect at low temperatures. Environ. Health Perspect.  106: 649-653.
53      Morris, R. D.; Naumova, E. N.; Munasinghe, R.  L. (1995) Ambient air pollution and hospitalization for congestive
54            heart failure among elderly people  in seven large US cities. Am. J.  Public Health 85: 1361-1365.
         March 2001                                     6-279        DRAFT-DO NOT QUOTE OR CITE

-------
  1       Naeher, L. P.; Holford, T. R.; Beckett, W. S.; Belanger, K.; Triche, E. W.; Bracken, M. B.; Leaderer, B. P. (1999)
  2             Healthy women's PEF variations with ambient summer concentrations of PM,0, PM2 5, SO42", H+, and O3.
  3             Am. J. Respir. Crit. Care Med. 160: 117-125.
  4       Nauenberg, E.; Basu, K. (1999) Effect of insurance coverage on the relationship between asthma hospitalizations
  5             and exposure to air pollution. Public Health Rep. 114: 135-148.
  6       Navidi, W.; Thomas, D.; Langholz, B.; Stram, D. (1999) Statistical methods for epidemiologic studies of the health
  7             effects of air pollution. Cambridge, MA: Health Effects Institute; research report no. 86.
  8       Neas, L. M.; Dockery, D. W.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr.; Speizer, F. E. (1994) Concentration of
  9             indoor paniculate matter as a determinant of respiratory health in children. Am. J. Epidemiol.
 10             139: 1088-1099.
 11       Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Tollerud, D. J.; Speizer, F. E. (1995) The association of ambient air
 12             pollution with twice daily peak expiratory flow rate measurements in children. Am. J. Epidemiol.
 13             141:111-122.
 14       Neas, L. M.; Dockery, D. W.; Burge, H.; Koutrakis, P.; Speizer, F. E. (1996) Fungus spores, air pollutants, and
 15             other determinants of peak expiratory flow rate in children. Am. J. Epidemiol. 143: 797-807.
 16       Neas, L. M.; Schwartz, J.; Dockery, D. (1999) A case-crossover analysis of air pollution and mortality in
 17             Philadelphia. Environ. Health Perspect. 107: 629-631.
 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       Norris, G.; Young-Pong, S. N.; Koenig, J. Q.; Larson, T. V.; Sheppard, L.; Stout, J. W. (1999) An association
21             between fine particles and asthma emergency department visits for children in Seattle. Environ. Health
22             Perspect. 107:489-493.
23       Norris, G.; Larson, T.; Koenig, J.; Claiborn, C.; Sheppard, L.; Finn, D. (2000) Asthma aggravation, combustion, and
24             stagnant air. Thorax 55: 466-470.
25       Ostro, B. (1995) Fine particulate air pollution and mortality in two Southern California counties. Environ. Res.
26             70:98-104.
27       Ostro, B. D.; Lipsett, M. J.; Wiener, M. B.; Seiner, J. C. (1991) Asthmatic responses to airborne acid aerosols.
28             Am. J. Public Health 81: 694-702.
29       Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Braxton-Owens, H.; White, M. C. (1995) Air pollution and asthma
30             exacerbations among African-American children in Los Angeles. In: Phalen, R. F.; Bates, D. V., eds.
31             Proceedings of the colloquium on particulate air pollution and human mortality and morbidity, part II;
32             January 1994; Irvine, CA. Inhalation Toxicol. 7: 711-722.
33       Ostro, B.; Chestnut, L.; Vichit-Vadakan, N.; Laixuthai, A. (1998) The impact of fine particulate matter in Bangkok,
34             Thailand. In: Chow, J.; Koutrakis, P., eds. PM2.5: a fine particle standard. Volume II: proceedings of an
35             international specialty conference; January; Long Beach, CA. Pittsburgh, PA: Air & Waste Management
36             Association; pp. 939-949. (A&WMA publication VIP-81).
37       Ostro, B. D.; Hurley, S.; Lipsett, M. J. (1999a) Air pollution and daily mortality in the Coachella Valley, California:
38             a study of PMIO dominated by coarse particles. Environ. Res. 81: 231-238.
39       Ostro, B. D.; Eskeland, G. S.; Sanchez, J. M.; Feyzioglu, T. (1999b) Air pollution and health effects: a study of
40             medical visits among children in Santiago, Chile. Environ. Health Perspect. 107: 69-73.
41       Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coachella
42             Valley, CA: a follow-up study. J. Exposure Anal. Environ. Epidemiol.: 10: 412-419.
43       Ott, W.; Wallace, L.; Mage, D. (2000) Predicting particulate (PM,0) personal exposure distributions using a random
44             component  superposition statistical model. J. Air Waste Manage. Assoc. 50: 1390-1406.
45       Ozkaynak, H.; Thurston, G. D. (1987) Associations between 1980 U.S. mortality rates and alternative measures of
46             airborne particle concentration. Risk Anal. 7: 449-461.
47       Ozkaynak, H.; Xue, J.; Zhou, H.; Raizenne, M. (1996) Associations between daily mortality and motor vehicle
48             pollution in Toronto, Canada. Boston, MA: Harvard University School of Public Health, Department of
49             Environmental Health; March 25.
50       Pantazopoulou, A.; Katsouyanni, K.; Kourea-Kremastinou, J.; Trichopoulos, D. (1995) Short-term effects of air
51             pollution on hospital emergency outpatient visits and admissions in the greater Athens, Greece area. Environ.
52             Res. 69:31-36.
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.
         March 2001                                     6-280        DRAFT-DO NOT QUOTE OR CITE

-------
   1       Pereira, L. A. A.; Loomis, D.; Conceicao, 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       Perry, H. M., Jr.; Schnaper, H. W.; Meyer, G.; Swatzell, R. (1982) Clinical program for screening and treatment of
   5             hypertension in veterans. J. Natl. Med. Assoc. 74: 433-444.
   6       Peters, A.; Goldstein, I. F.; Beyer, U.; Franke, K.; Heinrich, J.; Dockery, D. W.; Spengler, J. D.; Wichmann, H.-E.
   7             (1996) Acute health effects of exposure to high levels of air pollution in eastern Europe. Am. J. Epidemiol.
   8             144:570-581.
   9       Peters, A.; Doring,  A.; Wichmann, H.-E.; Koenig, W. (1997a) Increased plasma viscosity during an air pollution
 10             episode: a link to mortality? Lancet 349: 1582-1587.
 11       Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997b)  Respiratory effects are associated with the
 12             number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
 13       Peters, A.; Dockery, D. W.; Heinrich, J.; Wichmann, H. E. (1997c)  Short-term effects of paniculate air pollution on
 14             respiratory morbidity in asthmatic children. Eur. Respir. J. 10: 872-879.
 15       Peters, A.; Kotesovec, F.; Skorkovsky, J.; Brynda, J.; Heinrich, J. (1999a) Akute Auswirkung der
 16             Schwebstaubkonzentrationen in der Aufienluft auf die Mortahtat  - Vergleichsstudie Nordost-Bayern /
 17             Nordbohmen. AbschluBbericht [Acute effects of suspended particle concentrations in the atmosphere on
 18             mortality - a  study comparing northeast Bavaria and north Bohemia. Final report]. Bavaria, Federal Republic
 19             of Germany: Institut fur Epidemiologie; report no. GSF-EP S 1/99.
 20       Peters, J. M.; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
 21             Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999b) A study of twelve southern California communities with
 22             differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit. Care
 23             Med.  159:760-767.
 24       Peters, J. M.; Avol, E.; Gauderman, W. J.; Linn, W. S.; Navidi, W.; London, S. J.; Margolis, H.; Rappaport, E.;
 25             Vora,  H.; Gong, H., Jr.; Thomas, D. C. (1999c) A study of twelve southern California communities with
 26             differing levels and types of air pollution. II. Effects on pulmonary function. Am. J. Respir. Crit. Care Med.
 27             159: 768-775.
 28       Peters, A.; Liu, E.; Verrier, R. L.;  Schwartz, J.; Gold, D. R.; Mittleman, M.; Baliff, J.; Oh, J. A.; Allen, G.;
 29             Monahan, K.; Dockery, D. W. (2000a) Air pollution and incidence of cardiac arrhythmia. Epidemiology
 30             11:11-17.
 31       Peters, A.; Frohlich, M.; Doring, A.; Immervoll, T.; Wichmann, H.-E.; Hutchinson, W. L.; Pepys, M. B.; Koenig,
 32             W. (2000b) Particulate air pollution is associated with  an acute phase response in men: results from the
 33             MONICA-Augsburg Study. Eur. Heart J.: in press.
 34       Ponce de Leon, A.; Anderson, H. R.; Bland, J. M.; Strachan, D. P.; Bower, J. (1996) Effects of air pollution on
 35             daily hospital admissions for respiratory disease in London between 1987-88 and 1991-92. In: St Leger, S.,
 36             ed. The APHEA project. Short term effects of air pollution on health: a European approach using
 37             epidemiological time series data. J. Epidemiol. Community Health 50(suppl.  1): S63-S70.
 38       Ponka, A.; Savela, M.;  Virtanen, M. (1998) Mortality and air  pollution in Helsinki. Arch. Environ. Health
 39             53:281-286.
 40       Pope, C. A., III. (1991) Respiratory hospital  admissions associated with PMIO pollution in Utah, Salt Lake, and
 41             Cache Valleys. Arch. Environ. Health 46: 90-97.
 42       Pope, C. A., Ill; Kalkstein, L S. (1996) Synoptic weather modeling and estimates of the exposure-response
 43            relationship between daily mortality and particulate air pollution. Environ. Health Perspect. 104: 414-420.
 44       Pope, C. A., Ill; Dockery, D. W.; Spengler, J. D.; Raizenne, M. E. (1991) Respiratory health and PM)0 pollution:
 45            a daily time series analysis.  Am. Rev.  Respir. Dis. 144: 668-674.
 46       Pope, C. A., Ill; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F.  E.; Heath, C. W., Jr.
47            (1995) Particulate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am. J.
48            Respir. Crit. Care Med. 151: 669-674.
49       Pope, C. A, III; Hill, R. W.; Villegas, G. M.  (1999a) Particulate air pollution and daily mortality on Utah's Wasatch
 50            Front. Environ. Health Perspect.: 107: 567-573.
51       Pope, C. A., Ill; Dockery, D. W.; Kanner, R. E.; Vollegas, G.  M.; Schwartz, J. (1999b) Oxygen saturation, pulse
52            rate, and particulate air pollution: a daily time-series panel study. Am. J. Respir. Crit.  Care Med.
53            159:365-372.
         March 2001                                     6-281         DRAFT-DO NOT QUOTE OR CITE

-------
  1       Pope, C. A., Ill; Verrier, R. L.; Lovett, E. G.; Larson, A. C.; Raizenne, M. E.; Kanner, R. E.; Schwartz, J.; Villegas,
  2             G. M.; Gold, D. R.; Dockery, D. W. (1999c) Heart rate variability associated with paniculate air pollution.
  3             Am. Heart J. 138:890-899.
  4       Prescott, G. J.; Cohen, G. R.; Elton, R. A.; Fowkes, F. G. R.; Agius, R. M. (1998) Urban air pollution and
  5             cardiopulmonary ill health: a 14.5 year time series study. Occup. Environ. Med. 55: 697-704.
  6       Prescott, E.; Lange, P.; Vestbo, J.; Copenhagen City Heart Study Group. (1999) Socioeconomic status, lung
  7             function and admission to hospital for COPD: results from the Copenhagen City Heart Study. Eur. Respir. J.
  8             13:1109-1114.
  9       Qian, Z. Chapman, R. S.; Tian, Q.; Chen, Y.; Lioy, P. J.; Zhang, J. (2000) Effects of air pollution on children's
 10             respiratory health in three Chinese cities. Arch. Environ. Health 55: 126-133.
 11       Rahlenbeck, S. I.; Kahl, H. (1996) Air pollution and mortality in East Berlin during the winters of 1981-1989. Int. J.
 12             Epidemiol. 25: 1220-1226.
 13       Raizenne, M.; Neas, L. M.; Damokosh, A. I.; Dockery, D. W.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer,
 14             F. E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ.
 15             Health Perspect.  104: 506-514.
 16       Roemer, W.; Hoek, G.; Brunekreef, B. (1993) Effect of ambient winter air pollution on respiratory health of
 17             children with chronic respiratory symptoms. Am. Rev. Respir. Dis. 147: 118-124.
 18       Roemer, W.; Hoek, G.; Brunekreef, B.; Haluszka, J.; Kalandidi, A.; Pekkanen, J. (1998) Daily variations in air
 19             pollution and respiratory health in a multicentre study: the PEACE project. Eur. Respir. J. 12: 1354-1361.
 20       Roemer, W.; Hoek, G.; Brunekreef, B.; Clench-Aas, J.; Forsberg, B.; Pekkanen, J.; Schutz, A. (2000) PM,0
 21             elemental composition and acute respiratory health effects in European children (PEACE project). Eur.
 22             Respir. J. 15:553-559.
 23       Rojas-Bracho, L.; Suh, H. H.; Koutrakis, P. (2000) The relationship between personal exposures to PM25 and PM10
 24             and their corresponding indoor and outdoor concentrations in Boston, MA. Environ. Health Perspect.:
 25             submitted.
 26       Romieu, I.; Meneses, F.; Ruiz, S.; Sienra, J. J.; Huerta, J.; White, M. C.; Etzel, R. A. (1996) Effects of air pollution
 27             on the respiratory health of asthmatic children living in Mexico City. Am. J. Respir. Crit. Care Med.
 28             154:300-307.
 29       Romieu, I.; Meneses, F.; Ruiz, S.; Huerta, J.; Sienra, J. J.; White, M.; Etzel, R.; Hernandez, M. (1997) Effects of
 30             intermittent ozone exposure on peak expiratory  flow and respiratory symptoms among asthmatic children in
 31             Mexico City. Arch. Environ. Health 52: 368-376.
 32       Rooney, C.; McMichael, A. J.; Kovats, R. S.; Coleman, M. P. (1998) Excess mortality in England and Wales, and in
 33             greater London, during the 1995 heatwave. J. Epidemiol. Community Health 52: 482-486.
 34       Rosas, I.; McCartney, H. A.; Payne, R. W.; Calderon, C.; Lacey, J.; Chapela, R.; Ruiz-Velazco, S. (1998) Analysis
 35             of the relationships between environmental factors (aeroallergens, air pollution, and weather) and asthma
 36             emergency admissions to a hospital in Mexico City. Allergy (Copenhagen) 53:  394-401.
 37       Rossi, G.; Vigotti, M. A.; Zanobetti, A.; Repetto, F.; Gianelle, V.; and Schwartz, J. (1999) Air pollution and
 38             cause-specific mortality in Milan, Italy, 1980-1989. Arch. Environ. Health 54: 158-164.
 39       Rothman, K. J.; Greenland, S., eds. (1998) Modern epidemiology. 2nd ed. Philadelphia,  PA: Lippincott-Raven
40             Publishers.
41       Rutherford, S.; Clark, E.; McTainsh, G.; Simpson,  R.; Mitchell, C. (1999) Characteristics of rural dust events shown
42             to impact on asthma severity in Brisbane, Australia. Int. J. Biometeorol. 42: 217-225.
43       Samet, J. M.; Zeger, S. L.; Kelsall, J. E.; Xu, J.; Kalkstein, L. S. (1996) Weather, air pollution and mortality in
44             Philadelphia, 1973-1980, report to the Health Effects Institute on phase IB, Particle Epidemiology Evaluation
45             Project. Cambridge, MA: Health Effects Institute; review draft.
46       Samet, J. M.; Zeger, S. L.; Kelsall, J. E.; Xu, J.; Kalkstein, L. S. (1997) Paniculate air pollution and daily mortality:
47             analysis of the effects of weather and multiple air pollutants. The phase I.B report of the Particle
48             Epidemiology Evaluation Project. Cambridge, MA: Health effects Institute; HEI special report.
49       Samet, J.; Zeger, S.; Kelsall, J.; Xu, J.; Kalkstein, L. (1998) Does weather confound of modify the association of
50             particulate air pollution with mortality? An analysis of the Philadelphia data, 1973-1980. Environ. Res.
51             77:9-19.
52       Samet, J. M.; Dominici, F.; Zeger, S. L.; Schwartz, J.; Dockery, D. W. (2000a) National morbidity, mortality, and
53             air pollution study. Part I:  methods and methodologic issues. Cambridge, MA: Health Effects Institute;
54             research report no. 94.
         March 2001                                     6-282         DRAFT-DO NOT QUOTE OR CITE

-------
  1       Samet, J. M.; Zeger, S. L.; Dominici, F.; Curriero, F.; Coursac, I.; Dockery, D. W.; Schwartz, J.; Zanobetti, A.
  2             (2000b) The national morbidity, mortality, and air pollution study. Part II: morbidity, mortality, and air
  3             pollution in the United States. Cambridge, MA: Health Effects Institute; research report no. 94.
  4       Scarlett, J. F.; Abbott, K. J.; Peacock, J. L.; Strachan, D. P.; Anderson, H. R. (1996) Acute effects of summer air
  5             pollution on respiratory function in primary school children in southern England. Thorax 51:1109-1114.
  6       Schouten, J. P.; Vonk, J. M.; de Graaf, A. (1996) Short term effects of air pollution on emergency hospital
  7             admissions for respiratory disease: results of the APHEA project in two major cities in The Netherlands,
  8             1977-89. In: St Leger, S., ed. The APHEA project. Short term effects of air pollution on health: a European
  9             approach using epidemiological time series data. J. Epidemiol. Community Health 50(suppl. 1): S22-S29.
 10       Schwartz, J. (1994a) What are people dying of on high air pollution days? Environ. Res. 64: 26-35.
 11       Schwartz, J. (1994b) PM,0, ozone, and hospital admissions for the elderly in Minneapolis, MN. Arch. Environ.
 12             Health 49: 366-374.
 13       Schwartz, J. (1995) Short term fluctuations in air pollution and hospital admissions of the elderly for respiratory
 14             disease. Thorax 50: 531-538.
 15       Schwartz, J. (1997) Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology
 16             8:371-377.
 17       Schwartz, J. (1999) Air pollution and hospital admissions for heart disease in eight U.S. counties. Epidemiology
 18             10:17-22.
 19       Schwartz, J. (2000a) Assessing confounding, effect modification, and thresholds in the association between ambient
 20             particles and daily deaths. Environ. Health Perspect. 108: 563-568.
 21       Schwartz, J. (2000b) The distributed lag between air pollution and daily deaths. Epidemiology 11: 320-326.
 22       Schwartz, J. (2000c) Harvesting and long term exposure effects in the relation between air pollution and mortality.
 23             Am. J. Epidemiol.  151: 440-448.
 24       Schwartz, J. (2000d) Daily deaths are associated  with combustion particles rather than SO2 in  Philadelphia.
 25             Occup. Environ. Med. 57: 692-697.
 26       Schwartz, J.; Dockery, D. W. (1992) Increased mortality in Philadelphia associated with daily air pollution
 27             concentrations. Am. Rev. Respir. Dis. 145: 600-604.
 28       Schwartz, J.; Morris, R. (1995) Air pollution and hospital  admissions for cardiovascular disease in Detroit,
 29             Michigan. Am. J. Epidemiol. 142: 23-35.
 30       Schwartz, J.; Neas, L. M. (2000) Fine particles are more strongly associated than coarse particles with acute
 31             respiratory health effects in schoolchildren. Epidemiology.  11:6-10.
 32       Schwartz, J.; Zanobetti, A. (2000) Using meta-smoothing to estimate dose-response trends across multiple studies,
 33             with application to air pollution and daily death. Epidemiology 11: 666-672.
 34       Schwartz, J.; Dockery, D. W.; Neas, L. M.; Wypij, D.; Ware, J. H.; Spengler, J. D.; Koutrakis, P.; Speizer, F. E.;
 35             Ferris, B. G., Jr. (1994) Acute effects of summer air pollution on respiratory symptom reporting in children.
 36             Am. J. Respir. Crit. Care Med. 150: 1234-1242.
 37       Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996a) Is daily mortality associated specifically with fine particles?
 38             J. Air Waste Manage. Assoc. 46: 927-939.
 39       Schwartz, J.; Spix,  C.; Touloumi, G.; Bacharova, L.; Barumamdzadeh, T.; le Tertre, A.; Piekarksi, T.; Ponce  de
 40             Leon, A.; Ponka, A.; Rossi, G.; Saez, M.; Schouten, J. P. (1996b) Methodological issues in studies of air
 41              pollution and daily counts of deaths or hospital admissions. In: St Leger, S., ed. The APHEA project. Short
 42             term effects of air pollution on health: a European approach using epidemiological time series data.
 43              J. Epidemiol. Community Health 50(suppl. 1): S3-S11.
 44       Schwartz, J.; Norris, G.; Larson, T.; Sheppard, L.; Claiborne, C.; Koenig, J. (1999) Episodes of high coarse particle
 45              concentrations are not associated with increased mortality. Environ. Health Perspect. 107: 339-342.
46        Schwarz, G. (1978) Estimating the dimension of a model. Ann. Stat.  6: 461-464.
47        Seaton, A.; Soutar, A.; Crawford, V.; Elton, R.; McNerlan, S.; Cherrie, J.; Watt, M.; Agius, R.; Stout, R. (1999)
48              Particulate air pollution and the blood. Thorax 54: 1027-1032.
49        Segala, C.; Fauroux, B.; Just, J.;  Pascual, L.; Grimfeld, A.; Neukirch, F.  (1998) Short-term effect of winter air
 50              pollution on respiratory health of asthmatic children in Paris. Eur. Respir. J. 11: 677-685.
 51        Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects of ambient air pollution on
52              nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
 53       Shumway, R. H.; Tai, R. Y.; Tai, L. P.; Pawitan, Y. (1983) Statistical analysis of daily London mortality and
54              associated weather and pollution effects. Sacramento, CA: California Air Resources Board; contract no.
55             Al-154-33.


         March 2001                                     6-283        DRAFT-DO NOT QUOTE OR CITE

-------
  1       Shumway, R. H.; Azari, A. S.; Pawitan, Y. (1988) Modeling mortality fluctuations in Los Angeles as functions of
  2             pollution and weather effects. Environ. Res. 45: 224-241.
  3       Simpson, R. W.; Williams, G.; Petroeschevsky, A.; Morgan, G.; Rutherford, S. (1997) Associations between
  4             outdoor air pollution and daily mortality in Brisbane, Australia. Arch. Environ. Health 52: 442-454.
  5       Smith, M. A.; Jalaludin, B.; Byles, J. E.; Lim, L.; Leeder, S. R. (1996) Asthma presentations to emergency
  6             departments in western Sydney during the January 2194 bushfires. Int. J. Epidemiol. 25: 1227-1236.
  7       Smith, R. L.; Spitzner, D.; Kim, Y.; Fuentes, M. (2000) Threshold dependence of mortality effects for fine and
  8             coarse particles in Phoenix, Arizona. J. Air Waste Manage. Assoc. 50: 1367-1379.
  9       Spektor, D. M.; Hofmeister, V. A.; Artaxo, P.; Brague, J. A. P.; Echelar, F.; Nogueira, D. P.; Hayes, C.; Thurston,
10             G. D.; Lippmann, M. (1991) Effects of heavy industrial pollution on respiratory function in the children of
11             Cubatao, Brazil: a preliminary report. Environ. Health Perspect. 94: 51-54.
12       Spix, C.; Heinrich, J.; Dockery, D.; Schwartz, J.; Volksch, G.; Schwinkowski, K.; Collen, C.; Wichmann, H. E.
13             (1993) Air pollution and daily mortality in Erfurt, East Germany, 1980-1989. Environ. Health Perspect.
14             101:518-526.
15       Spix, C.; Anderson, H. R.;  Schwartz, J.; Vigotti, M. A.; LeTertre, A.; Vonk, J. M.; Touloumi, G.; Balducci, F.;
16             Piekarski, T.; Bacharova, L.; Tobias, A.; Ponka, A.; Katsouyanni, K. (1998) Short-term effects of air
17             pollution on hospital admissions  of respiratory diseases in Europe: a quantitative summary of APHEA  study
18             results. Arch. Environ. Health 53: 54-64.
19       Stieb, D. M.; Burnett, R. T.; Beveridge, R. C.;  Brook, J. R. (1996) Association between ozone and asthma
20             emergency department visits in Saint John, New Brunswick, Canada. Environ. Health Perspect.
21             104: 1354-1360.
22       Stieb, D. M.; Beveridge, R. C.; Rowe, B. H.; Walter, S. D.; Judek, S. (1998a) Assessing diagnostic classification in
23             an emergency department: implications for daily time series studies of air pollution. Am. J. Epidemiol.
24             148:666-670.
25       Stieb, D. M.; Brook, J. R.; Broder, I.; Judek, S.; Burnett, R.  T.; Beveridge, R. C. (1998b) Personal exposure of
26             adults with cardiorespiratory disease to particulate acid and sulfate in Saint John, New Brunswick, Canada.
27             Appl. Occup. Environ. Hyg. 13: 461-468.
28       Stieb, D. M.; Beveridge, R. C.; Brook, J. R.; Smith-Doiron,  M.; Burnett, R. T.; Dales, R. E.; Beaulieu, S.; Judek, S.;
29             Mamedov, A. (2000) Air pollution, aeroallergens and cardiorespiratory emergency department visits in Saint
30             John, Canada. J. Exposure Anal.  Environ. Epidemiol.: 10: 461-477.
31       Studnicka, M. J.; Frischer, T.; Meinert,  R.; Studnicka-Benke, A.; Hajek, K.; Spengler, J. D.; Neumann, M. G.
32             (1995) Acidic particles and lung  function in children: a summer camp study in the Austrian Alps. Am.  J.
3 3             Respir. Crit. Care Med. 151: 423-430.
34       Sunyer, J.; Spix, C.; Quenel, P.; Ponce-de-Leon, A.; Ponka,  A.; Barumandzadeh, T.; Touloumi, G.; Bacharova, L.;
35             Wojtyniak, B.; Vonk, J.; Bisanti, L.; Schwartz, J.; Katsouyanni, K. (1997) Urban air pollution and emergency
36             admissions for asthma  in four European cities: the APHEA project. Thorax 52: 760-765.
37       Sunyer, J.; Schwartz, J.; Tobias, A.; Macfarlane, D.; Garcia, J.; Anto, J. M. (2000) Patients with chronic obstructive
38             pulmonary disease are  at increased risk of death associated with urban particle air pollution: a case-crossover
39             analysis. Am. J. Epidemiol. 151:  50-56.
40       Taggart, S. C. O.; Custovic, A.; Francis, H. C.; Faragher, E. B.; Yates, C. J.; Higgins, B. G.; Woodcock, A. (1996)
41             Asthmatic bronchial hyperresponsiveness varies with ambient levels of summertime air pollution. Eur.
42             Respir. J. 9: 1146-1154.
43       Tanaka, H.; Honma, S.; Nishi, M.; Igarashi, T.; Teramoto, S.; Nishio, F.; Abe, S. (1998) Acid fog and hospital visits
44             for asthma: an epidemiological study. Eur. Respir. J.  11: 1301-1306.
45       Tenias, J. M.; Ballester, F.; Rivera, M. L. (1998) Association between hospital emergency visits for asthma and air
46             pollution in Valencia, Spain. Occup. Environ. Med. 55: 541-547.
47       Thurston, G. D.; Lippmann, M.; Scott, M. B.; Fine, J. M. (1997) Summertime  haze air pollution and children  with
48             asthma. Am. J. Respir. Crit. Care Med. 155: 654-660.
49       Tiittanen, P.; Timonen, K. L.; Ruuskanen, J.; Mirme, A.; Pekkanen, J. (1999) Fine particulate air pollution,
50             resuspended road dust  and respiratory health among symptomatic children. Eur. Respir. J. 13: 266-273.
51       Timonen, K. L.; Pekkanen, J.  (1997) Air pollution and respiratory health among children with asthmatic or cough
52             symptoms. Am. J. Respir. Crit. Care Med. 156: 546-552.
53       Tobias, A.; Campbell, M. J. (1999) Modelling  influenza epidemics in the  relation between black smoke and total
54             mortality. A sensitivity analysis. J. Epidemiol. Community Health 53: 583-584.
         March 2001                                     6-284        DRAFT-DO NOT QUOTE OR CITE

-------
   1      Tolbert, P. E.; Klein, M.; Metzger, K. B.; Flanders, W. D.; Todd, K.; Mulholland, J. A.; Ryan, P. B.; Frumkin, H.
   2            (2000a) Interim results of the study of particulates and health in Atlanta (SOPHIA). J. Exposure Anal.
   3            Environ. Epidemiol. 10: 446-460.
   4      Tolbert, P. E.; Mulholland, J. A.; Macintosh, D. L.; Xu, F.; Daniels, D.; Devine, O. J.; Carlin, B. P.; Klein, M.;
   5            Dorley, J.; Butler, A. J.; Nordenberg, D. F.; Frumkin, H.; Ryan, P. B.; White, M. C. (2000b) Air quality and
   6            pediatric emergency room visits for asthma in Atlanta, Georgia. Am. J. Epidemiol. 151: 798-810.
   7      Touloumi, G.; Katsouyanni, K.; Zmirou, D.; Schwartz, J.; Spix, C.; Ponce de Leon, A.; Tobias, A.; Quennel, P.;
   8            Rabczenko, D.;  Bacharova, L.; Bisanti, L.; Vonk, J. M.; Ponka, A. (1997) Short-term effects of ambient
   9            oxidant exposure on mortality: a combined  analysis within the APHEA project. Am. J. Epidemiol.
 10            146: 177-185.
 11      Tsai, F. C.; Apte, M. G.; Daisey, J. M. (2000) An exploratory analysis of the relationship between mortality and the
 12            chemical composition of airborne particulate matter. Inhalation Toxicol. 12(suppl.): 121-135.
 13      Turnovska, T.; Kostianev, S. (1999)  Effects of reduced air pollution on children's pulmonary function. Cent. Eur. J.
 14            Public Health 7: 77-79.
 15      U.S. Environmental Protection Agency. (1986) Second addendum to air quality criteria for particulate matter and
 16            sulfur oxides (1982): assessment of newly available health effects information. Research Triangle Park, NC:
 17            Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; EPA report
 18            no. EPA-600/8-86-020F. Available from: NTIS, Springfield, VA;  PB87-176574.
 19      U.S. Environmental Protection Agency. (1996) Air quality criteria for particulate matter. Research Triangle Park,
 20            NC: National  Center for Environmental  Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
 21       Van Der Zee,  S. C.; Hoek, G.; Boezen, H. M.; Schouten, J. P.; Van Wijnen, J. H.; Brunekreef, B. (1999) Acute
 22            effects of urban  air pollution on respiratory  health of children with and without chronic respiratory
 23            symptoms. Occup. Environ. Med. 56: 802-813.
 24      Van Der Zee, S. C.; Hoek, G.; Boezen, M. H.; Schouten, J. P.; Van Wijnen, J. H.; Brunekreef, B. (2000) Acute
 25            effects of air pollution on respiratory health of 50-70 yr old adults. Eur. Respir. J. 15: 700-709.
 26      Vedal, S.; Petkau, J.; White, R.; Blair, J. (1998) Acute effects of ambient inhalable particles in asthmatic and
 27            nonasthmatic children. Am. J. Respir. Crit. Care Med. 157: 1034-1043.
 28      Veterans Administration Cooperative Study Group on Antihypertensive Agents. (1967) Effects of treatment on
 29            morbidity in hypertension. Results in patients with diastolic blood  pressures averaging 115 through 129 mm
 30            Hg. JAMA J. Am. Med.  Assoc. 202: 1028-1034.
 31       Veterans Administration Cooperative Study Group on Antihypertensive Agents. (1970) Effects of treatment on
 32             morbidity in hypertension. II. Results in patients with diastolic blood pressure averaging 90 through 114 mm
 33             Hg. JAMA J. Am. Med.  Assoc. 213: 1143-1152.
 34       Vigotti, M. A.; Rossi, G.; Bisanti, L.; Zanobetti, A.; Schwartz, J. (1996) Short term effects of urban air pollution on
 35            respiratory health in Milan, Italy, 1980-89. In: Leger, S., ed. The APHEA project. Short term effects of air
 36            pollution on health: a European approach using epidemiological time series data. J. Epidemiol. Community
 37            Health 50(suppl. 1): S71-S75.
 38       Wallace, L. (1996) Indoor particles: a review. J. Air Waste Manage. Assoc. 46: 98-126.
 39       Wallace, L. (2000) Correlations of personal exposure to particles with outdoor air measurements: a review of recent
 40            studies. Aerosol Sci. Technol.  32: 15-25.
 41       Wang, B.; Peng, Z.; Zhang, X.;  Xu, Y.; Wang, H.; Allen, G.; Wang, L.; Xu, X. (1999) Particulate matter, sulfur
 42            dioxide, and pulmonary function in never-smoking adults in Chongqing, China. Int. J. Occup. Environ
 43            Healths: 14-19.
 44       Ward, D. J.; Miller, M. R.; Walters, S.; Harrison, R. M.;  Ayres, J. G. (2000) Impact of correcting peak flow for
 45            nonlinear errors on air pollutant effect estimates from a panel study. Eur. Respir. J. 15: 137-140.
 46       Ware, J. H.; Ferris, B. G., Jr.; Dockery, D. W.; Spengler, J. D.; Stram, D. O.; Speizer, F. E. (1986) Effects of
47            ambient sulfur oxides and suspended particles on respiratory health of preadolescent children. Am. Rev.
48            Respir. Dis. 133:  834-842.
49      Wichmann, H.-E.; Spix, C.; Tuch, T.; Wolke, G.; Peters,  A.; Heinrich, J.; Kreyling, W. G.; Heyder, J.  (2000) Daily
 50            mortality and fine and ultrafine particles in Erfurt, Germany. Part I: role of particle number and particle mass.
51            Cambridge, MA:  Health Effects Institute; Research Report no. 98.
52      Wilson, W. E.; Suh, H. H. (1997) Fine particles and coarse particles: concentration relationships relevant to
53            epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
         March 2001                                    6-285         DRAFT-DO NOT QUOTE OR CITE

-------
  1      Wilson, W. E.; Mage, D. T.; Grant, L. D. (2000) Estimating separately personal exposure to ambient and
  2            nonambient participate matter for epidemiology and risk assessment: why and how. J. Air Waste Manage.
  3            Assoc. 50: 1167-1183.
  4      Wolff, G. T.; Stroup, C. M.; Stroup, D. P. (1983) The coefficient of haze as a measure of particulate elemental
  5            carbon. J. Air Pollut. Control Assoc. 33: 746-750.
  6      Wong, C. M.; Lam, T. H.; Peters, J.; Hedley, A. J.; Ong, S. G.; Tarn, A. Y. C.; Liu, J.; Spiegelhalter, D. J. (1998)
  7            Comparison between two districts of the effects of an air pollution intervention on bronchial responsiveness
  8            in primary school children in Hong Kong. J. Epidemiol. Community Health 52: 571-578.
  9      Wong, C. M.; Hu, Z. G.;  Lam, T. H.; Hedley, A. J.; Peters, J. (1999) Effects of ambient air pollution and
 10            environmental tobacco smoke on respiratory health of non-smoking women in Hong Kong. Int. J. Epidemiol.
 11            28: 859-864.
 12      Woodruff, T. J.; Grille, J.; Schoendorf, K. C. (1997) The relationship between selected causes of postneonatal
 13            infant mortality and particulate air pollution in the United States. Environ. Health Perspect. 105: 608-612.
 14      Wordley, J.; Walters, S.;  Ayres, J. G. (1997) Short term variations in hospital admissions and mortality and
 15            particulate air pollution. Occup. Environ. Med. 54:  108-116.
 16      Xu, Z.; Yu, D.; Jing, L.; Xu, X. (2000) Air pollution and daily mortality in Shenyang, China. Arch. Environ. Health
 17            55:115-120.
 18      Yang, W.; Jennison, B. L.; Omaye, S. T. (1997) Air pollution and asthma emergency room visits in Reno, Nevada.
 19            Inhalation Toxicol. 9: 15-29.
 20      Ye, S.-H.; Zhou, W.; Song, J.; Peng, B.-C.; Yuan, D.; Lu, Y.-M.; Qi, P.-P. (2000) Toxicity and health effects of
 21            vehicle emissions in Shanghai.  Atmos. Environ. 34: 419-429.
 22      Zanobetti, A.; Schwartz, J. (2000) Race, gender, and social status as modifiers of the effects of PM10 on mortality.
 23            J. Occup. Environ. Med. 42: 469-474.
 24      Zanobetti, A.; Wand, M.  P.; Schwartz, J.; Ryan, L. M. (2000a) Generalized additive distributed lag models:
 25            quantifying mortality displacement. Biostatistics 1: 279-292.
 26      Zanobetti, A.; Schwartz, J.; Dockery, D. W. (2000b) Airborne particles are a risk factor for hospital admissions for
 27            heart and lung disease. Environ. Health Perspect. 108: 1071-1077.
 28      Zeger, S. L.; Dominici, F.; Samet, J. (1999) Harvesting-resistant estimates of air pollution effects on mortality.
 29            Epidemiology 10:  171-175.
 30      Zeger, S. L.; Thomas, D.; Dominici, F.; Samet, J. M.; Schwartz, J.; Dockery, D.; Cohen, A. (2000) Exposure
 31            measurement error in time-series studies of air pollution: concepts and consequences. Environ. Health
 32            Perspect. 108: 419-426.
 33      Zeghnoun, A.; Beaudeau, P.; Carrat, F.;  Delmas, V.; Boudhabhay, O.; Gayon, F.; Guincetre, D.; Czernichow, P.
 34            (1999) Air pollution and respiratory drug sales in the city of Le Havre, France, 1993-1996. Environ. Res.
 35            81:224-230.
 36      Zemp, E.; Elsasser, S.; Schindler, C.; Kiinzli, N.; Perruchoud, A. P.; Domenighetti, G.; Medici, T.;
 37            Ackermann-Liebrich, U.; Leuenberger, P.; Monn, C.; Bolognini, G.; Bongard, J.-P.; Brandli, O.; Karrer, W.;
 38            Keller, R.; Schoni, M. H.; Tschopp, J.-M.; Villiger,  B.; Zellweger, J.-P.; SAPALDIA Team. (1999)
 39            Long-term ambient air pollution and respiratory symptoms in adults (SAPALDIA study). Am. J. Respir. Crit.
40            Care Med. 159: 1257-1266.
41      Zhang, J.; Qian, Z.; Kong, L.; Zhou, L.; Yan, L.; Chapman, R. S. (1999) Effects of air pollution on respiratory
42            health of adults in three Chinese cities. Arch. Environ. Health 54: 373-381.
43      Zhang, H.; Triche, E.; Leaderer, B. (2000) Model for the analysis of binary time series of respiratory symptoms.
44            Am.J. Epidemiol.  151: 1206-1215.
45      Zidek, J. V.; Wong, H.; Le, N. D.;  Burnett, R. (1996) Causality, measurement error and multicollinearity in
46            epidemiology. Environmetrics 7: 441-451.
47      Zmirou, D.; Schwartz, J.; Saez, M.; Zanobetti, A.;  Wojtyniak, B.; Touloumi, G.; Spix, C.; Ponce de Leon, A.;
48            Le Moullec, Y.; Bacharova, L.; Schouten, J.; Ponka, A.; Katsouyanni, K. (1998) Time-series analysis of air
49            pollution and cause-specific mortality. Epidemiology 9: 495-503.
         March 2001                                    6-286        DRAFT-DO NOT QUOTE OR CITE

-------
i                            APPENDIX 6A
2
3
4               Demographic and Pollution Data for 90-City Analysis
                            of NMMAPS Project
6
7
    March 2001                       6A-1      DRAFT-DO NOT QUOTE OR CITE

-------
2
P
3
£3"
O
O











ON
K>



O
>
Tl
H
1
O
o
^
0
H
c
o
H
W
O
J*3

H- -t
H
m
TABLE 6A-1. THE 90 CITIES AND THEIR INCLUDED COUNTIES BY POPULATION SIZE WITH MEAN DAILY
NUMBER OF DEATHS BY CATEGORY (1987-1994). Evaluated in NMMAPS 90-Cities Analyses, Samet et al. (2000a,b).
CVD/Respiratory
City
Los Angeles
New York

Chicago
Dallas/Fort Worth
Houston
San Diego
Santa Ana/ Anaheim
Phoenix
Detroit
Miami
Philadelphia
Minneapolis/St. Paul
Seattle
San Jose
Cleveland
San Bernardino
Pittsburgh

Oakland

Atlanta
San Antonio

Riverside
Denver

Sacramento

St. Louis
Buffalo



Abbreviation
la
ny

chic
dlft
hous
sand
staa
phoe
del
miam
phil
minn
seat
sanj
clev
sanb
pitt

oakl

atla
sana

rive
denv

sacr

stlo
buff



County
Los Angeles
Bronx, Kings, New York,
Richmond, Queens, Westchester
Cook
Collin, Dallas, Rockwall, Tarrant
Harris
San Diego
Orange
Maricopa
Wayne
Dade
Philadelphia
Hennepin, Ramsey
King
Santa Clara
Cuyahoga
San Bernardino
Allegheny

Alameda

Fulton, De Kalb
Bexar

Riverside
Denver, Adams, Arapahoe

Sacramento

St. Louis City
Erie



State
CA
NY

IL
TX
TX
CA
CA
AZ
MI
FL
PA
MN
WA
CA
OH
CA
PA

CA

GA
TX

CA
CO

CA

MO
NY



Population
8,863,164
8,197,430

5,105,067
3,312,553
2,818,199
2,498,016
2,410,556
2,122,101
2,111,687
1,937,094
1,585,577
1,518,196
1,507,319
1,497,577
1,412,140
1,418,380
1,336,449

1,279,182

1,194,788
1,185,394

1,170,413
1,124,159

1,041,219

993,529
968,532



Total
148.1
190.9

113.9
47.9
39.9
41.6
32.4
38.4
46.9
43.8
42.3
26.3
25.6
19.7
36.5
20.6
37.6

22.2

17.5
20.1

20.1
9.1

17.2

10.7
25.2



Disease
87.0
108.3

62.0
26.0
20.0
22.6
18.7
20.9
26.5
23.6
21.5
13.9
13.4
10.7
20.1
12.1
21.0

12.2

8.8
10.5

12.4
5.0

9.5

6.0
14.8



Other
61.1
82.6

51.9
21.9
19.8
19.0
13.6
17.5
20.4
20.2
20.8
12.4
12.2
9.0
16.4
8.5
16.9

10.0

8.7
9.6

7.7
4.1

7.7

4.7
10.3




-------
P3
cr
0
0










>
1



a
£>
"Tl
H
a
o
2;
o
H
O
c
o
H
W
O
n
DAILY NUMBER OF DEATHS BY CATEGORY (1987-1994). Evaluated in NMMAPS 90-Cities Analyses,
Samet et al. (2000a,b).
City
Columbus
Cincinnati
St. Petersburg
Kansas City
Honolulu
Tampa
Memphis
Indianapolis
Newark
Baltimore
Salt Lake City
Rochester
Worcester
Orlando
Jacksonville
Fresno
Louisville

Boston
Birmingham
Washington

Oklahoma City
Providence
El Paso
Tacoma
Austin
Abbreviation County
clmo
cine
stpe
kan
hono
tamp
memp
indi
nwk
bait
salt
roch
wor
orla
jckv
fres
loui

bost
birm
dc

okla
prov
elpa
taco
aust
Franklin
Hamilton
Pinellas
Clay, Jackson, Platte
Honolulu
Hillsborough
Shelby
Marion
Essex
Baltimore City
Salt Lake
Monroe
Worcester
Orange
Duval
Fresno
Jefferson

Suffolk
Jefferson
Washington DC

Oklahoma
Providence
El Paso
Pierce
Travis
State
OH
OH
FL
MO
HI
FL
TN
IN
NJ
MD
UT
NY
MA
FL
FL
CA
KY

MA
AL
DC

OK
RI
TX
WA
TX
Population
961,437
866,228
851,659
844,510
836,231
834,054
826,330
797,159
778,206
736,014
725,956
713,968
709,705
677,491
672,971
667,490
664,937

663,906
651,525
606,900

599,611
596,270
591,610
586,203
576,407
CVD/Respiratory
Total Disease
16.8
19.9
29.3
16.7
11.9
16.9
17.5
16.9
18.4
20.2
9.3
14.6
15.2
11.0
13.0
11.1
16.3

13.2
16.2
15.5

12.9
14.6
7.7
10.0
7.0
8.9
11.0
17.7
9.3
6.4
9.1
9.7
9.0
8.7
9.8
4.9
7.9
8.2
5.8
7.0
6.2
8.8

6.5
8.5
7.0

7.3
7.9
3.8
5.7
3.4
Other
7.9
8.9
11.6
7.5
5.5
7.8
7.7
8.0
9.7
10.4
4.4
6.7
6.9
5.2
6.0
4.9
7.5

6.7
7.7
8.5

5.6
6.7
3.9
4.3
3.6

-------
p
o
       TABLE 6A-1 (cont'd). THE
            DAILY NUMBER OF
90 CITIES AND THEIR INCLUDED COUNTIES BY POPULATION SIZE WITH MEAN
DEATHS BY CATEGORY (1987-1994). Evaluated in NMMAPS 90-Cities Analyses,
bo
o
I— t











ON
•^



O
^
H
6
o
z
0
H
O
G
0
H
W
O
o
H
w


aanu
;i ei ai. (zuuua,c
>;•



CVD/Respiratory
City
Dayton
Jersey City
Bakersfield
Akron
Charlotte
Nashville
Tulsa
Grand Rapids
New Orleans
Stockton
Albuquerque
Syracuse
Toledo
Raleigh
Wichita
Colorado Springs
Baton Rouge
Modesto
Madison
Spokane
Little Rock
Greensboro
Knoxville

Shreveport
Des Moines


Abbreviation
dayt
jers
bake
akr
char
nash
tuls
gdrp
no
stoc
albu
syra
tole
ral
wich
colo
batr
mode
madi
spok
Itrk
grnb
knox

shr
desm


County
Montgomery
Hudson
Kern
Summit
Mecklenburg
Davidson
Tulsa
Kent
Orleans
San Joaquin
Bernalillo
Onondaga
Lucas
Wake
Sedwick
El Paso
East Baton Rouge
Stanislaus
Dane
Spokane
Pulaski
Guilford
Knox

Bossier, Caddo
Polk


State
OH
NJ
CA
OH
NC
TN
OK
MI
LA
CA
NM
NY
OH
NC
KS
CO
LA
CA
WI
WA
AR
NC
TN

LA
IA


Population
573,809
553,099
543,477
514,990
511,433
510,784
503,341
500,631
496,938
480,628
480,577
468,973
462,361
423,380
403,662
397,014
380,105
370,522
367,085
361,364
349,660
347,420
335,749

334,341
327,140


Total
11.9
11.5
8.6
10.7
8.5
11.0
10.0
8.7
12.0
8.5
7.6
9.7
10.8
5.6
7.2
5.0
6.3
6.6
5.3
7.8
7.0
6.9
6.7

6.8
6.1


Disease
6.5
5.9
5.0
5.8
4.3
6.0
5.8
4.9
5.9
4.8
3.8
5.4
6.3
2.9
4.0
2.8
3.4
3.8
2.9
4.5
3.7
3.8
3.5

3.7
3.4


Other
5.4
5.6
3.6
4.9
4.2
5.0
4.2
3.8
6.1
3.6
3.8
4.3
4.5
2.7
3.3
2.3
3.0
2.8
2.4
3.3
3.3
3.1
3.1

3.1
2.6



-------
o
TABLE 6A-1 (cont'd). THE 90 CITIES AND THEIR INCLUDED COUNTIES BY POPULATION SIZE WITH MEAN
     DAILY NUMBER OF DEATHS BY CATEGORY (1987-1994). Evaluated in NMMAPS 90-Cities Analyses,
                                      Samet et al. (2000a,b).
'•w'
O











ON
i
L/1



O
£
H
1
D
0
O
H
0
d
0
H
W
O
7*
o
H
m
CVD/Respiratory
City
Fort Wayne
Corpus Chrisit
Norfolk
Jackson
Hunts ville
Anchorage
Lexington
Lubbock
Richmond
Arlington
Kingston

Evansville
Kansas City
Olympia
Topeka


















Abbreviation
ftwa
corp
nor
jcks
hunt
anch
lex
lubb
rich
arlv
king

evan
kans
olym
tope


















County
Allen
Nueces
Norfolk
Hinds
Madison
Anchorage
Fayette
Lubbock
Richmond City
Arlington
Ulster

Vanderburgh
Wyandotte
Thurston
Shawnee


















State
IN
TX
VA
MS
AL
AK
KY
TX
VA
VA
NY

IN
KS
WA
KS


















Population
300,836
291,145
261,229
254,441
238,912
226,338
225,366
222,636
203,056
170,936
165,304

165,058
161,993
161,238
160,976


















Total
5.9
4.9
4.8
5.3
3.9
1.9
4.1
3.9
5.1
2.4
3.0

4.4
3.2
2.8
3.6


















Disease
3.4
2.5
2.6
3.0
2.2
0.8
2.1
2.3
2.7
1.3
1.8

2.5
1.8
1.5
2.0


















Other
2.5
2.4
2.2
2.3
1.7
1.1
2.0
1.6
2.4
1.2
1.2

1.9
1.4
1.3
1.6



















-------
      TABLE 6A-2. MEAN DAILY POLLUTION LEVELS BY CITY (1987-1994)
           Evaluated in NMMAPS 90-Cities Analyses (Samet et al., 2000a,b)
City
Los Angeles
New York
Chicago
Dallas/Ft. Worth
Houston
San Diego
Santa Ana/Anaheim
Phoenix
Detroit
Miami
Philadelphia
Minneapolis/St. Paul
Seattle
San Jose
Cleveland
San Bernardino
Pittsburgh
Oakland
Atlanta
San Antonio
Riverside
Denver
Sacramento
St. Louis
Buffalo
Columbus
Cincinnati
St. Petersburg
Kansas City
Honolulu
Tampa
Memphis
Indianapolis
Newark
Baltimore
PM,o
46.0
28.8
35.6
23.8
30.0
33.6
37.4
40.3
40.9
25.7
35.4
26.9
25.3
30.4
45.1
37.0
31.6
26.3
36.1
23.8
52.0
29.6
33.3
30.1
21.7
29.0
34.2
23.5
25.9
15.3
28.3
30.3
32.0
32.9
32.9
03
ppb
22.8
19.6
18.6
25.3
20.5
31.6
23.0
22.5
22.6
25.9
20.5
NA
19.4
17.9
27.4
35.9
20.7
17.2
25.1
22.2
33.4
21.4
26.7
22.8
22.9
26.0
25.8
24.6
27.6
18.9
23.5
29.0
31.9
15.2
21.2
NO2
39.4
38.9
24.3
13.8
18.8
22.9
35.1
16.6
21.3
11.0
32.2
17.6
NA
25.1
25.2
27.9
27.6
21.2
26.0
NA
25.0
27.9
16.3
22.5
19.0
NA
26.7
11.8
9.2
NA
21.2
26.8
20.2
33.6
32.9
SO2
1.9
12.8
4.6
1.1
2.8
1.7
1.3
3.5
6.4
NA
9.9
2.6
NA
NA
10.3
0.7
14.2
NA
6.0
NA
0.4
5.5
NA
11.3
8.6
5.9
11.9
NA
2.4
NA
7.8
6.8
7.7
9.6
8.4
CO
ppm
15.1
20.4
7.9
7.4
8.9
11.0
12.3
12.7
6.6
10.6
11.8
11.8
17.8
9.4
8.5
10.3
12.2
9.1
8.9
10.1
11.2
10.3
9.4
10.5
7.3
7.6
10.0
7.1
6.2
8.3
7.8
11.9
9.0
8.7
9.2
March 2001                         6A-6       DRAFT-DO NOT QUOTE OR CITE

-------
   TABLE 6A-2 (cont'd). MEAN DAILY POLLUTION LEVELS BY CITY (1987-1994)
           Evaluated in NMMAPS 90-Cities Analyses (Samet et al., 2000a,b)
City
Salt Lake City
Rochester
Worcester
Orlando
Jacksonville
Fresno
Louisville
Boston
Birmingham
Washington DC
Oklahoma City
Providence
El Paso
Tacoma
Austin
Dayton
Jersey City
Bakersfield
Akron
Charlotte
Nashville
Tulsa
Grand Rapids
New Orleans
Stockton
Albuquerque
Syracuse
Toledo
Raleigh
Wichita
Colorado Springs
Baton Rouge
Modesto
Madison
Spokane
PM,o
/•ig/m3
32.9
21.9
22.2
22.7
29.9
43.4
30.8
26.0
31.2
28.2
25.0
30.9
41.2
28.0
21.1
27.4
30.5
53.2
22.4
30.7
32.4
26.6
22.8
29.0
39.0
16.9
24.5
25.6
25.6
25.6
26.3
27.3
41.7
19.9
36.0
03
ppb
23.0
22.7
30.0
24.1
28.2
29.4
19.8
17.9
22.4
17.5
28.4
25.4
24.4
23.8
25.5
26.6
19.7
33.3
30.5
29.3
16.2
31.4
27.7
20.5
22.6
25.8
23.7
27.1
35.4
24.2
24.3
21.2
26.1
29.7
32.6
NO2
yUg/m3
29.6
NA
25.2
11.4
14.8
21.7
22.4
29.9
NA
25.6
13.9
21.9
23.6
NA
NA
NA
28.7
19.4
NA
16.2
NA
16.6
NA
21.3
24.2
NA
NA
NA
12.7
NA
NA
16.6
24.2
NA
NA
SO2
/"g/m3
4.4
10.4
6.7
1.5
2.2
1.9
8.4
10.0
6.6
11.2
NA
9.5
9.1
6.5
NA
NA
10.7
3.0
12.0
NA
11.6
6.9
3.0
NA
1.7
NA
3.6
5.9
NA
4.8
NA
5.2
1.9
3.3
NA
CO
ppm
13.5
6.3
8.9
9.3
9.2
6.8
11.2
11.3
10.5
12.3
7.1
10.0
12.5
16.6
NA
8.2
20.1
10.5
7.0
11.1
11.2
6.5
5.7
9.4
8.2
7.9
11.7
10.3
16.1
6.5
10.9
4.3
9.1
10.4
21.9
March 2001                         6A-7      DRAFT-DO NOT QUOTE OR CITE

-------
TABLE 6A-2 (cont'd). MEAN DAILY POLLUTION LEVELS BY CITY (1987-1994)
        Evaluated in NMMAPS 90-Cities Analyses (Samet et al., 2000a,b)
City
Little Rock
Greensboro
Knoxville
Shreveport
Des Moines
Fort Wayne
Corpus Christi
Norfolk
Jackson
Huntsville
Anchorage
Lexington
Lubbock
Richmond
Arlington
Kingston
Evansville
Kansas City
Olympia
Topeka
PMIO
Aig/m3
25.8
27.5
36.3
24.7
35.5
23.2
24.7
26.0
26.4
26.0
23.0
24.5
25.1
25.4
22.0
20.4
32.4
43.4
22.7
29.0
03
ppb
27.7
NA
29.6
28.2
11.8
32.1
23.9
NA
23.9
30.4
NA
32.8
NA
NA
29.0
NA
NA
18.5
NA
NA
NO2
/"g/iri3
9.3
NA
NA
NA
NA
NA
NA
19.6
NA
12.9
NA
16.4
NA
23.7
25.5
NA
NA
17.6
NA
NA
S02
/•ig/m3
2.6
4.2
NA
2.3
NA
4.0
1.0
6.7
NA
NA
NA
6.2
NA
5.8
NA
NA
NA
4.7
NA
NA
CO
ppm
NA
12.2
13.6
NA
8.6
14.4
NA
7.3
7.9
6.3
16.1
8.8
NA
6.6
6.6
NA
NA
8.2
12.7
NA
March 200 1
6A-8
                                          DRAFT-DO NOT QUOTE OR CITE

-------
  TABLE 6A-3. NUMBER OF DAYS FOR WHICH MONITORING WAS AVAILABLE
     BY POLLUTANT FOR CITIES (1987-1994). Evaluated in 90-Cities NMMAPS
                        Analyses of Sametet al. (2000b)
City
Los Angeles
New York
Chicago
Dallas/Fort Worth
Houston
San Diego
Santa Ana/Anaheim
Phoenix
Detroit
Miami
Philadelphia
Minneapolis/St. Paul
Seattle
San Jose
Cleveland
San Bernardino
Pittsburgh
Oakland
Atlanta
San Antonio
Riverside
Denver
Sacramento
St. Louis
Buffalo
Columbus
Cincinnati
St. Petersburg
Kansas City
Honolulu
PM,0
580
489
2,683
624
793
521
480
376
1,348
484
495
2,764
2,205
945
1,298
538
2,899
508
482
670
545
1,645
488
487
489
1,564
1,705
367
670
415
03
2,922
2,922
2,922
2,922
2,922
2,922
2,922
2,554
1,861
2,882
2,901
0
1,820
2,922
1,712
2,922
2,883
2,922
2,200
2,918
2,922
2,922
2,922
1,731
2,884
1,494
1,712
2,920
2,856
1,681
NO2
2,922
2,493
2,922
2,557
2,557
2,922
2,922
740
2,686
2,863
2,554
2,725
0
1,957
2,555
2,922
2,537
2,921
2,922
0
2,904
2,484
2,916
2,919
2,522
0
2,554
2,235
2,922
0
SO2
2,922
2,920
1,409
2,908
2,922
2,922
2,922
1,272
2,922
0
2,919
2,914
0
0
2,922
2,922
2,922
0
2,918
0
2,908
2,860
0
2,919
2,922
964
2,905
0
1,094
0
CO
2,922
2,920
2,922
2,922
2,922
2,922
2,922
2,554
2,922
2,919
2,919
2,918
2,922
2,922
2,897
2,922
2,920
2,922
2,839
2,891
2,921
2,922
2,922
2,920
2,921
2,557
2,922
2,922
2,922
2,919
March 2001
6A-9
DRAFT-DO NOT QUOTE OR CITE

-------
    TABLE 6A-3 (cont'd). NUMBER OF DAYS FOR WHICH MONITORING WAS
    AVAILABLE BY POLLUTANT FOR CITIES (1987-1994). Evaluated in 90-Cities
                    NMMAPS Analyses of Samet et al. (2000b)
City
Tampa
Memphis
Indianapolis
Newark
Baltimore
Salt Lake City
Rochester
Worcester
Orlando
Jacksonville
Fresno
Louisville
Boston
Birmingham
Washington
Oklahoma City
Providence
El Paso
Tacoma
Austin
Dayton
Jersey City
Bakersfield
Akron
Charlotte
Nashville
Tulsa
Grand Rapids
New Orleans
Stockton
Albuquerque
Syracuse
Toledo
Raleigh
Wichita
PMIO
508
480
1,269
484
1,220
1,356
486
450
421
555
517
485
631
900
417
563
485
2,587
482
646
461
1,367
550
1,495
454
1,989
411
111
531
488
1,200
485
416
480
366
03
2,922
1,707
1,588
2,726
2,063
2,409
2,886
1,763
2,920
2,791
2,922
2,603
2,882
2,200
2,847
2,832
1,634
2,922
1,601
2,909
1,696
2,843
2,557
1,677
1,936
2,861
2,834
1,615
2,889
2,475
2,922
2,864
1,711
1,267
2,913
NO2
941
2,254
2,874
2,882
2,843
1,903
0
2,864
2,024
2,727
2,922
1,604
2,922
0
2,842
2,295
2,441
2,472
0
0
0
2,496
2,557
0
1,593
0
2,462
0
2,879
2,379
0
0
0
1,219
0
SO2
1,818
2,823
2,922
2,896
2,912
2,739
2,921
2,452
2,878
2,738
2,398
2,841
2,922
1,916
2,286
0
2,922
2,906
2,756
0
0
2,918
2,557
2,827
0
2,619
2,426
2,907
0
867
0
2,857
2,921
0
1,423
CO
2,922
2,922
2,922
2,894
2,865
2,922
2,921
2,899
2,921
2,922
2,922
2,922
2,922
2,922
2,341
2,909
2,921
2,922
2,766
0
2,922
2,883
2,659
2,922
2,922
2,771
2,836
2,903
2,922
2,906
2,922
2,908
2,897
2,160
2,922
March 2001
6A-10
DRAFT-DO NOT QUOTE OR CITE

-------
    TABLE 6A-3 (cont'd). NUMBER OF DAYS FOR WHICH MONITORING WAS
    AVAILABLE BY POLLUTANT FOR CITIES (1987-1994). Evaluated in 90-Cities
                    NMMAPS Analyses of Samet et al. (2000b)
City
Colorado Springs
Baton Rouge
Modesto
Madison
Spokane
Little Rock
Greensboro
Knoxville
Shreveport
Des Moines
Fort Wayne
Corpus Christi
Norfolk
Jackson
Hunts ville
Anchorage
Lexington
Lubbock
Richmond
Arlington
Kingston
Evansville
Kansas City
Olympia
Topeka
PM10
481
474
199
338
2,393
516
445
577
349
1,334
336
613
474
508
1,382
2,379
816
1,306
474
313
323
404
551
1,135
269
03
2,920
2,922
2,496
1,698
974
2,922
0
1,679
2,922
2,782
1,587
2,919
0
2,191
2,173
0
1,709
0
0
1,705
0
0
2,890
0
0
NO2
0
2,880
2,449
0
0
2,921
0
0
0
0
0
0
1,787
0
1,090
0
2,871
0
2,537
2,306
0
0
324
0
0
SO2
0
2,891
845
2,432
0
2,908
1,077
0
2,881
0
1,219
2,920
2,148
0
0
0
2,906
0
2,907
0
0
0
2,909
0
0
CO
2,922
2,888
2,892
2,709
2,922
0
1,855
2,511
0
2,825
1,822
0
2,921
2,574
2,532
1,488
2,865
0
2,922
2,896
0
0
2,775
950
0
March 2001
6A-11
DRAFT-DO NOT QUOTE OR CITE

-------
REFERENCES

Samet, J. M.; Dominici, F.; Zeger, S. L.; Schwartz, J.; Dockery, D. W. (2000a) National morbidity, mortality, and
      air pollution study. Part I: methods and methodologic issues. Cambridge, MA: Health Effects Institute;
      research report no. 94.
Samet, J. M.; Zeger, S. L.; Dominici, F.; Curriero, F.; Coursac, I.; Dockery, D. W.; Schwartz, J.; Zanobetti, A.
      (2000b) The national morbidity, mortality, and air pollution study. Part II: morbidity, mortality, and air
      pollution in the United States. Cambridge, MA: Health Effects Institute; research report no. 94.
March 2001                                 6A-12        DRAFT-DO NOT QUOTE OR CITE

-------
                         APPENDIX 6B





     Heart Rate Variability as a Predictor of Serious Cardiac Outcomes
March 2001                       6B-1      DRAFT-DO NOT QUOTE OR CITE

-------
  1           As an adjunct to discussion of newly emerging literature evaluating relationships between
  2      ambient PM and heart rate variability (HRV), factors affecting HR are reviewed briefly and
  3      summarized here.  More detail description of HRV, and its measurement and interpretation can
  4      be found elsewhere (1996 Task Force of the European Society of Cardiology and the North
  5      American Society of Pacing and Electrophysiology).
  6
  7      Factors Affecting Heart Rate. The heart has a spontaneous rhythm of approximately 100
  8      beats/mm in the absence of extrinsic influences, because the electrical signal triggering heartbeat
  9      originates in and spreads throughout the heart via a specialized conduction system. The tissue
 10      structures that comprise the conduction system of the heart include the sinoatrial (SA) node, the
 11      internodal pathways, the atrioventricular (AV) node, the bundle of His and its branches, and the
 12      Purkinje system. Although all parts of the conduction system are capable of spontaneous
 13      electrical discharge and heartbeat initiation, it is the SA node (with its higher rate of discharge)
 14      that is the normal cardiac pacemaker in the healthy heart.  The spontaneous discharge rate of the
 15      SA node, and therefore heartbeat, is modulated by nervous impulses and by circulating
 16      substances,  such as epinephrine that originate outside the heart. One category of modulating
 17      input to the  heart is through the sympathetic and parasympathetic divisions of the autonomic
 18      nervous system via numerous nerve fibers that innervate the heart.
 19           Stimulation of the heart via parasympathetic nerve fibers decreases the rate of discharge of
20      the SA node, thereby decreasing HR (bradycardia), and decreases the excitability of the AV
21      junctional fibers between the atrial musculature and the AV node, thereby  slowing transmission
22      of the impulse into the ventricles.  Stimulation of the heart via sympathetic fibers increases the
23      rate of discharge of the SA node, thereby increasing HR (tachycardia) and  increasing the
24      excitability of the AV node and increasing transmission of the cardiac impulse into the ventricle.
25      During the resting  state parasympathetic input to the heart predominates, so the normal resting
26      HR is well below the inherent rate of 100 beats/min.  The HR along with stroke volume
27      determines cardiac output, which interacts with peripheral resistance to determine blood pressure.
28           The autonomic control of HR is modulated by the vasomotor center located in the brain in
29      the reticular substance of the medulla and lower third of the pons. Impulses sent forth from the
30      vasomotor center through the parasympathetic and sympathetic neurons regulate HR and
31      vasomotor tone. The medial portion of the vasomotor center transmits inhibitory impulses that

        March 2001                               6B-2        DRAFT-DO NOT QUOTE OR CITE

-------
  1      decrease HR through the parasympathetic nerve fibers (vagus nerve).  The lateral portions of the
  2      vasomotor center transmit excitatory impulses that increase both HR and contractility through
  3      sympathetic nerve fibers to the heart. In this way, the vasomotor center can either increase or
  4      decrease HR, as well as vasomotor tone. The vasomotor center, in turn, is influenced by
  5      impulses arising in higher centers of the brain.
  6           Thus, HR is the resultant of the intrinsic rate of the heart modified by various internal and
  7      external factors. Most important of these is the output of the vasomotor center delivered via the
  8      autonomic nervous system. Other factors affecting HRV include exercise and changes in
  9      ambient temperature and oxygen tension.
 10
 11      Measures of Heart Rate Variability. Heart rate variability is being used increasingly in
 12      applications from basic research to clinical practice (Berntson et al., 1997).  Meaningful analysis
 13      of HRV is dependent on fidelity of the basic cardiac input signal that is derived from the
 14      electrocardiogram (ECG). This signal is digitized and a series of intervals between successive
 15      R (R-R) waves are determined.  The population of R-R intervals or pairs of R-R intervals are
 16      treated as if they were a set of temporarily unordered data.  The variability of these measures is
 17      expressed either by conventional statistical measures (Malik,  1995) or other analytical methods,
 18      whereby specific patterns of HRV may be related to specific physiological processes and
 19      mechanisms.
 20           A wide variety of estimates of HRV have been described. The Task Force of the European
 21      Society of Cardiology and the North American Society of Pacing and Electrophysiology (1996)
 22      has recommended standard time domain measures that index  overall heart rate variability,  short-
 23      term heart rate variability, and long-term heart rate variability (Table 9-9).  A measure
 24      recommended for HRV is the standard deviation of all normal-to-normal (N-N), also designated
 25      as R-R, heart beat intervals (SDNN). The recommended estimate of short-term variability is the
 26      root mean square of the successive beat differences (rMSSD) and that for longer-term variability
 27      is the standard deviation of the mean N-N interval for each 5-min segment of recording
28      (SDANN).
29           Periodic components of HRV tend to aggregate within several frequency domains (see
30      Table 6B-1). In young  healthy individuals at rest, the most conspicuous frequency band is  at the
31      normal respiratory frequency of 0.15 to 0.4 Hz and is termed high-frequency (HF) domain. The

        March 2001                              6B-3        DRAFT-DO NOT QUOTE OR CITE

-------
            TABLE 6B-1. TERMS USED IN EXPRESSING HEART RATE VARIABILITY
Abbreviation
HR
HRV
SDNN
rMSSD
SPANN
HF Domain
LF Domain
VLF Domain
Definition
Heart rate (beats/min)
Heart rate variability
Standard deviation of all normal to normal heart
beat intervals
Root mean square of successive beat differences
Standard deviation of the mean normal to normal
(N-N) interval for each 5 min
0.15 -0.4 HZ
0.05 -0.1 5 HZ
0.003 - 0.05 HZ
Domain
—
—
Time
Time
Time
Frequency
Frequency
Frequency
 1     band from 0.05 to 0.15 Hz is termed low-frequency (LF) domain. Other domains have been
 2     described including very low frequencies (VLF), 0.003 to 0.05 Hz, and ultra-low frequencies
 3     (ULF) that include circadian rhythms. Thus, HRV is quantitated by both time domain metrics
 4     (NN, SDNN, rMSSD, and SDANN) and frequency domain metrics (HF, LF, and ULF).
 5
 6     Factors Affecting Heart Rate Variability.  Heart rate variability after a myocardial infarction is
 7     associated with increased mortality (Kleiger et al., 1987). Aging and gender also are associated
 8     with depressed HRV (Umetani et al., 1998).  Reardon and Malik (1996) examined the affect of
 9     aging in healthy subjects (age range 40 to 102 years; 39 women) with normal resting ECGs.
10     In all subjects, 24-h Holter recordings were performed and used to  measure HRV.  The HRV
11     triangular index decreased significantly with age, whereas rMSSD  did not change. There was a
12     significant difference in HRV index in subjects >70 years compared with those <70 years. There
13     was no significant difference in rMSSD  between the two age groups. The authors conclude that
14     aging reduces HRV and decreased HRV may reflect reduced responsiveness of autonomic
15     activity to external environmental stimuli with age.
16
       March 2001
6B-4
DRAFT-DO NOT QUOTE OR CITE

-------
 1          Umetani et al. (1998) studied the effects of age and gender on 24-h HR and HRV in healthy
 2     subjects (10 to 99 years old; 112 male and 148 female).  The authors conclude that (1) HRV in
 3     healthy subjects declines with aging; (2) HRV of healthy subjects, particularly those >65 years
 4     old, may decrease to below levels associated with increased risk of mortality; (3) gender
 5     influences HRV (gender differences in HRV are age and measure dependent); and (4) age and
 6     gender also affect HRV.
 7          Tsuji et al. (1994) studied HRV in the original subjects of the Framingham Heart Study.
 8     Subjects  with transient or persistent nonsinus rhythm, 50% of recorded time, and those taking
 9     antiarrhythmic medications were excluded.  The associations between HRV measures and all-
10     cause mortality during 4 years of follow-up were assessed. A 1-SD decrement in low-frequency
11     power was associated with 1.70 times greater hazard for all-cause mortality (95% confidence
12     interval of 1.37 to 2.09). The authors concluded that estimation of HRV offers prognostic
13     information beyond that provided by the evaluation of traditional risk factors.
14
15
       March 2001                               6B-5        DRAFT-DO NOT QUOTE OR CITE

-------
  1       REFERENCES

  2       Berntson, G. G.; Bigger, J. T., Jr.; Eckberg, D. L.; Grossman, P.; Kaufmann, P. G.; Malik, M.; Nagaraja, H. N.;
  3             Forges, S. W.; Saul, J. P.; Stone, P. H.; Van Der Molen, M. W. (1997) Heart rate variability: origins,
  4             methods, and interpretive caveats. Psychophysiology 34: 623-648.
  5       Bigger, J. T., Jr.; Fleiss, J. L.; Steinman, R. C.; Rolnitzky, L. M.; Kleiger, R. E.; Rottman, J. N. (1992) Frequency
  6             domain measures of heart period variability and mortality after myocardial infarction. Circulation
  7             85: 164-171.
  8       Burnett, R. T.; Dales, R.; Krewski, D.; Vincent, R.; Dann, T.; Brook, J. R. (1995) Associations between ambient
  9             particulate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. Am. J. Epidemiol.
10             142: 15-22.
11       Godleski, J. J.; Sioutas, C.; Katler, M.; Koutrakis, P. (1996) Death from inhalation of concentrated ambient air
12             particles in animal models of pulmonary disease. Am. J. Respir. Crit. Care Med. 153: A15.
13       Godleski et al. (1998) Increased cardiac vulnerability during exposure to inhaled environmental particles [abstract].
14             Presented at: Health Effects Institute annual meeting; May; Boston, MA. Boston, MA: Health Effects
15             Institute.
16       Gold, D. R.; Litonjua, A.; Schwartz, J.; Lovett, E.; Larson, A.; Nearing, B.; Allen, G.; Verrier, M.; Cherry, R.;
17             Verrier, R. (2000) Ambient pollution and heart rate variability. Circulation 101:1267-1273.
18       Hayano, J.; Sakakibara, Y.; Yamada, M.; Ohte, N.; Fujinami, T.; Yokoyama, K.; Watanabe, Y.; Takata, K. (1990)
19             Decreased magnitude of heart rate spectral components in coronary artery disease. Its relation  to
20             angiographic severity. Circulation 81:  1217-1224.
21       Killingsworth, C. R.; Alessandrini, F.; Krishna Murthy, G. G.; Catalano, P. J.; Paulauskis, J. D.; Godleski, J. J.
22             (1997) Inflammation, chemokine expression, and death in monocrotaline-treated rats following fuel oil fly
23             ash inhalation. Inhalation Toxicol. 9: 541-565.
24       Kleiger, R. E.; Miller, J. P.; Bigger J. T. Jr.; Moss, A.  J.; Multicenter Post-infarction Research Group. (1987)
25             Decreased heart rate variability and its association with increased mortality after acute myocardial infarction.
26             Am. J. Cardiol. 59: 256-262.
27       Liao, D.; Creason, J.; Shy, C.; Williams, R.; Watts, R.; Zweidinger, R. (1999) Daily variation of particulate air
28             pollution and poor cardiac autonomic control in the elderly. Environ. Health Perspect. 107: 521-525.
29       Malik, M. (1995) Graphical representation of circadian patterns of heart rate variability components.  Pacing Clin.
30             Electrophysiol. 18: 1575-1580.
31       Martin, G. J.; Magid, N. M.; Myers, G.; Barnett, P. S.; Schaad, J. W.; Weiss, J. S.; Lesch, M.; Singer, D. H. (1987)
32             Heart rate variability and sudden death secondary to coronary artery disease during ambulatory
33             electrocardiographic monitoring. Am. J. Cardiol. 60: 86-89.
34       Morris, R. D.; Naumova, E. N.; Munasinghe, R. L. (1995) Ambient air pollution and hospitalization for congestive
35             heart failure among elderly people in seven large US cities. Am. J. Public Health 85: 1361-1365.
36       Pope, C.  A., Ill; Verrier, R. L.; Lovett, E. G.; Larson, A. C.; Raizenne, M. E.; Kanner, R. E.; Schwartz, J.; Villegas,
37             G. M.; Gold, D. R.; Dockery, D. W. (1999) Heart rate variability associated with particulate air pollution.
38             Am. Heart J. 138: 890-899.
39       Reardon, M.;  Malik, M. (1996) Changes in heart rate variability with age. Pacing Clin. Electrophysiol.
40             19: 1863-1866.
41       Schwartz, J. (1997) Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology
42             8:371-377.
43       Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit,
44             Michigan. Am. J. Epidemiol. 142: 23-35.
45       Singer, D. H.; Martin, G. J.; Magid, N.; Weiss, J. S.; Schaad, J. W.; Kehoe, R.; Zheutlin, T.; Fintel, D. J.;
46             Hsieh, A.-M.; Lesch,  M. (1988) Low heart rate variability and sudden cardiac death. J. Electrocardiol.
47             21(suppl.):S46-S55.
48       Task Force of the European  Society of Cardiology and the North American Society of Pacing and
49             Electrophysiology. (1996) Heart rate variability: standards of measurement, physiological interpretation and
50             clinical use. Circulation 93: 1043-1065.
51       Tsuji, H.; Venditti, F. J.; Manders, E. S.; Evans, J. C.; Larson, M. G.; Feldman, C. L.; Levy, D. (1994) Reduced
52             heart rate variability and mortality risk in an elderly cohort: The Framingham Heart Study. Circulation
53             90: 878-883.
         March 2001                                    6B-6         DRAFT-DO NOT QUOTE OR CITE

-------
1      Umetani, K.; Singer, D. H.; McCraty, R.; Atkinson, M. (1998) Twenty-four hour time domain heart rate variability
2            and heart rate: relations to age and gender over nine decades. J. Am. Coll. Cardiol. 31: 593-601.
3      Watkinson, W. P.; Campen, M. J.; Costa, D. L. (1998) Cardiac arrhythmia induction after exposure to residual oil
4            fly ash particles in a rodent model of pulmonary hypertension. Toxicol. Sci. 41: 209-216.
      March 2001                                   6B-7         DRAFT-DO NOT QUOTE OR CITE

-------

-------
  i             7. DOSIMETRY OF PARTICULATE MATTER
  2
  3
  4     7.1  INTRODUCTION
  5          A basic principle in health effects evaluation is that the dose delivered to the target site of
  6     concern, rather than the external exposure, is the proximal cause of any biological response.
  7     Characterization of the exposure-dose-response continuum for particulate matter (PM), a
  8     fundamental objective of any dose-response assessment for evaluation of health effects, requires
  9     the elucidation and understanding of the mechanistic determinants of inhaled particle dose,
 10     which is dependent initially on the deposition of particles within the respiratory tract. Particle
 11     deposition refers to the removal of particles from their airborne state because of their
 12     aerodynamic or thermodynamic behavior in air. Once particles have deposited onto the surfaces
 13     of the respiratory tract, they subsequently will be subjected to either absorptive or nonabsorptive
 14     particulate removal processes. This may result in their removal  from airway surfaces, as well as
 15     their removal to various degrees from the respiratory tract. Particulate matter translocated from
 16     initial deposition sites is said to have undergone clearance. Clearance of deposited particles
 17     depends upon the initial site of deposition and upon the physicochemical properties of the
 18     particles, both of which impact upon specific translocation mechanisms.  Retained particle
 19     burdens are determined by the dynamic relationship between deposition and clearance
 20     mechanisms.
 21           This chapter is concerned with particle dosimetry, the study of the deposition, translocation,
 22     clearance and retention of particles within the respiratory tract and extrapulmonary tissues.
 23      It summarizes basic concepts as presented in the 1996 EPA document,  Air Quality Criteria for
 24      Particulate Matter or "PM AQCD" (U.S. Environmental Protection Agency, 1996), specifically
 25      in Chapter 10; and it updates the state of the science based upon  new literature on particle
 26      deposition, clearance and retention appearing since publication of the 1996 PM AQCD.
 27      Although the basic mechanisms governing deposition and clearance of inhaled particles have not
28      changed, there has been significant additional information on the role of certain biological
29      determinants of the  deposition/clearance process, such as gender and age. Also, the
       March 2001                                7-1         DRAFT-DO NOT QUOTE OR CITE

-------
 1      understanding of regional dosimetry and the particle size range over which this has been
 2      evaluated has been expanded.
 3           The dose from inhaled particles deposited and retained in the respiratory tract is governed
 4      by a number of factors. These include exposure concentration and exposure duration, respiratory
 5      tract anatomy and ventilatory parameters, and by physicochemical properties of the particles
 6      themselves (e.g., particle size, hygroscopicity, solubility). The basic characteristics of particles
 7      as they relate to deposition and retention, as well as anatomical and physiological factors
 8      influencing particle deposition and retention, were discussed in depth in the  1996 PM AQCD.
 9      Thus, in this current chapter, only an overview of basic information related to one critical factor
10      in deposition, namely particle size, is provided (Section 7.1.1), so as to allow the reader to
11      understand the different terms used in the remainder of this chapter and subsequent ones dealing
12      with health effects. This is followed, in Section 7.1.2, by a basic overview of respiratory tract
13      structure as it relates to deposition evaluation.  The ensuing major sections of this chapter then
14      provide updated information on particle deposition, clearance, and retention in the respiratory
15      tract of humans, as well as laboratory animals, which are useful in the evaluation of PM health
16      effects.  Issues related to the phenomenon of particle overload as it may apply to human exposure
17      and the use of instillation as an exposure technique to evaluate PM health effects also are
18      discussed. The final sections of the chapter deal with mathematical models of particle
19      disposition in the respiratory tract.
20           It must be emphasized that any dissection into discrete topics of factors that control dose
21      from inhaled particles tends to mask the dynamic and interdependent nature  of the intact
22      respiratory system. For example, although deposition is discussed separately from clearance
23      mechanisms, retention (i.e., the actual amount of particles found in the respiratory tract at any
24      point in time) is determined by the relative rates of both deposition and clearance. Thus,
25      assessment of overall dosimetry requires integration of these various components of the overall
26      process.
27
28      7.1.1  Size Characterization of Inhaled Particles
29           Information about particle size distribution is important in the evaluation of effective
30      inhaled dose. This section summarizes particle attributes requiring characterization and provides
31      general definitions important in understanding particle fate within the respiratory tract.
        March 2001                                 7-2         DRAFT-DO NOT QUOTE OR CITE

-------
  1           Particles exist in the atmosphere as aerosols, which are airborne suspensions of finely
  2      dispersed solid or liquid particles.  Because aerosols can consist of almost any material, their
  3      description in simple geometric terms can be misleading unless important factors relating to
  4      constituent particle size, shape, and density are considered. Although the size of particles within
  5      aerosols can be described based on actual physical measurements (such as those obtained with a
  6      microscope), in many cases it is better to use some equivalent diameter in place of the physical
  7      diameter.  The most commonly used metric is aerodynamic equivalent diameter (AED), whereby
  8      particles of differing geometric size, shape and density are compared in terms of aerodynamic
  9      behavior (i.e., terminal setting velocity) of particles that are unit density (1 gm/cm3) spheres. The
 10      aerodynamic behavior of unit density spherical particles constitutes a useful standard by which
 11      many types of particles can be compared in terms of certain deposition mechanisms.
 12           It is important to note that aerosols present in natural and work environments have
 13      polydisperse size distributions. This means that the constituent particles within an aerosol have a
 14      range of sizes and are more appropriately described in terms of a size distribution parameter.
 15      The lognormal distribution (i.e., the situation whereby the logarithms of particle diameter are
 16      distributed normally) can be used for describing size distributions of most aerosols.  In linear
 17      form, the logarithmic mean is the median of the distribution, and the metric of variability around
 18      this central tendency is the geometric standard deviation (og). The og, a dimensionless term, is
 19      the ratio of the 84th (or 16th) percentile particle size to the 50th percentile size. Thus, the only
 20      two parameters needed to describe  a log normal distribution of aerosol particle sizes are the
 21      median diameter and the geometric standard deviation. However, the actual size distribution
 22      may be obtained in  various ways. For example, when a distribution is described by counting
 23      particles, the median is called the count median diameter (CMD). On the other hand, the median
 24      of a distribution based  on particle mass in an aerosol is the mass median diameter (MMD).
 25      When using aerodynamic diameters, a term  that is encountered frequently is mass median
 26      aerodynamic diameter  (MMAD), which refers to the median of the distribution of mass with
 27      respect to aerodynamic equivalent diameter. Most of the present discussion will  focus on
28      MMAD because it is the most commonly used measure of aerosol distribution. However,
29      alternative distributions should be used for particles with actual physical size below about
30      0.5 yum, because, for these, aerodynamic properties become less important.  One such metric is


        March 2001                                7-3         DRAFT-DO NOT QUOTE OR CITE

-------
  1      thermodynamic-equivalent size, which is the diameter of a spherical particle that has the same
  2      diffusion coefficient in air as the particle of interest.
  3
  4      7.1.2  Structure of the Respiratory Tract
  5           Detailed discussion of respiratory tract structure was provided in the 1996 PM AQCD (U.S.
  6      Environmental Protection Agency, 1996), and only a brief synopsis is presented here.
  7      For dosimetry purposes, the respiratory tract can be divided into three regions: (1) extrathoracic
  8      (ET), (2) tracheobronchial (TB), and (3) alveolar (A). The ET region consists of head airways
  9      (i.e., nasal or oral passages) through the larynx and represents the areas through which inhaled air
10      first passes. In humans, inhalation can occur through the nose or mouth (or both, known as
11      oronasal breathing).  However, most laboratory animals commonly used in respiratory
12      toxicological studies are obligate nose breathers.
13           From the ET region, inspired air enters the TB region at the trachea.  From the level of the
14      trachea, the conducting airways then undergo branching for a number of generations. The
15      terminal bronchiole is the most peripheral of the distal conducting airways and these lead,
16      in humans, to the respiratory bronchioles, alveolar ducts, alveolar sacs and alveoli (all of which
17      comprise the A region). All of the conducting airways, except the trachea and portions of the
18      mainstem bronchi, are surrounded by parenchymal tissue. This is composed primarily of the
19      alveolated structures of the A region and associated blood and lymphatic vessels.  It should be
20      noted that these respiratory tract regions are comprised of various cell types, and that there are
21      distinct differences in the cellular composition of the ET, TB, and A regions. Although a
22      discussion of cellular structure of the respiratory tract is beyond the scope of this section, details
23      may be found in a number of sources (e.g., Crystal et al., 1997).
24
25
26      7.2 PARTICLE DEPOSITION
27          This section discusses the deposition of particles in the respiratory tract. It begins with an
28      overview of the basic physical mechanisms that govern deposition. This is followed by an
29      update on both total respiratory tract and regional deposition patterns in humans. Some critical
        March 2001                                7-4        DRAFT-DO NOT QUOTE OR CITE

-------
  1     biological factors that may modulate deposition are then presented. The section ends with a
  2     discussion of new information related to interspecies patterns of particle deposition.
  3
  4     7.2.1  Mechanisms of Deposition
  5          Particles may deposit within the respiratory tract by five mechanisms: (1) inertial
  6     impaction, (2) sedimentation, (3) diffusion, (4) electrostatic precipitation, and (5) interception.
  7          Sudden changes in airstream direction and velocity cause particles to fail to follow the
  8     streamlines of airflow. As a consequence, the particles contact, or impact, onto airway surfaces.
  9     The ET and upper TB airways are characterized by high air velocities and sharp directional
 10     changes and, thus, dominate as sites of inertial impaction.  Impaction is a significant deposition
 11     mechanism for particles larger than 1 /^m AED.
 12          All aerosol particles are continuously influenced by gravity, but particles with an AED
 13     >0.5 /urn are affected to the greatest extent. A particle will acquire a terminal settling velocity
 14     when a balance is achieved between the acceleration of gravity acting on the particle and the
 15     viscous resistance of the air,  and it is this settling out of the airstream that takes it into contact
 16     with airway surfaces.  Both sedimentation and inertial impaction can influence the deposition of
 17     particles within the same size range. These deposition processes act together in the ET and TB
 18     regions, with inertial impaction dominating in the upper airways and gravitational settling
 19     becoming increasingly dominant in the lower conducting airways, especially for the largest
 20     particles, which can penetrate into the smaller bronchi.
 21          Particles having actual  physical diameters <1 //m are subjected increasingly to diffusive
 22     deposition because of random bombardment by air molecules, which results in contact with
 23     airway surfaces.  The root mean square displacement that a particle experiences in a unit of time
 24     along a given cartesian coordinate is a measure of its diffusivity. The density of a particle is
 25     unimportant in determining particle diffusivity. Thus, instead of having an aerodynamic
 26     equivalent size, diffusive particles of different shapes can be related to the diffusivity of a
 27     thermodynamic equivalent size based on spherical particles.
28          The particle size region around 0.3 to 0.5 ^m frequently is described as consisting of
29     particles that are small enough to be minimally influenced by impaction or sedimentation and
 30     large enough to be minimally influenced by diffusion.  Such particles are the most persistent in
31      inhaled air and undergo the lowest extent of deposition in the respiratory tract.
        March 2001                                 7-5         DRAFT-DO NOT QUOTE OR CITE

-------
  1           Interception is deposition by physical contact with airway surfaces.  The interception
  2      potential of any particle depends on its physical size, and fibers are the chief concern in relation
  3      to the interception process. Their aerodynamic size is determined predominantly by their
  4      diameter, rather than their length.
  5           Electrostatic precipitation is deposition related to particle charge. The minimum charge an
  6      aerosol particle can have is zero when it is electrically neutral. This condition rarely is achieved
  7      because of the random charging of aerosol particles by air ions. Aerosol particles will acquire
  8      charges from these ions by collisions with them because of their random thermal motion.
  9      Furthermore, many laboratory generated aerosols are charged. Such aerosols will lose their
10      charge  slowly as they attract oppositely charged ions.  An equilibrium state of these competing
11      processes eventually is achieved.  This Boltzmann equilibrium represents the charge distribution
12      of an aerosol in charge equilibrium with bipolar ions.  The minimum amount of charge is very
13      small, with a statistical probability that some particles within the aerosol will have no charge, and
14      others will have one or more charges.
15           The electrical charge on  some particles may result in an enhanced deposition over what
16      would be expected from size alone. This results from image charges induced on the surface of
17      the airway by these particles or to space-charge effects, whereby repulsion of particles containing
18      like charges results in increased migration toward the airway wall.  The effect of charge on
19      deposition is inversely proportional to particle size and airflow rate. This type of deposition is
20      probably small compared to the effects of turbulence and other deposition mechanisms, and
21      generally has been considered  to be a minor contributor to overall particle deposition. However,
22      a recent study (Cohen et al., 1998) employing hollow airway casts of the human tracheobronchial
23      tree assessed deposition of ultrafine (0.02 jum) and fine (0.125 /um) particles; the deposition of
24      singly charged particles was found to be 5 to 6 times that of particles having no charge and 2 to
25      3 times that of particles at Boltzmann equilibrium. This suggests that electrostatic precipitation
26      may, in fact, be a significant deposition mechanism for ultrafine, and some fine, particles within
27      the TB region.
28
29      7.2.2  Deposition Patterns in  the Human Respiratory Tract
30           Knowledge of sites where particles of different sizes deposit in the respiratory tract and the
31      amount of deposition is necessary for understanding and interpreting the health effects associated
        March 2001                                7-6         DRAFT-DO NOT QUOTE OR CITE

-------
  1      with exposure to particles. Particles deposited in the various respiratory tract regions are
  2      subjected to large differences in clearance mechanisms and pathways and, consequently,
  3      retention times. This section summarizes concepts of particle deposition in humans and
  4      laboratory animals as reported in U.S. Environmental Protection Agency (1996), and provides
  5      additional information based on studies published since the release of that earlier document.
  6           The ambient air often contains particles that are too massive to be inhaled. The descriptor
  7      "inhalability" is used to denote the overall spectrum of particle sizes that potentially are capable
  8      of entering the respiratory tract. Inhalability is defined as the ratio of the number concentration
  9      of particles of a certain aerodynamic diameter that are inspired through the nose or mouth to the
 10      number concentration of the same diameter particle present in an inspired volume of ambient air
 11      (International Commission on Radiological Protection, 1994).  In general, for humans, unit
 12      density particles >100-yam diameter have a low probability of entering the mouth or nose in still
 13      air.  However, there is no sharp cutoff to zero probability. Furthermore, there is no lower limit to
 14      inhalability as long as the particle exceeds a critical size where the aggregation of atomic or
 15      molecular units is stable enough to endow it with "particulate" properties, in contrast to those of
 16      free ions or gas molecules.
 17
 18      7.2.2.1 Total Respiratory Tract Deposition
 19           Total human respiratory tract deposition, as a function of particle size, is depicted in
20      Figure 7-1. These data were obtained by various investigators using different sizes of spherical
21      test particles in healthy male adults under different ventilation conditions; the large standard
22      deviations reflect interindividual and breathing pattern-related variability of deposition
23      efficiencies. Deposition with nose breathing is generally higher than that with mouth breathing
24      because of the superior filtration capabilities of the nasal passages. For particles with
25      aerodynamic diameters greater than 1 //m, deposition is governed by impaction  and
26      sedimentation, and it increases  with increasing AED. When AED is >10 //m, almost all inhaled
27      particles are deposited.  As the  particle size decreases from ~0.5 /j.m, diffusional deposition
28      becomes dominant and total deposition depends more on the actual physical diameter of the
29      particle, with decreasing particle diameter leading to an increase in total deposition. Total
30      deposition shows a minimum for particle diameters in the range of 0.3 to 0.5 /^m where, as noted
31      above, neither sedimentation, impaction or diffusion deposition are very effective.

        March 2001                                 7-7        DRAFT-DO NOT QUOTE OR CITE

-------
            c
            o
            o
                100
                 80
                 60
                 40
                 20
                    .2   o
      I
O Human (Oral)
• Human (Nasal)
                                             1
                        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
       Source: Modified from Schlesinger (1988).


1          Besides particle size, breathing pattern is the most important factor affecting lung
2      deposition. Recently, Kim (2000) reported total lung deposition values in healthy adults for a
3      wide range of breathing patterns; tidal volume 375 to 1500 mL, flow rate 150 to 1000 mL/s, and
4      respiratory time 2 to 12s. Total lung deposition increased with increasing tidal volume at a
5      given flow rate and increased with increasing flow rate at a given respiratory time. Various
6      deposition values were correlated with a single composite parameter consisting of particle size,
7      flow rate, and tidal volume.
      March 2001
                  DRAFT-DO NOT QUOTE OR CITE

-------
  1           One of the specific size modes of the ambient aerosol that is being evaluated in terms of
  2      potential toxicity is the ultrafme mode (i.e., particles having diameters <0.1 yum [CMD]).  There
  3      is little information on total respiratory tract deposition of such particles. Frampton et al. (2000)
  4      exposed healthy adult females to 26.7-nm diameter carbon particles (at 10 /^g/m3) for 2 h. The
  5      inspired and expired particle number concentration and size distributions were evaluated.  Total
  6      respiratory tract deposition fraction was determined for six particle size fractions, ranging from
  7      7.5 to 133.4 nm. They found an overall total lung deposition fraction of 0.66 (by particle
  8      number) or 0.58 (by particle mass), indicating that exhaled mean particle diameter was slightly
  9'     larger than inhaled diameter. The deposition fraction decreased with increasing particle size
 10      within the ultrafme range,  from 0.76 at the smallest size to 0. 47 at the largest.  Jaques and Kim
 11      (2000) found the greatest deposition fraction for smaller particles and for breathing patterns with
 12      longer residence times (i.e., low flow and higher tidal volume) consistent with deposition by
 13      diffusion.
 14           A property of some ambient particulate species that affects deposition is hygroscopicity, the
 15      propensity of a material for taking up and retaining moisture under certain conditions of humidity
 16      and temperature. Such particles can increase in size in the humid air within the respiratory tract
 17      and, when inhaled, will deposit according to their hydrated size rather than their initial size.  The
 18      implications of hygroscopic growth on deposition has been reviewed extensively by Morrow
 19      (1986) and Hiller (1991), whereas the complications of studying lung deposition of hygroscopic
 20      aerosols have been reviewed recently by Kim (2000). In general, compared to nonhygroscopic
 21      particles of the same initial size, the deposition of hygroscopic aerosols in different regions of the
 22      lung may be higher or lower, depending on the initial size. Thus, for particles with initial sizes
 23      larger than =0.5 yum, the influence of hygroscopicity is to increase total deposition, whereas for
 24      smaller ones total deposition is decreased.
 25
 26      7.2.2.2  Deposition in the Extrathoracic Region
27           The fraction of inhaled particles depositing in the ET region is quite variable, depending on
28      particle  size, flow rate, breathing frequency and whether breathing is through the nose or the
29      mouth.  Mouth breathing bypasses much of the filtration capabilities of the nasal airways, leading
30      to increased deposition in the lungs (TB and A regions).  The ET region is clearly the site of first
31      contact with particles in the inhaled air, and essentially acts as a "prefilter" for the lungs.

        March 2001                                 7-9        DRAFT-DO NOT QUOTE OR CITE

-------
 1           Since release of the 1996 PM AQCD, a number of studies have explored ET deposition
 2      with in vivo studies, as well as in both physical and mathematical model systems. In one study,
 3      the relative distribution of particle deposition between the oral and nasal passages was assessed
 4      during "inhalation" by use of a physical model (silicone rubber) of the human upper respiratory
 5      system, extending from the nostrils and mouth through the main bronchi (Lennon et al., 1998).
 6      Monodisperse particles ranging in size from 0.3 to 2.5 /um were used at various flow rates
 7      ranging from 15 to 50 L/min. Total deposition was assessed, as was regional deposition in the
 8      oral passages, lower oropharynx-trachea, nasal passages, and nasopharynx-trachea. Deposition
 9      within the nasal passages was found to agree with available data obtained from a human
10      inhalation study (Heyder and Rudolf, 1977), being proportional to particle size, density, and
11      inspiratory flow rate. It also was found that for oral inhalation, the relative distribution between
12      the oral cavity and the oropharynx-trachea was similar, whereas for nasal inhalation, the nasal
13      passages contained most of the particles deposited in the model, with only about 10% depositing
14      in the nasopharynx-trachea section. Furthermore, the deposition efficiency of the
15      nasopharynx-trachea region was greater than that of the oropharynx-trachea region.
16      For simulated oronasal breathing, deposition in the ET  region depended primarily on particle size
17      rather than flow rate. For all flows and for all breathing modes, total deposition in the ET region
18      increased as particle diameter increased. Such information on deposition patterns in the ET
19      region is useful in refining empirical deposition models.
20           Deposition within the nasal passages was further evaluated by Kesavanathan and Swift
21      (1998), who examined the deposition of 1- to 10-yum particles in the nasal passages of normal
22      adults under an inhalation regime in which the particles were drawn through the nose and out
23      through the mouth at flow rates ranging from 15 to 35 L/min.  At any particle size, deposition
24      increased with increasing flow rate; whereas, at any flow rate, deposition increased with
25      increasing particle size.  In  addition, as was shown experimentally by Lennon et al. (1998) under
26      oronasal breathing conditions, deposition of 0.3- to 2.5-yum particles within the nasal passages
27      was significantly greater than within the oral passages,  and nasal inhalation resulted in greater
28      total deposition in the model than did oral inhalation. These results are consistent with other
29      studies discussed in the 1996 PM AQCD and with the known dominance of impaction deposition
30      within the ET region.

        March 2001                                7-10        DRAFT-DO NOT QUOTE OR CITE

-------
 1           For ultrafme particles (d < 0.1 /mi), deposition in the ET region is controlled by diffusion,
 2      which depends only on the particle's geometric diameter.  Prior to 1996, ET deposition for this
 3      particle size range had not been studied extensively in humans, and this remains the case.  In the
 4      earlier document, the only data available for ET deposition of ultrafme particles were from cast
 5      studies.  More recently, deposition in the ET region was examined using mathematical modeling.
 6      Three dimensional numerical simulations of flow and particle diffusion in the human upper
 7      respiratory tract, which included the nasal region, oral region, larynx, and first two generations of
 8      bronchi, were performed by Yu et al. (1998).  Deposition of particles ranging from 0.001 to
 9      0.1 /mi in these different regions was calculated under inspiratory and expiratory flow conditions.
10      Deposition efficiencies in the total model were lower on expiration than inspiration, although the
11      former were quite high.  Nasal deposition of ultrafme particles can also be quite high. For
12      example, nasal deposition accounted for up to 54% of total deposition in the model system for
13      0.001-/on particles.  The total deposition efficiency in the model was 75% (of the amount
14      entering), for this size particle.  With oral breathing, deposition efficiency was estimated at 48%
15      (of amount entering) (Yu et al., 1998).
16           Swift and Strong (1996) examined the deposition of ultrafine particles, ranging in size from
17      0.053 to 0.062 /mi, in the nasal passages of normal adults during constant inspiratory flows of
18      6 to 22 L/min.  The results are consistent with results noted in studies above, namely that the
19      nasal passages are highly efficient collectors for ultrafine particles.  In this case, fractional
20      deposition ranged from 94 to 99% (of amount inhaled).  There was found to be only a weak
21      dependence of deposition on flow rate, which contrasts with results noted above (i.e., Lennon et
22      al., 1998) for particles >0.3 //m, but is consistent with diffusion as the deposition mechanism.
23           Cheng et al. (1997) examined oral airway deposition in  a replicate cast of the human nasal
24      cavity, oral cavity, and laryngeal-tracheal sections.  Particle sizes ranged from 0.005 to 0.150 /mi,
25      and constant inspiratory and expiratory flow rates of 7.5 to 30 L/min were used. They noted that
26      the deposition fractions within the oral cavity were essentially the same as that in the
27      laryngeal-tracheal sections for all particle sizes and flowrates. They ascribed this to the balance
28      between flow turbulence and residence time in these two regions. Svartengren et al. (1995)
29      examined the effect of changes in external resistance on oropharyngeal deposition of 3.6-/mi
30      particles in asthmatics. Under control mouthpiece breathing conditions (flow rate 0.5 L/s), the
31      median deposition as a percentage of inhaled particles in the mouth and throat was 20%

        March 2001                                7-11       DRAFT-DO NOT QUOTE OR CITE

-------
  1      (mean = 33%; range 12 to 84%). Although the mean deposition fell to 22% with added
  2      resistance, the median value remained at 20% (range 13 to 47%).  Fiberoptic examination of the
  3      larynx revealed that there was a trend for increased mouth and throat deposition associated with
  4      laryngeal narrowing.  Katz et al. (1999) indicate, on the basis of mathematical model
  5      calculations, that turbulence plays a key role in enhancing particle deposition in the larynx and
  6      trachea.
  7           The results of all of the above studies support the previously known ability of the ET
  8      region, and especially the nasal passages, to act as an efficient filter for inhaled particles.  Even
  9      ultrafine particles have significant deposition within the ET region, and this region, therefore,
 10      serves as an important filter for such particles as well as for larger ones, potentially reducing the
 11      amount of particles within a wide range that are available for deposition in the TB and A regions.
 12
 13      7.2.2.3 Deposition in the Tracheobronchial and Alveolar Regions
 14           Particles that do not deposit in the ET region enter the lung, but their regional deposition in
 15      the lung cannot be precisely measured. Much of the  available regional deposition data have been
 16      obtained from experiments with radioactive labeled poorly soluble particles. These have been
 17      described previously (U.S. Environmental Protection Agency, 1996). Although there are no new
 18      regional data obtained by means of the radioactive aerosol method since the publication of that
 19      document, a novel serial bolus delivery method has been introduced. Using this bolus technique,
20      regional deposition has been measured for fine and coarse aerosols (Kim et al.,  1996) and for
21      ultrafine aerosols (Kim and Jacques, 2000). The serial bolus method uses nonradioactive
22      aerosols and can measure regional deposition in virtually an unlimited number of lung
23      compartments. The Kim and Jaques studies cited above measured regional deposition in
24      10 serial compartments of the lung, and obtained tracheobronchial and alveolar deposition for
25      particles ranging  from 0.04 to 5.0 //m in diameter. TB and alveolar deposition also have been
26      measured in men and women using this method (Kim and Hu, 1998).
27
28      7.2.2.4 Local Distribution of Deposition
29           Airway structure and its associated air flow patterns are exceedingly complex and
30      ventilation distribution of air in different parts of the  lung is uneven. Thus, it is expected that
31      particle deposition patterns within the ET, TB, and A regions would be highly nonuniform, with

        March 2001                                7-12       DRAFT-DO NOT QUOTE OR CITE

-------
  1      some sites exhibiting deposition that is much greater than average levels within these regions.
  2      This was discussed in detail previously in the 1996 PM AQCD. Basically, using deposition data
  3      from living subjects as well as from mathematical and physical models, enhanced deposition has
  4      been shown to occur in the nasal passages and trachea and at branching points in the TB and
  5      A regions.  Recently, Churg and Vedal (1996) examined retention of particles on carinal ridges
  6      and tubular sections of airways from lungs obtained at necropsy.  Results indicated significant
  7      enhancement of particle retention on carinal ridges through the segmental bronchi; the ratios
  8      were similar in all airway generations examined.
  9           Deposition "hot spots" at airway bifurcations have undergone additional analyses using
 10      mathematical modeling techniques. Using calculated deposition sites, a number of studies
 11      showed a strong correlation between secondary flow patterns and deposition sites and density for
 12      large (10 ^m) particles, as well as for ultrafine particles (0.01 /urn) (Heistracher and  Hofmann,
 13      1997; Hofmann et al., 1996). This supports experimental work, noted in U.S. Environmental
 14      Protection Agency (1996), indicating that, like larger particles, ultrafine particles also show
 15      enhanced deposition at airway branch points, even in the upper tracheobronchial tree.
 16           The pattern of particle distribution on a more regional scale was evaluated by Kim et al.
 17      (1996). Deposition patterns were measured in situ in healthy nonsmoking young adult males,
 18      using an aerosol bolus technique that delivered 1-, 3-, or 5-//m particles into specific volumetric
 19      depths within the lungs. The distribution of particle deposition was uneven, and it was noted that
 20      sites of peak deposition shifted from distal to proximal regions of the lungs with increasing
 21      particle size. Furthermore, the surface dose was found to be greater in the conducting airways
 22      than in the alveolar region for all of the particle sizes evaluated. Within the conducting airways,
 23      the largest airway regions (i.e., 50 to 100 mL volume) received the greatest surface doses.
 24           Kim and Fisher (1999) studied local deposition efficiencies and deposition patterns of
 25      aerosol particles (2.9 to 6.7 jwm) in sequential double bifurcation tube models  with two different
26      branching geometries:  one with in-plane (A) and  another with out of plane (B) bifurcation. The
27      deposition efficiencies (DE) in each bifurcation increased with increasing Stokes number (Stk).
28      With symmetric flow conditions, DE was somewhat smaller in the second than the first
29      bifurcation in both models.  DE was greater in the second bifurcation in model B than in model
30      A. With asymmetric flows, DE was greater in the low-flow side compared to  the high-flow side
31      and this was consistent in both models. Deposition pattern analysis showed highly localized

        March 2001                               7-13        DRAFT-DO NOT QUOTE OR CITE

-------
 1      deposition on and in the immediate vicinity of each bifurcation ridge, regardless of branching
 2      pattern and flow pattern.
 3           Comer et al. (2000) used the same three-dimensional computer simulation technique to
 4      measure local deposition patterns in sequentially bifurcating, airway models.  The simulation was
 5      for 3-, 5-, and 7-^m particles and assumed steady, laminar, constant property air flow with
 6      symmetry about the first bifurcation.  The overall trend of the particle deposition efficiency (i.e.,
 7      an exponential  increase with Stokes number) was similar for all bifurcations.  Local deposition
 8      patterns consistently showed that the majority of the deposition occurred in the carinal region.
 9           Kim and Jaques (2000) used the respiratory bolus technique to measure the respiratory dose
10      of fine particles (0.04, 0.06, 0.08, and 0.1 /^m) in young adults. Under normal breathing
11      conditions (tidal volume 500 mL, respiratory flow rate 250 mL/s), bolus aerosols were delivered
12      sequentially to  a lung depth ranging from 50 to 500 mL in 50-mL increments. The results
13      indicate that regional deposition varies widely along the depth of the lung regardless of the
14      particle sizes used.  Peak deposition occurred in the lung regions  situated between 150 and
15      200 mL from the mouth and sites of peak deposition shifted proximally with a decrease in
16      particle size. Deposition dose per unit surface area was greatest in the proximal lung regions and
17      decreased rapidly with increased lung depth.  Peak surface dose was 5 to 7 times greater than the
18      average lung dose.  These results indicate that local enhancement of dose occurs in healthy lungs,
19      and dose enhancement could be an important factor in eliciting pathophysiological effects.
20
21      7.2.2.5  Deposition of Specific Size Modes of Ambient Aerosol
22           The studies described in previous sections generally evaluated deposition using individual
23      particle sizes within certain ranges, without consideration of specific relevant ambient size
24      ranges. Some recent studies, however, have considered the deposition profiles of particle modes
25      that exist in ambient air, so as to  provide information on dosimetry of these "real world" particle
26      size fractions. One such study (Venkataraman and Kao, 1999) examined the contribution of two
27      specific size modes of the PM,0 ambient aerosol, namely the fine  mode (defined as particles with
28      diameters up to 2.5  ,um) and the coarse mode (defined as particles with diameters 2.5 to 10 /^m),
29      to total lung and regional lung doses (i.e., a daily dose expressed as jUg/day, and a surface  dose
30      expressed a yug/cm2/day) resulting from a 24-h exposure to a particle concentration of 150 /ug/m3.
31      The study also evaluated deposition in terms of two metrics, namely mass dose and number dose.

        March 2001                               7-14        DRAFT-DO NOT QUOTE OR CITE

-------
  1      Deposition was calculated using a mathematical model for a healthy human lung under both
  2      moderate exertion and vigorous exertion. Regional deposition values were obtained for the
  3      nasopharyngeal region (NP), the tracheobronchial tree (TB), and the pulmonary airways (A).
  4           The daily mass dose from exposure to PM10 for three breathing cycles resulted in 36% of
  5      the inhaled coarse particle mass deposited in the lung and 30% in the NP, 4% in TB, and 2% in
  6      A. About 9% of the fine particle mass was deposited in the lungs, 1.5% in NP and TB and 6% in
  7      A. The daily mass dose peaked in the A airways (generation 20) for all breathing patterns,
  8      whereas that for the coarse fractions was comparable in the TB and A regions. The mass per unit
  9      surface area of various airways from the fine and coarse fractions was larger in the trachea and
10      first few generations of bronchi (gen 3 to 5). It was suggested that these large surface doses may
11      be related to aggravation of upper respiratory tract illness in geographical areas where coarse
12      particles were present.
13           The daily number dose from exposure to PM,0 resulted in 18% of the inhaled coarse
14      particles being deposited in the lungs, 13% in the NP, 2% in the TB, and 3% in A. About 11% of
15      inhaled fine particle number was  deposited in the lungs, 0.06% in NP, 2% in TB, and 9% in A.
16      Daily number dose was different  for fine and coarse fractions in all lung airways, with the dose
17      from the fine fraction higher by about 100 times in the NP and about 105 times in internal lung
18      airways. The surface number dose (particles/cm2/day) was 103 to 105 times higher for fine than
19      for coarse particles in all lung airways, indicating the larger number of fine particles depositing.
20      Particle number  doses did not follow trends in mass doses and are much higher for fine than
21      coarse particles and are higher for different breathing patterns.  It also was concluded that the fine
22      fraction contributes 10,000 times greater particle number per alveolar macrophage than the
23      coarse fraction particles. These results must be viewed with caution because they were obtained
24      using a pure mathematical model that must be validated.
25           Another evaluation of deposition that included consideration of size mode of the ambient
26      aerosol was that  of Broday and Georgopoulos (2000). In this case, a mathematical model was
27      used to account for particle hygroscopic growth, transport, and deposition in tracking the changes
28      in the size distribution of inhaled  aerosols. It was concluded that different rates of particle
29      growth in the inspired air resulted in a change in the  size distribution of the aerosol, such that
30      increased mass and number fractions of inspired fine particles are found in the size range
31      between 0.1 to 1  yum and, therefore, deposit to a lesser extent due to a decrease in diffusion

        March 2001                              7-15         DRAFT-DO NOT QUOTE OR CITE

-------
 1      deposition. On the other hand, particles that were originally in the 0.1 - to 1 -//m size range when
 2      inhaled will undergo enhanced deposition because of their increase in size resulting from
 3      hygroscopic growth. Thus, the speciation of the inhaled polydisperse aerosol and its initial size
 4      distribution affect the evolution of size distribution once inhaled and, thus, its deposition profile
 5      in the respiratory tract. Hygroscopicity of respirable particles must be considered for accurate
 6      predictions of deposition.  Because different fractions likely have different chemical
 7      composition, such changes in deposition patterns will affect dosimetry and biological responses.
 8
 9      7.2.3 Biological Factors Modulating Deposition
10           Experimental deposition data in humans are commonly derived using healthy adult
11      Caucasian males. Various factors can act to alter deposition patterns from those obtained in this
12      group. Evaluation of these factors is important to help understand potentially susceptible
13      subpopulations, because differences in biological response following pollutant exposure may be
14      caused by dosimetry differences as well as by differences in innate sensitivity. The effects of
15      different biological factors on deposition were discussed in U.S. Environmental Protection
16      Agency (1996) and are summarized below, with additional information obtained from more
17      recent studies.
18
19      7.2.3.1 Gender
20           Males and females differ in body size and ventilatory parameters; so, it is expected that
21      there would be gender differences in deposition.  Using particles in the 2.5- to 7.5-,um size range
22      Pritchard et al. (1986) indicated that, for comparable particle sizes and inspiratory flow rates,
23      females had higher ET and TB deposition and smaller A deposition than did males.  The ratio of
24      A deposition to total thoracic deposition in females also was found to be smaller. These
25      differences were attributed to gender differences in airway size.
26           In a recent study (Bennett et al., 1996), the total respiratory tract deposition of 2-/um
27      particles was examined in adult males and females aged 18 to 80 years who breathed with a
28      normal resting pattern. Deposition was assessed in terms of a deposition fraction, which was the
29      difference between the amount of particles inhaled and exhaled during oral breathing. Although
30      there was a tendency for a greater deposition fraction in females compared to males, because

        March 2001                                7-16        DRAFT-DO NOT QUOTE OR CITE

-------
  1     males had greater minute ventilation, the deposition rate (i.e., deposition per unit time) was
  2     greater in males than in females.
  3          Kim and Hu (1998) assessed regional deposition patterns in healthy adult males and
  4     females using aerosols with median aerodynamic sizes of 1, 3, and 5 /j.m and a bolus delivery
  5     technique, which involved controlled breathing. The total deposition in the lungs was similar for
  6     both genders with the smaller particle, but was greater in women for the 3- and 5-/um particles,
  7     regardless of the inhalation flow rate used; this difference ranged from 9 to 31 %, with higher
  8     values associated with higher flow rates. The pattern of deposition was similar for both genders,
  9     although females showed enhanced deposition peaks for all three particle sizes. The volumetric
 10     depth location of these peaks was found to shift from peripheral (increased volumetric depth) to
 11     proximal (shallow volumetric depth) regions of the lung with increasing particle size, but the
 12     shift was greater in females than in males. Thus, deposition appeared to be more localized in the
 13     lungs of females compared to those of males. These differences were attributed to a smaller size
 14     of the upper airways in females than in males (particularly of the laryngeal structure).  Local
 15     deposition of l-/u.m particles was somewhat flow dependent, but for larger (5-^m) particles was
 16     largely independent of flow (flows did not include those that  would be typical of exercise).
 17          In a related study, Kim et al. (2000) evaluated differences in deposition between males and
 18     females related to exercise levels of ventilation and breathing patterns.  Using particles at the
 19     same size noted above and a number of breathing conditions, total lung deposition was
 20     comparable between men and women for l-/um particles but was greater in women than men for
 21      3- and 5-^m particles with all breathing patterns. The gender difference was about 15% at rest,
 22     and variable during exercise, depending on particle size. However, total lung deposition rate
 23      (deposition per unit time) was found to be 3 to 4 times greater during moderate exercise than
 24      during rest for all particle sizes. Thus, it was concluded that exercise may increase the health risk
 25      from particles because of increased deposition, and that women may be more susceptible to this
 26      exercise-induced change.
 27           Jaques and Kim (2000) and Kim and Jaques (2000) expanded the evaluation of deposition
28      in males and females to particles <1 /^m.  They measured total lung deposition in healthy adults
29      using sizes in the ultrafine mode (0.04 to 0.1 /wm), in addition to those having diameters of 1 and
30      5  ,um. Total lung deposition was greater in females than in males for 0.04- and 0.06-^m
31      particles. The difference was negligible for 0.08- and 0.1 -/um particles. Therefore, the gender

        March 2001                               7-17         DRAFT-DO NOT QUOTE OR CITE

-------
 1     effect was particle-size dependent, showing a greater deposition in females for very small
 2     ultrafine and large coarse particles, but not for fine particles ranging from 0.08 to 1 //m. A local
 3     deposition fraction was determined in each volumetric compartment of the lung to which
 4     particles are injected based on the inhalation procedure (Kim and Jaques, 2000).  The deposition
 5     fraction was found to increase with increasing lung depth from the mouth, reach a peak value and
 6     then decrease with further increase in lung volumetric depth. The height of the peak and its
 7     depth did vary with particle size and breathing pattern. Peak deposition  for the 5-/um particles
 8     was more proximal than that for the  \-jj.m particles, whereas that for the ultrafine particles
 9     occurred between these two peaks. For the ultrafine particles, the peak deposition became more
10     proximal as particle size decreased.  Although this pattern of deposition  distribution was similar
11     for both men and women, the region of peak deposition was shifted closer to the mouth and peak
12     height was slightly greater for women than for men for all exposure conditions.
13
14     7.2.3.2 Age
15           Airway structure and respiratory conditions vary with age, and these variations may alter
16     the deposition pattern of inhaled particles. The limited experimental studies reported in the
17     earlier PM AQCD (U. S. Environmental Protection Agency, 1996) indicated results ranging from
18     no clear dependence of total deposition on age to slightly higher deposition in children than
19     adults. Potential regional deposition differences between children and adults were assessed to a
20     greater extent using mathematical models. These indicated that if the entire respiratory tract and
21     a complete breathing cycle at normal rate are considered, that ET deposition in children generally
22     would be higher than that in adults, but that TB and A regional deposition in children may be
23     either higher or lower than the adult, depending on particle size (Xu and Yu, 1986). Enhanced
24     deposition in the TB region would occur for particles <5  ,um in children  (Xu and Yu, 1986;
25     Hofmann et al, 1989a).
26           An age dependent theoretical model to predict regional particle deposition in childrens'
27     lungs, and that incorporates breathing parameters and morphology of the growing lung, was
28     developed by Musante and Martonen (1999). The model was used to compare deposition, at rest,
29     of monodisperse aerosols, ranging from 0.25 to 5 /urn, in the lungs of children (aged 7, 22, 48,
30     and 98 mo) to that in adults (aged 30 years). Compared to adults, A deposition was highest in the
31     48- and 98-mo subjects for all particle sizes, TB deposition was found to be a monotonically

       March 2001                               7-18        DRAFT-DO NOT QUOTE OR CITE

-------
  1      decreasing function of age for all sizes; and total lung deposition (i.e., TB+A) was generally
  2      higher in children than adults, with children of all ages showing similar total deposition fractions.
  3           This model was used by Musante and Martonen (2000a) to evaluate the deposition of a
  4      polydisperse aerosol that has been extensively used in toxicological studies, namely residual oil
  5      fly ash (ROFA) having an MMAD of 1.95 ^m.  Deposition was evaluated under resting
  6      breathing conditions. The mass based deposition fraction of the particles was found to decrease
  7      with age from 7 mo to adulthood, but the mass deposition per unit surface area in the lungs of
  8      children could be  significantly greater than that in the adult.
  9           Cheng et al. (1995) examined deposition of ultrafme particles in replica casts of the nasal
10      airways  of children aged 1.5 to 4 years. Particle sizes ranged from 0.0046 to 0.2 /^m, and both
11      inspiratory and expiratory flowrates were used (3 to  16 L/min). Deposition efficiency was found
12      to decrease with increasing age for a given particle size and flowrate.
13           Oldham et al. (1997) examined the deposition of monodisperse particles, having diameters
14      of 1, 5, 10, and 15 ^m, in hollow airway models that were designed to represent the trachea and
15      the first  few bronchial airway generations of an adult, a 7-year-old child, and a 4-year-old child.
16      They noted that in most cases, the total deposition efficiency was greater in the child-size models
17      than in the adult model.
18           Bennett et al. (1997a) analyzed the regional deposition of 4.5 /urn, poorly soluble particles
19      in children and in adults with mild cystic fibrosis (CF), but who likely had normal upper airway
20      anatomy, such that intra- and extrathoracic deposition would be similar to that in healthy adults.
21      The mean age of the children was 13.8 years and adults were 29.1 years.  ET deposition,  as a
22      percentage of total respiratory tract deposition, was higher by about 50% in children compared to
23      CF and healthy adults (30.7%, 20.1%, and 16.0%, respectively). There was an age dependence
24      of ET deposition in the children, in that the percentage ET deposition tended to be higher at a
25      younger age; the younger group (<14 years) had almost twice the percentage ET deposition of the
26      older group (>14 years). Additional analyses showed an inverse correlation of extrathoracic
27      deposition with body height.  There was no significant difference in lung or total respiratory tract
28      deposition between the children and adults. Because ET deposition was age dependent and total
29      deposition was not, this suggests that the ET region does a more effective job in children of
30      filtering out the particles that would otherwise reach  the TB region.  However, because the lungs


        March 2001                                7-19        DRAFT-DO NOT QUOTE OR CITE

-------
 1      of children are smaller than those of adults, children may still have comparable deposition per
 2      unit surface area as would adults.
 3           Bennett and Zeman (1998) measured the deposition of monodisperse 2 //m (MMAD)
 4      particles in children aged 7 to 14 years and adolescents aged 14 to 18 years for comparison to
 5      that in adults (19 to 35 years). Each subject inhaled the particles by following their previously
 6      determined individual spontaneous resting breathing pattern. Deposition was assessed by
 7      measuring the amount of particles inhaled and exhaled. There was no age-related difference in
 8      deposition within the children group.  There was also no significant difference in deposition
 9      between the children and adolescents, between the children and adults, or between the
10      adolescents and adults.  However, the investigators noted that, because the children had smaller
11      lungs and higher minute volumes relative to lung size, they likely would receive greater doses of
12      particles per lung surface area compared to adults.  Furthermore, deposition in children did vary
13      with tidal volume, increasing with increasing volume to a greater extent than was seen in adults.
14      These additional studies still do not provide unequivocal evidence for significant differences in
15      deposition between adults and children, even  when considering differences in lung surface area.
16      However, it should be noted that differences in levels of activity between adults and children are
17      likely to play a fairly large role in age-related  differences in deposition patterns of ambient
18      particles. Children generally have higher activity levels during the day, and higher associated
19      minute ventilation per lung size, which can contribute to a greater size-specific dose of particles.
20      Activity levels in relationship to exposure are discussed more fully in Chapter 5.
21           Another subpopulation of potential concern related to susceptibility to inhaled particles is
22      the elderly. In the study of Bennett et al. (1996), in which the total respiratory tract deposition of
23      2-^m particles was examined in people aged 18 to 80 years, the deposition fraction in the lungs
24      of people with normal lung function was found to be independent of age, depending solely on
25      breathing pattern and airway resistance.
26
27      7.2.3.3 Respiratory Tract Disease
28           The presence of respiratory tract disease can affect airway structure and ventilatory
29      parameters, thus altering deposition compared to that in healthy individuals. The effect of airway
30      diseases on deposition has been studied extensively, as described in the  earlier PM AQCD (U.S.
31      Environmental Protection Agency, 1996).  Studies described therein had shown that people with

        March 2001                                7-20         DRAFT-DO NOT QUOTE OR CITE

-------
  1      chronic obstructive pulmonary disease (COPD) had very heterogeneous deposition patterns, with
  2      differences in regional deposition compared to normals. People with asthma and obstructive
  3      pulmonary disease tended to have greater TB deposition than did healthy people. Furthermore,
  4      there tended to be an inverse relationship between bronchconstriction and the extent of
  5      deposition in the A region, whereas total respiratory tract deposition generally increased with
  6      increasing level of airway obstruction. The described studies were performed during controlled
  7      breathing, where all subjects breathed with the same tidal volume and respiratory rate. However,
  8      although resting tidal volume is similar or elevated in people with COPD compared to normals,
  9      the former tend to breathe at a faster rate, resulting in higher than normal tidal peak flow and
 10      resting minute ventilation.  Thus, some of the reported differences in the deposition of particles
 11      could have been caused by increased fractional deposition with each breath. Although the extent
 12      to which lung deposition may change with respect to particle size, breathing pattern, and disease
 13      status in people with COPD is still unclear, some recent studies have attempted to provide
 14      additional insight into this issue.
 15           Bennett et al. (1997b) measured the fractional deposition of insoluble 2-^m particles in
 16      people with severe to moderate COPD (mix of emphysema and chronic bronchitis,  mean age
 17      62 years) and compared this to healthy older adults (mean age  67 years) under conditions where
 18      the subjects breathed using their individual resting breathing pattern, as well as a controlled
 19      breathing pattern. People with COPD tended to breathe with elevated tidal volume and at a
 20      faster rate than people with healthy lungs, resulting in about 50% higher resting minute
 21      ventilation. Total respiratory tract  deposition was assessed in terms of deposition fraction, a
 22      measure of the amount deposited based on measures of aerosol inhaled and amount exhaled, and
 23      deposition rate, the particles deposited per unit time.  Under typical breathing conditions, people
 24      with COPD had about 50% greater deposition fraction than did age-matched healthy adults.
 25      Because of the elevation in minute ventilation, people with COPD had  average deposition rates
26      about 2.5 times that of healthy adults.  Similar to previously reviewed studies (U.S.
27      Environmental Protection Agency, 1996), these investigators observed  an increase in deposition
28      with an increase in airway resistance, suggesting that, at rest, COPD resulted in increased
29      deposition of fine particles in proportion to the severity  of airway disease.  The investigators also
30      reported a decrease in deposition with increasing mean effective airspace diameter;  this
31      suggested that the enhanced deposition was associated more with the chronic bronchitic

        March 2001                                7-21         DRAFT-DO NOT QUOTE OR CITE

-------
  1      component of COPD than with the emphysematous component of the disease. Greater
  2      deposition was noted with natural breathing compared to the fixed pattern.
  3           Kim and Kang (1997) measured lung deposition of l-yum particles inhaled via the mouth by
  4      healthy adults (mean age 27 years) and by those with various degrees of airway obstruction,
  5      namely smokers (mean age 27 years), smokers with small airway disease (SAD; mean age
  6      37 years), asthmatics (mean age 48 years), and patients with COPD (mean age 61 years)
  7      breathing under the same controlled pattern.  Deposition fraction was obtained by measuring the
  8      number of particles inhaled and exhaled, breath by breath. There was a marked increase in
  9      deposition in people with COPD. Deposition was 16%, 49%, 59%, and 103% greater in
10      smokers, smokers with SAD, asthmatics and people  with COPD, respectively, than healthy
11      adults.  Deposition in COPD patients was significantly greater than that associated with either
12      SAD or asthma; there was no significant difference in deposition between people with SAD and
13      asthma. Deposition fraction was  found to be correlated with percent predicted forced expiratory
14      volume (FEV,) and forced expiratory flow (FEF25-75%). Airway resistance was not correlated
15      strongly with total lung deposition.  Kohlhaufl et al.  (1999) also showed increased deposition of
16      fine particles (0.9 //m) in women  with bronchial hyperresponsiveness.
17           Segal et al. (2000a) developed a mathematical model for airflow and particle motion in the
18      lungs that was used to evaluate how lung cancer affects deposition patterns in the lungs of
19      children.  It was noted that the presence of airway tumors could affect deposition, by increasing
20      probability of inertial deposition and diffusion.  The  former would occur on the upstream
21      surfaces of tumors, whereas the latter would occur on downstream surfaces. It was concluded
22      that particle deposition is affected by the presence of airway disease, but that effects may be
23      systematic and could be predicted and incorporated into dosimetry models.
24           Thus, the database related to particle deposition and lung disease suggests that total lung
25      deposition generally is increased with obstructed airways, regardless of deposition distribution
26      between the TB and A regions. Airflow distribution is very uneven in COPD because of the
27      irregular pattern of obstruction, and there can be closure of small airways. In this situation, a part
28      of the lung is inaccessible, and particles can penetrate deeper into other better ventilated regions.
29      Thus, deposition can be enhanced locally in regions of active ventilation, particularly in the
30      A region.  The relationships between lung deposition and airway obstruction or ventilation
31      distribution were previously studied in  vivo in animal models (Kim, 1989; Kim et al., 1989).

        March 2001                               7-22        DRAFT-DO NOT QUOTE OR CITE

-------
  1      7.2.3.4 Anatomical Variability
  2           As indicated above, variations in anatomical parameters between genders and between
  3      healthy people and those with obstructive lung disease can affect deposition patterns.  However,
  4      previous analyses generally have overlooked the effect on deposition of normal interindividual
  5      variability in airway structure in healthy individuals. This is an important consideration in
  6      dosimetry modeling, which often is based on a single idealized structure.  Studies available since
  7      1996 have attempted to assess the influence of such variation in respiratory tract structure on
  8      deposition patterns.
  9           The ET region is the first to contact inhaled particles and, therefore, deposition within this
 10      region would reduce the amount of particles available for deposition in the lungs. Variations in
 11       relative deposition within the ET region will, therefore, propagate through the rest of the
 12      respiratory tract, creating differences in calculated doses from individual to individual.
 13      A number of studies have examined the influence of variations in airway geometry on deposition
 14      in the ET region.
 15           Cheng et al. (1996) examined nasal airway deposition in healthy adults using particles
 16      ranging in size from 0.004 to 0.15 //m at two constant inspiratory flow rates, 167 and 33 mL/s.
 17      Deposition was evaluated in relation to measures of nasal geometry as determined by magnetic
 18      resonance imaging and acoustic rhinometry. They noted that interindividual variability in
 19      deposition was correlated with the wide variation of nasal dimensions, in that greater surface
 20      area, smaller cross-sectional area and increasing complexity of airway shape were all associated
 21      with enhanced deposition.
 22           Using a regression analysis of data on nasal airway deposition derived from Cheng et al.
 23      (1996), Guilmette et al. (1997) noted that the deposition efficiency within this region was highly
 24      correlated with both nasal airway surface area and volume; this indicated that airway size and
 25      shape factors were important in explaining intraindividual variability noted in experimental
 26      studies of human nasal airway aerosol deposition. Thus, much of the variability in measured
 27      deposition among people resulted from differences in the size and shape of airway regions.
 28           Kesavanathan and Swift (1998) also evaluated the influence of geometry in affecting
 29      deposition in the nasal passages of normal adults from two ethnic groups.  Mathematical
30      modeling of the results indicated that the shape of the nostril affected particle deposition in the
31      nasal passages, but that there still remained large intersubject variations in deposition when this

        March 2001                                7-23        DRAFT-DO NOT QUOTE OR CITE

-------
  1      was accounted for, and that likely was caused by geometric variability in the mid and posterior
  2      regions of the nasal passages.
  3           Bennett et al. (1998) studied the role of anatomic dead space (ADS) in particle deposition
  4      and retention in bronchial airways using an aerosol bolus technique. They found that the
  5      fractional deposition was dependant on the subject's ADS, and that a significant number of
  6      particles were retained beyond 24 h. This finding of prolonged retention of insoluble particles in
  7      the airways is consistent with the findings of Scheuch et al. (1995) and Stahlhofen et al. (1986a).
  8      Bennett et al. (1999) also found a lung volume-dependent asymmetric distribution of particles
  9      between the left and right lung; the  left:right ratio was increased at increased percentage of total
10      lung capacity (e.g., at 70% TLC,  L:R was 1.60).
11           From the analysis of detailed deposition patterns measured by a serial bolus delivery
12      method, Kim and Hu (1998) and  Kim and Jaques (2000) found a marked enhancement in
13      deposition in the very shallow region of the lungs in females. The enhanced local deposition for
14      both ultrafine and coarse particles was attributed to a smaller size of the upper airways,
15      particularly of the laryngeal structure.
16           Hofmann et al. (2000) examined the role of heterogeneity of airway structure in the rat
17      acinar region in affecting deposition patterns within this area of the lungs.  By the use of different
18      morphometric models, they showed that substantial variability in predicted particle deposition
19      would result.
20
21      7.2.4 Interspecies Patterns of Deposition
22           The primary purpose of this document is to assess the health effects of particles in humans.
23      As such, human dosimetry studies have been stressed.  Such studies avoid uncertainties
24      associated with extrapolation of dosimetry from laboratory animals to humans. Nevertheless,
25      animal models have been and are currently being used in evaluations of health effects from
26      particulate matter, because there are ethical limits to the types of studies that can be performed on
27      human subjects.  Because of this, there is considerable need to understand dosimetry in animals,
28      and to understand dosimetric differences between animals and humans, hi this regard, there has
29      been a number of new studies that were designed to assess particle dosimetry in commonly used
30      animals and to relate this to dosimetry in humans.

        March 2001                               7-24        DRAFT-DO NOT QUOTE OR CITE

-------
  1          The various species used in inhalation toxicology studies that serve as the basis for
  2     dose-response assessment may not receive identical doses in a comparable respiratory tract
  3     region (i.e., ET, TB, or A) when exposed to the same aerosol at the same inhaled concentration.
  4     Such interspecies differences are important, because any adverse toxic effect is often related to
  5     the quantitative pattern of deposition within the respiratory tract as well as to the exposure
  6     concentration; this pattern determines not only the initial respiratory tract tissue dose, but also the
  7     specific pathways by which deposited material is cleared and redistributed (Schlesinger, 1985).
  8     Differences in patterns of deposition between humans and animals were summarized previously
  9     in the earlier PM AQCD (U.S. Environmental Protection Agency, 1996; Schlesinger et al., 1997).
 10     Such differences in initial deposition must be considered when relating biological responses
 11     obtained in laboratory animal studies to effects in humans.
 12          One of the issues that must be considered in interspecies comparisons of hazards from
 13     inhaled particles is inhalability of the aerosol in the atmosphere of concern. Although this may
 14     not be an issue for humans per se as far as exposure to ambient particles are concerned, it can be
 15     an important issue when attempting to relate results of studies using animal species employed in
 16     inhalation toxicological studies (Miller et al.,  1995). For example, differences in inhalability
 17     between rat and human become very pronounced for particles >5 /um, and some differences are
 18     also evident for particles as small as 1 /urn.
 19          Several recent studies have addressed various aspects of interspecies differences in
 20     deposition using mathematical modeling approaches.  Hofmann et al. (1996) compared
 21      deposition between rat and human lungs using three-dimensional asymmetric bifurcation models
 22     and mathematical procedures for obtaining air flow and particle trajectories. Deposition in
 23      segmental bronchi and terminal bronchioles was evaluated under both inspiration and expiration,
 24      at particle sizes of 0.01, 1, and 10 //m (which covered the range of deposition mechanisms from
 25      diffusion to impaction). Total deposition efficiencies of all particles in the upper and lower
 26      airway bifurcations were comparable in magnitude for both rat and human. However, the
 27      investigators noted that penetration probabilities from preceding airways must be considered.
28      When considering the higher penetration probability in the human lung, the resulting bronchial
29      deposition fractions were generally higher in human than rat. For all particle sizes, deposition at
30      rat bronchial bifurcations was less enhanced on the carinas compared to that found in human
31      airways.

        March 2001                                7-25         DRAFT-DO NOT QUOTE OR CITE

-------
 1           Hoftnann et al. (1996) attempted to account for interspecies differences in branching
 2     patterns in deposition analyses. Numerical simulations of three-dimensional particle deposition
 3     patterns within selected (species-specific) bronchial bifurcations indicated that morphologic
 4     asymmetry was a major determinant of the heterogeneity of local deposition patterns.  They noted
 5     that many interspecies deposition calculations used morphometry that was described by
 6     deterministic lung models (i.e., the number of airways in each airway generation adopts a
 7     constant value, and all airways in a given generation have identical lengths and diameters). Such
 8     models cannot account for variability and branching asymmetry of airways in the lungs. Thus,
 9     their study employed computations that used stochastic morphometric models of human and rat
10     lungs (Koblinger and Hofmann, 1985, 1988; Hofmann et al., 1989b) and evaluated regional and
11     local particle deposition. Stochastic models of lung structure describe, in mathematical terms,
12     the inherent asymmetry and variability of the airway system, including diameter, length and
13     angle. They are based on statistical analyses of actual morphometric analyses of lungs. The
14     model also incorporated breathing patterns for humans and rats.  The dependence of deposition
15     on particle size was found to be similar in both rats and humans, with deposition minima in the
16     size range of 0.1 to 1 /um for both total deposition and deposition within the TB region.  This was
17     not found to occur in the A region, where a deposition maximum occurred at about 0.02 to
18     0.03 /urn in both species followed by a decline, and then another maximum between 3 and 5 /urn.
19     The deposition decrease in the A region at the smallest and largest sizes resulted from the
20     filtering efficiency of upstream airways. Although  deposition patterns were qualitatively similar
21     in rat and human, total respiratory tract and TB deposition in the human lung appeared to be
22     consistently higher than in the rat. Alveolar region deposition fraction in humans was lower than
23     in the rat over the size range of 0.001 to 10 /^m. Furthermore, both species showed a similar
24     pattern of dependence of deposition on flow rate.
25           The above model also assessed local deposition.  In both human and rat, deposition of
26     0.001- and 10-^m particles was highest in the upper bronchial airways, whereas  0.1- and l-/um
27     particles showed higher deposition in more peripheral airways, namely the bronchiolar airways
28     in rat and the respiratory bronchioles in humans.  Deposition was variable within any branching
29     generation because of differences in airway dimensions, and regional and total deposition also
30     exhibited intrasubject variations.  Airway geometric differences between rats and humans were
31     reflected in deposition. Because of the greater branching asymmetry in rats, prior to about

       March 2001                               7-26        DRAFT-DO NOT QUOTE OR CITE

-------
  1      generation 12, each generation showed deposition maxima at two particle sizes, reflecting
  2      deposition in major and minor daughters. These geometric differences became reduced with
  3      depth into the lung; beyond generation 12, these two maxima were no longer seen.  A later
  4      analysis (Hofmann and Bergmann, 1998), using a stochastic morphometric model of human and
  5      rat lungs to compare regional and local particle deposition in the human and rat lungs over a wide
  6      range of particle sizes (1 to 10 //m) and flow rates, noted that, although there were quantitative
  7      differences in the deposition patterns within the lungs of these two species, the dependence of
  8      deposition on particle size and flow rate was qualitatively similar.  This indicates that the
  9      dependence of deposition on physical factors is similar for all species.
 10           Another comparison of deposition in lungs of humans and rats was performed by Musante
 11      and Martonen (2000b). An interspecies mathematical dosimetry model was used to determine
 12      the deposition of residual oil fly ash (ROFA) in the lungs under sedentary and light activity
 13      breathing patterns. This latter was mimicked in the rat by increasing the CO2 level in the
 14      exposure system. The MMAD of the aerosol was 1.95 jum.  They noted that physiologically
 15      comparable respiratory intensity levels did not necessarily correspond to comparable dose
 16      distribution in the lungs. Because of this, the resting rat  may not be a good model for the resting
 17      human.  The ratio of aerosol mass deposited in the TB region to that in the A region for the
 18      human at rest was 0.961, indicating fairly uniform deposition throughout the lungs. On the other
 19      hand, in the resting rat, the ratio was 2.24, indicating greater deposition in the TB region than in
 20      the A region. However, by mimicking light activity in the rat, the ratio was reduced to 0.97,
 21      similar to the human.  This suggests that ventilatory characteristics in animal models may have to
 22      be adjusted to provide for comparable regional deposition to that in humans.
 23           The relative distribution of particles deposited in the bronchial and alveolar region airways
 24      may differ in the lungs of animals and humans, for the same total amount of deposited matter,
 25      because  of structural differences. The effect of such structural difference between rat and human
26      airways on particle deposition patterns was examined by Hofmann et al. (1999) in an attempt to
27      find the most appropriate morphometric parameter to characterize local particle deposition for
28      extrapolation modeling purposes.  Particle deposition patterns were evaluated as functions of
29      three morphometric parameters, namely (1) airway generation, (2) airway diameter, and
30      (3) cumulative path length.  It was noted that airway diameter was a more appropriate


        March 2001                               7-27        DRAFT-DO NOT QUOTE OR CITE

-------
 1      morphometric parameter for comparison of particle deposition patterns in human and rat lungs
 2      than was airway generation.
 3           The influence of exposure concentration on the pattern of particle retention in rats (exposed
 4      to diesel soot) and humans (exposed to coal dust) was examined by Nikula et al. (2000) using
 5      histological lung sections obtained from both species. The exposure concentrations for diesel
 6      soot were 0.35, 3.5, or 7.0 mg/m3, and exposure duration was 7 h/day, 5 days/week for 24 mo.
 7      The human lung sections were obtained from nonsmoking nonminers, nonsmoking coal miners
 8      exposed to levels <2 mg dust/m3 for 3 to 20 years, or nonsmoking miners exposed to <10 mg/m3
 9      for 33 to 50 years. In both species, the volume density of deposition increased with increasing
10      dose (which is related to exposure duration and concentration).  In rats, the diesel exhaust
11      particles were found to be primarily in the lumens of the alveolar duct and alveoli, whereas, in
12      humans, retained dust was found primarily in the interstitial tissue. Thus, different lung cells
13      contact retained particles in the two species and may result in different biological responses with
14      chronic dust exposure.
15           The manner in which particle dose is expressed, that is, the specific dose metric, may
16      impact on relative differences in deposition between humans and other animal species.
17      For example, although deposition when expressed on a mass per unit alveolar surface area basis
18      may not be different between rats and humans, dose metrics based on particle number per various
19      anatomical parameters (e.g., per alveolus or alveolar macrophage) can differ between rats and
20      humans, especially for particles  around 0.1 to 0.3 yum (Miller et al., 1995). Furthermore, in
21      humans with lung disease such as asthma or COPD, differences between rat and human can be
22      even more pronounced.
23           The probability of any biological effect in humans or animals depends on deposition and
24      retention of particles, as well as  the underlying dose-response relationship. Interspecies
25      dosimetric extrapolation must consider differences in deposition, clearance, and  dose response.
26      Thus, even similar deposition patterns may not result in similar effects in different species
27      because dose also is affected by clearance  mechanisms and species sensitivity. In addition, the
28      total number of particles deposited in the lung may not be the most relevant dose metric to
29      compare species. For example,  it may be the number of deposited particles per unit surface area
30      that determines response. More specifically, even if deposition is similar in rat and human, there
31      would be a higher deposition density in the rat because of the smaller surface area of rat lung.

        March 2001                               7-28        DRAFT-DO NOT QUOTE OR CITE

-------
  1     Thus, species-specific differences in deposition density should be considered when health effects
  2     observed in laboratory animals are being evaluated in terms of the human situation.
  3
  4
  5     7.3  PARTICLE CLEARANCE AND TRANSLOCATION
  6          This section discusses the clearance and translocation of particles that have deposited in the
  7     respiratory tract. A basic overview of biological mechanisms and pathways of clearance in the
  8     various region of the respiratory tract is presented first. This is followed by an update on
  9     regional kinetics of particle clearance. Interspecies patterns of clearance are then addressed,
 10     followed by new information on biological factors that may modulate clearance.
 11
 12     7.3.1  Mechanisms and Pathways of Clearance
 13          Particles that deposit on airway surfaces may be cleared from the respiratory tract
 14     completely, or may be translocated to other sites within this system, by various regionally distinct
 15     processes.  These clearance mechanisms, which are outlined in Table 7-1, can be categorized as
 16     either absorptive (i.e., dissolution) or nonabsorptive (i.e., transport of intact particles) and may
 17     occur simultaneously or with temporal variations.  It  should be mentioned that particle solubility
 18     in terms of clearance refers to solubility within the respiratory tract fluids and cells.  Thus, a
 19     poorly soluble particle is considered to be one whose rate of clearance by dissolution is
 20     insignificant compared to its rate of clearance as an intact particle.  For the most part, all
 21      deposited particles are subject to clearance by the same mechanisms, with their ultimate fate a
 22     function of deposition site, physicochemical properties (including solubility and any toxicity),
 23      and sometimes deposited mass or number concentration. Clearance routes from the various
 24      regions of the respiratory tract have been discussed previously in detail (U.S. Environmental
 25      Protection Agency, 1996; Schlesinger et al., 1997). They are schematically shown in Figure 7-2
26      (for extrathoracic and tracheobronchial regions) and in Figure 7-3 (for poorly soluble particle
27      clearance from the alveolar region) and are reviewed  only briefly below.
28
29
       March 2001                               7-29        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 and absorption into blood

                    Tracheobronchial region (TB)
                       Mucociliary transport
                       Endocytosis by macrophages/epithelial cells
                       Coughing
                       Dissolution and absorption into blood/lymph

                    Alveolar region (A)
                       Macrophages, epithelial cells
                       Interstitial
                       Dissolution and absorption into blood/lymph	

                    Source: Schlesinger (1995).
                                                    (  Nasal Passages
                             Blood
)
                                          Dissolution
                                                                              Posterior
                       Extrinsic Clearance
                                                                                    Mucociliary
                                                                                    Transport
                                                                               Pharynx
Tracheobronchial Tree
                   Figure 7-2.  Major clearance pathways for particles deposited in the extrathoracic region
                               and tracheobronchial tree.

                   Source:  Adapted from Schlesinger et al. (1997).
                   March 2001
    7-30
DRAFT-DO NOT QUOTE OR CITE

-------
L'CjyL'tS^f It^U f df ll\fl^
T
Phagocytosis by ^j
Alveolar Macrophages I
1 \
T p
Movement within _^ .
Alveolar Lumen
1
Bronchiolar / Bronchial ^ 	
Lurncn ^
*
Mucociliary Blanket
i
Gl Tract
Endocyt
^ lype i A
Epithelic
T
assage Through
Iveolar Epithelium ^"
i

Interstitium ^

i
i
mphatic Channels *^~
v
1 ^ f mi^h NI/^/Hoc*
osis by
Iveolar
il Pnllr- Rlnnrl ^
11 L/cllo DIUUU -^ 	
A
Passage through
Pulmonary Capillary
Endothelium
A A
i
] Phagocytosis by \
^ Interstitial
n \^ Macrophages J


       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      7.3.1.1 Extra thoracic Region
 2           The clearance of poorly soluble particles deposited in the posterior portions of the nasal
 3      passages occurs via mucociliary transport, with the general flow of mucus towards the
 4      nasopharynx.  Mucus flow in the most anterior portion of the nasal passages is forward, clearing
 5      deposited particles to the vestibular region where removal is by sneezing, wiping, or blowing.
 6           Soluble material deposited on the nasal epithelium is accessible to underlying cells via
 7      diffusion through the mucus.  Dissolved substances may be translocated subsequently into the
 8      bloodstream. The nasal passages have a rich vasculature, and uptake into the blood from this
 9      region may occur rapidly.
10
       March 2001
7-31
DRAFT-DO NOT QUOTE OR CITE

-------
 1           Clearance of poorly soluble particles deposited in the oral passages is by coughing and
 2      expectoration or by swallowing into the gastrointestinal tract.  Soluble particles are likely to be
 3      rapidly absorbed after deposition.
 4
 5      7.3.1.2 Tracheobronchial Region
 6           Poorly soluble particles deposited within the TB region are cleared by mucociliary
 7      transport towards the oropharynx, followed by swallowing.  Poorly soluble particles also may
 8      traverse the epithelium by endocytotic processes, entering the peribronchial region, be engulfed
 9      via phagocytosis by airway macrophages, which can then move cephalad on the mucociliary
10      blanket, or enter the airway lumen from the bronchial or bronchiolar mucosa.  Soluble particles
11      may be absorbed through the epithelium into the blood. There is, however, evidence that even
12      some soluble particles may be cleared by mucociliary transport (Bennett and Ilowite, 1989;
13      Matsuietal., 1998).
14
15      7.3.1.3 Alveolar Region
16           Clearance from the A region occurs via a number of mechanisms and pathways. Particle
17      removal by macrophages comprises the main nonabsorptive clearance process in this region.
18      These cells, which reside on the epithelium, phagocytize and transport deposited material that
19      they contact by random motion or via directed migration under the influence of chemotactic
20      factors.
21           Although alveolar macrophages normally comprise up to about 5% of the total alveolar
22      cells in healthy, nonsmoking humans and other mammals, the actual cell count may be altered by
23      particle loading.  The magnitude of any increase in cell number is related to the number of
24      deposited particles rather than to total deposition by weight. Thus, equivalent masses of an
25      identically deposited substance would not produce the same response if particle sizes differed,
26      and the deposition of smaller particles would tend to result in a greater elevation in macrophage
27      number than would deposition of larger particles.
28           Particle-laden macrophages may be cleared from the A region  along a number of pathways.
29      As noted in Figure 7-3, this includes cephalad transport via the mucociliary system after the cells
30      reach the distal terminus of the mucus blanket; movement within the interstitium to a lymphatic
31      channel; or perhaps traversing of the alveolar-capillary endothelium, directly entering the
        March 2001                               7-32        DRAFT-DO NOT QUOTE OR CITE

-------
   1      bloodstream.  Particles within the lymphatic system may be translocated to tracheobronchial
   2      lymph nodes, which can become reservoirs of retained material. Particles subsequently reaching
   3      the postnodal lymphatic circulation will enter the blood.  Once in the systemic circulation, these
   4      particles, or transmigrated macrophages, can travel to extrapulmonary organs. Deposited
   5      particles that are not ingested by alveolar macrophages may enter the interstitium, where they are
   6      subject to phagocytosis by resident interstitial macrophages, and may travel to perivenous,
   7      peribronchiolar or subpleural sites, where they become trapped, increasing particle burden. The
   8      migration and grouping of particles and macrophages within the lungs can lead to the
   9      redistribution of initially diffuse deposits into focal aggregates. Some particles or components
 10      can bind to epithelial cell membranes or macromolecules, or other cell components, delaying
 11      clearance from the lungs.
 12           Churg and Brauer (1997) examined lung autopsy tissue from 10 never-smokers from
 13      Vancouver, Canada. They noted that the geometric mean particle diameter (GMPD) in lung
 14      parenchymal tissue was 0.38 ^m (og = 2.4). Ultrafmes were less than 5% of the total retained
 15      particulate matter.  Metal particles had a GMPD of 0.17 //m and silicates 0.49 f^m. Ninety-six
 16      percent of retained PM was less than 2.5 //m.
 17           Clearance by the absorptive mechanism involves dissolution in the alveolar surface fluid,
 18      followed by transport through the epithelium and into  the interstitium, and diffusion into the
 19      lymph or blood. Although factors affecting the dissolution of deposited particles are poorly
 20      understood, solubility is influenced by the particle's surface to volume ratio and other properties,
 21      such as hydrophilicity and lipophilicity  (Mercer, 1967; Morrow, 1973; Patten, 1996).  Thus, as
 22     noted, materials generally considered to be relatively insoluble  still may have high dissolution
 23     rates and short dissolution half-times if the particle size is small.
 24           Some deposited particles may undergo dissolution in the acidic milieu of the
 25     phagolysosomes after ingestion by macrophages. Intracellular dissolution may be the initial step
 26     in translocation from the lungs for these particles and for material associated with these particles
 27     (Kreyling, 1992; Lundborg et al., 1985). Following dissolution, the material can be absorbed
 28     into the blood. Dissolved materials may then leave the lungs at rates that are more rapid than
29     would be expected based on an "expected" normal dissolution rate in lung fluid.
30
31

        March 2001                                7-33         DRAFT-DO NOT QUOTE OR CITE

-------
  1      7.3.2 Clearance Kinetics
  2           The kinetics of clearance has been reviewed in U.S. Environmental Protection Agency
  3      (1996) and in a number of monographs (e.g., Schlesinger et al., 1997) and is discussed only
  4      briefly here.  The actual time frame over which clearance occurs affects the cumulative dose
  5      delivered to the respiratory tract, as well as that delivered to extrapulmonary organs.
  6
  7      7.3.2.1 Extrathoracic Region
  8           Mucus flow rates in the posterior nasal passages are highly nonuniform, but the median rate
  9      in a healthy adult human is about 5 mm/min, resulting in a mean anterior to posterior transport
10      time of about 10 to 20 min for poorly soluble particles (Rutland and Cole, 1981; Stanley et al.,
11      1985). Particles deposited in the anterior portion of the nasal passages are cleared more slowly
12      by mucus transport, and are usually more effectively removed by sneezing, wiping, or nose
13      blowing (Fry and Black, 1973; Morrow, 1977).
14
15      7.3.2.2 Tracheobronchial Region
16           Mucus transport in the tracheobronchial tree occurs at different rates in different local
17      regions; the velocity of movement is fastest in the trachea, and it becomes progressively slower
18      in more distal airways. In healthy nonsmoking humans, using noninvasive procedures and no
19      anesthesia, average tracheal mucus transport rates have been measured at 4.3 to 5.7 mm/min
20      (Yeates et al., 1975, 1981; Foster et al., 1980; Leikauf et  al.,  1981, 1984), whereas that in the
21      main bronchi has been measured at  -2.4 mm/min (Foster et al., 1980). Estimates for human
22      medium bronchi range between 0.2  to 1.3 mm/min, whereas those in the most distal ciliated
23      airways range down to 0.001 mm/min (Morrow et al., 1967; Cuddihy and Yeh, 1988; Yeates and
24      Aspin, 1978).
25           The total duration of bronchial clearance, or some other time parameter, often is used as an
26      index of mucociliary kinetics.  Although clearance from the TB region is generally rapid, there is
27      experimental evidence, discussed in U.S. Environmental  Protection Agency (1996), that a
28      fraction of material deposited in the TB region is retained much longer than the 24 h commonly
29      used as the outer range of clearance time for particles within this region (Stahlhofen et al.,
30      1986a,b; Scheuch and Stahlhofen, 1988; Smaldone et al., 1988). Some recent studies described

        March 2001                              7-34       DRAFT-DO NOT QUOTE OR CITE

-------
   1      below continue to support the concept that TB regional clearance consists of both a fast and a
  2      slow component.
  3           Falk et al. (1997) studied clearance in healthy adults using monodisperse Teflon particles
  4      (6.2 ^m) inhaled at two flow rates. A considerable fraction (about 50%) of particles deposited in
  5      small airways had not cleared within 24 h following exposure. These particles cleared with a
  6      half time of 50 days. Although the deposition sites of the particles were not confirmed
  7      experimentally, calculations suggested these to be in the smaller ciliated airways. Camner et al.
  8      (1997) also noted that clearance from the TB region was incomplete by 24 h postexposure, and
  9      suggested that this may be caused by incomplete clearance from bronchioles. Healthy adults
 10      inhaled teflon particles (6, 8, and 10 //m) under low flow rates to maximize deposition in the
 11      small ciliated airways.  The investigators noted a decrease in 24-h  retention with increasing
 12      particle size, indicating a shift with increasing size toward either a smaller retained fraction,
 13      deposition more proximally in the respiratory tract, or both. They  calculated that a large fraction,
 14      perhaps as high as 75%, of particles depositing in generations 12 through 16 was still retained at
 15      24 h postexposure.
 16           In a study to examine retention kinetics in the tracheobronchial tree (Falk et al., 1999),
 17      normal nonsmoking adults inhaled radioactively tagged 6.1 -/^m particles at both a normal flow
 18      rate and slow flow rate designed to deposit particles preferentially  in the small ciliated airways.
 19      Lung retention was measured from 24 h to 6 mo after exposure.  Following the normal
 20      inhalation, 14% of the particles retained at 24 h cleared with a half time of 3.7 days, and 86%
 21      with a half time of 217 days. Following the slow inhalation, 35% of the particles retained at 24 h
 22      cleared with a half time of 3.6 days, and 65% with a half time of 170 days. Deposition calculated
 23     using a number of mathematical models indicated higher deposition in the bronchiolar region
 24     (generations 9 through 15) with the slow rate inhalation compared to the normal rate. The
 25      experimental data and predictions of the deposition modeling indicated that 40%  of the particles
 26      deposited in the conducting airways during the slow inhalation were retained after 24 h. The
 27      particles that cleared with the shorter half time were mainly deposited in the bronchiolar region,
 28      but only about 25% of the particles deposited in this region cleared in this phase.  This study
 29      provided additional confirmation for a phase of slow clearance from the bronchial tree.
30           The underlying sites and mechanisms of long-term TB retention in the smaller airways are
31      not known.  Some proposals were presented in the earlier PM AQCD (U.S. Environmental

        March 2001                                7-35        DRAFT-DO NOT QUOTE OR CITE

-------
  1      Protection Agency, 1996).  This slow clearing tracheobronchial compartment likely is associated
  2      with bronchioles <1 mm in diameter (Lay et al., 1995; Kreyling et al., 1999; Falk et al., 1999).
  3      Based on a study in which an adrenergic agonist was used to stimulate mucus flow, so as to
  4      examine the role of mucociliary transport in the bronchioles, it was found that clearance from the
  5      smaller airways was not influenced by the drug, suggesting to the investigators that mucociliary
  6      transport was not as an effective clearance mechanism from this region as in larger airways
  7      (Svartengren et al., 1998, 1999). Although  slower or less effective mucus transport may result in
  8      longer retention times in these small airways, other factors may account for long-term TB
  9      retention.  One such proposal is the movement of particles into the gel phase because of surface
10      tension forces in the liquid lining of the small airways (Gehr et al., 1990, 1991).  The issue of
11      particle retention in the tracheobronchial tree certainly is not resolved.
12           Long-term TB retention patterns are not uniform.  There is an enhancement at bifurcation
13      regions (Radford and Martell,  1977; Henshaw and Fews, 1984; Cohen et al.,  1988), the likely
14      result of both greater deposition and less effective mucus clearance within these areas.  Thus,
15      doses calculated based on uniform surface retention density may be misleading, especially if the
16      material is, toxicologically, slow acting.
17
18      7.3.2.3  Alveolar Region
19           Particles deposited in the A region generally are retained longer than those deposited in
20      airways cleared by mucociliary transport.  There are limited data on alveolar clearance rates in
21      humans. Within any species, reported clearance rates vary widely because, in part, of different
22      properties of the particles used in the various studies.  Furthermore, some chronic experimental
23      studies have employed high concentrations of poorly soluble particles, which may have interfered
24      with normal clearance mechanisms, resulting in clearance rates different from those that would
25      typically occur at lower exposure levels.  Prolonged exposure to high particle concentrations is
26      associated with what is termed particle "overload".  This is discussed later in greater detail in
27      Section 7.4.
28           There are numerous pathways of A region clearance, and the utilization of these may
29      depend on the nature of the particles being cleared.  Little is known concerning relative rates
30      along specific pathways.  Thus, generalizations about clearance kinetics are difficult to make.
31      Nevertheless, A region clearance is usually described as a multiphasic process, each phase

        March 2001                                7-36         DRAFT-DO NOT QUOTE OR CITE

-------
  1     considered to represent removal by a different mechanism or pathway, and often characterized by
  2     increased retention half times following exposure.
  3          The initial uptake of deposited particles by alveolar macrophages is very rapid and
  4     generally occurs within 24 h of deposition (Lehnert and Morrow, 1985; Naumann and
  5     Schlesinger, 1986; Lay et al., 1998).  The time for clearance of particle-laden alveolar
  6     macrophages via the mucociliary system depends on the site of uptake relative to the distal
  7     terminus of the mucus blanket at the bronchiolar level. Furthermore, clearance pathways, and
  8     subsequent kinetics, may depend to some extent on particle size.  For example, some smaller
  9     ultrafme particles (perhaps <0.02 /urn) may be less effectively phagocytosed than larger ones
 10     (Oberdorster, 1993).
 11          Uningested particles may penetrate into the interstitium within a few hours following
 12     deposition. This transepithelial passage seems to increase as particle loading increases,
 13     especially to that level above which macrophage numbers increase (Ferin, 1977; Perm et al.,
 14     1992; Adamson and Bowden, 1981).  It also maybe particle size dependent, because insoluble
 15     ultrafme particles (<0.1 //m diameter) of low intrinsic toxicity show increased access to the
 16     interstitum and greater lymphatic uptake than do larger particles of the same material
 17     (Oberdorster et al.,  1992; Ferin et al., 1992).  However, ultrafme particles of different materials
 18     may not enter the interstitium to the same extent.  Similarly, a depression of phagocytic activity,
 19     a reduction in macrophage ability to migrate to sites of deposition (Madl et al., 1998), or the
 20     deposition of large numbers of ultrafme particles may increase the number of free particles in the
 21     alveoli, perhaps enhancing removal by other routes. In any case, free particles may reach the
 22     lymph nodes, perhaps within a few days after deposition (Lehnert et al., 1988; Harmsen et al.,
 23     1985), although this route is not certain and may be species dependent.
 24          The extent of lymphatic uptake of particles may depend on the effectiveness of other
 25     clearance pathways, in that lymphatic translocation probably increases when phagocytic activity
26     of alveolar macrophages is decreased.  This may be a factor in lung overload. However, it seems
27     that the deposited mass or number of particles must exceed some threshold below which
28     increases in loading do not affect translocation rate to the lymph nodes  (Ferin and Feldstein,
29     1978; LaBelle and Brieger, 1961).  In addition, the rate of translocation to the lymphatic system
30     may be somewhat particle size dependent. Although no human data are available, translocation
31      of latex particles to the lymph nodes of rats was greater for 0.5- to 2-^m particles than for 5- and

        March 2001                               7-37        DRAFT-DO NOT QUOTE OR CITE

-------
  1      9-yUm particles (Takahashi et al., 1992), and smaller particles within the 3- to 15-yum size range
  2      were found to be translocated at faster rates than were larger sizes (Snipes and Clem, 1981).
  3      On the other hand, translocation to the lymph nodes was similar for both 0.4-,um barium sulfate
  4      or 0.02-,um gold colloid particles (Takahashi et al., 1987). It seems that particles <2 ,wm clear to
  5      the lymphatic system at a rate independent of size, and it is particles of this size, rather than those
  6      > 5 /u.m, that would have significant deposition within the A region following inhalation,  hi any
  7      case, the normal rate of translocation to the lymphatic system is quite slow, and elimination from
  8      the lymph nodes is even slower, with half times  estimated in tens of years (Roy, 1989).
  9           Soluble particles depositing in the A region may be cleared rapidly via absorption through
10      the epithelial surface into the blood. Actual rates depend on the size of the particle (i.e., solute
11      size), with smaller molecular weight solutes clearing faster than larger ones.  Absorption may be
12      considered as a two stage process, with the first  stage being dissociation of the deposited
13      particles into material that can be absorbed into the circulation (i.e., dissolution), and the second
14      stage being uptake of this material. Each of these stages may be time dependent. The rate of
15      dissolution depends on a number of factors, including particle surface area and chemical
16      structure. A portion of the dissolved material may be absorbed more slowly because of binding
17      to respiratory tract components. Accordingly, there is a very wide range for absorption rates,
18      depending on the physicochemical  properties of the material deposited.
19
20      7.3.3  Interspecies Patterns of Clearance
21           The inability to study the retention of certain materials in humans for direct risk assessment
22      requires use of laboratory animals.  Because dosimetry depends on clearance rates and routes,
23      adequate toxicologic assessment necessitates that clearance kinetics in these animals be related to
24      those in humans. The basic mechanisms and overall patterns of clearance from the respiratory
25      tract are similar in humans and most other mammals. However, regional clearance rates can
26      show substantial variation between species, even for similar particles deposited under
27      comparable exposure conditions, as extensively reviewed elsewhere (U.S. Environmental
28      Protection Agency, 1996; Schlesinger et al., 1997; Snipes et al., 1989).
29           In general, there are species-dependent rate constants for various clearance pathways.
30      Differences in regional and total clearance rates between some  species are a reflection of
31      differences in mechanical clearance processes. For example, the relative proportion of particles
        March 2001                               7-38       DRAFT-DO NOT QUOTE OR CITE

-------
  1      cleared from the A region in the short and longer term phases differs between laboratory rodents
  2      and larger mammals, with a greater percentage cleared in the faster phase in rodents. A recent
  3      study (Oberdorster et al., 1997) showed interstrain differences in mice and rats in the handling of
  4      particles by alveolar macrophages. Macrophages of B6C3F1 mice could not phagocytize 10-//m
  5      particles, but those of CSV black/6Jmice did. In addition, the nonphagocytized lQ-/um particles
  6      were efficiently eliminated from the alveolar region, whereas previous work in rats found that
  7      these large particles, after uptake by macrophages, were retained persistently (Snipes and Clem,
  8      1981; Oberdorster et al., 1992).  The end result of interspecies differences in  clearance for
  9      consideration in assessing particle dosimetry is that the retention of deposited particles can differ
 10      between species, and this may result in differences in  response to similar particulate exposure
 11      atmospheres.
 12           Hsieh and Yu (1998) summarized the existing data on pulmonary clearance of inhaled,
 13      poorly soluble particles in the rat, mouse, guinea pig,  dog, monkey, and human.  Clearance at
 14      different initial lung burdens, ranging from  0.001 to 10 mg particles/g lung, was analyzed using a
 15      two-phase exponential decay function. Two clearance phases in the alveolar region, namely fast
 16      and slow, were associated with mechanical  clearance  along two pathways, the former with the
 17      mucociliary system and the latter with the lymph nodes. Rats and mice were noted to be fast
 18      clearers compared to the other species. Increasing the initial lung burden resulted in an
 19      increasing mass fraction of particles cleared by the slower phase. As lung burden increased
 20      beyond 1  mg particles/g lung, the fraction cleared by the slow phase increased to almost 100%
 21      for all species. However, the rate for the fast phase was similar in all species and did not change
 22      with increasing lung burden of particles, while the rate for the slow phase decreased with
 23      increasing lung burden. At elevated burdens, the "overload" effect on clearance  rate was greater
 24      in rats than in humans, an observation consistent with previous findings (Snipes, 1989).
 25
 26      7.3.4 Biological Factors  Modulating Clearance
 27           A number of factors have been assessed in terms of modulation of normal clearance
28      patterns.  These include aging, gender, workload, disease, and irritant inhalation, and have been
29      discussed in detail previously (U.S. Environmental Protection Agency, 1996).
30
31
        March 2001                               7-39         DRAFT-DO NOT QUOTE OR CITE

-------
  1      7.3.4.1  Age
  2           Studies previously described (U.S. Environmental Protection Agency, 1996) indicated that
  3      there appeared to be no clear evidence for any age-related differences in clearance from the
  4      respiratory tract, either from child to adult, or young adult to elderly.  Studies of mucociliary
  5      function have shown either no changes or some slowing in mucous clearance function with age
  6      after maturity, but at a rate that would be unlikely to significantly affect overall clearance
  7      kinetics.
  8
  9      7.3.4.2  Gender
10           Previous studies (U.S. Environmental Protection Agency, 1996) indicated no gender related
11      differences in nasal mucociliary clearance rates in children (Passali and Bianchini Ciampoli,
12      1985) nor in tracheal transport rates in adults (Yeates et al., 1975).
13
14      7.3.4.3  Physical Activity
15           The effect of increased physical activity on mucociliary clearance is unresolved, with
16      previously discussed  studies (U.S. Environmental Protection Agency, 1996) indicating either no
17      effect or an increased clearance rate with exercise. However, it is possible to have an enhanced
18      mucus transport by nonmucociliary mechanisms such as a two-phase gas-liquid interaction.
19      During exercise, breathing patterns become similar to "huffing", fast expiration compared to
20      inspiration. With this breathing mode, effective mucus transport has been demonstrated in
21      simulated airway models (Kim et al., 1987).  There are no data concerning changes in A region
22      clearance with increased activity levels. Breathing with an increased tidal volume was noted to
23      increase the rate of particle clearance from the A region, and this was suggested to result from
24      distension-related evacuation of surfactant into proximal airways, resulting in a facilitated
25      movement of particle-laden macrophages or uningested particles because of the accelerated
26      motion of the alveolar fluid film (John et al., 1994).
27
28      7.3 A A  Respiratory Tract Disease
29           Various respiratory tract diseases are associated with clearance alterations. The
30      examination of clearance in individuals with lung disease requires careful interpretation of results
31      because differences in deposition of particles used to assess clearance function may occur

        March 2001                                 7-40         DRAFT-DO NOT QUOTE OR CITE

-------
  1     between normal individuals and those with respiratory disease; this would impact directly on the
  2     measured clearance rates, especially in the tracheobronchial tree. Earlier studies reported in U.S.
  3     Environmental Protection Agency (1996) noted findings of slower nasal mucociliary clearance in
  4     humans with chronic sinusitis, bronchiectasis, rhinitis, or cystic fibrosis and slowed bronchial
  5     mucus transport associated with bronchial carcinoma, chronic bronchitis, asthma, and various
  6     acute respiratory infections. However, a recent study by Svartengren et al. (1996a) concluded,
  7     based on deposition and clearance patterns, that particles cleared equally effectively from the
  8     small ciliated airways of healthy humans and those with mild to moderate asthma.  However, this
  9     similarity was ascribed to effective therapy for the asthmatics.
 10          In another study, Svartengren et al. (1996b) examined clearance from the TB region in
 11     adults with chronic bronchitis who inhaled 6-/zm Teflon particles.  Based on calculations,
 12     particle deposition was assumed to be in small ciliated airways at low flow and in larger airways
 13     at higher flow.  The results were compared to that obtained in healthy subjects from other
 14     studies. At low flow, a larger fraction of particles was retained over 72 h in people with chronic
 15     bronchitis compared to healthy subjects, indicating that clearance resulting from spontaneous
 16     cough could not fully compensate for impaired mucociliary transport in small airways.  For larger
 17     airways, patients with chronic bronchitis cleared a larger fraction of the deposited particles over
 18     72 h than did healthy subjects, but this was reportedly because of differences in deposition
 19     resulting from airway obstruction.
 20          An important mechanism of clearance from the tracheobronchial region, under some
 21     circumstances, is cough. Although cough is generally a reaction to an inhaled stimulus, in some
 22     individuals  with respiratory disease, spontaneous coughing also serves to clear the upper
 23     bronchial airways of deposited substances by dislodging mucus from the airway surface. Recent
 24     studies confirm that this mechanism likely plays a significant role in clearance for people with
 25     mucus hypersecretion, at least for the upper bronchial tree, and for a wide range of deposited
26     particle sizes (0.5 to 5 /wm) (Toms et al., 1997; Groth et al., 1997).  There appears to be a general
27     trend towards an association between the extent (i.e., number) of spontaneous coughs and the rate
28     of particle clearance, with faster clearance associated with a greater number of coughs (Groth
29     et al., 1997). Thus, recent evidence continues to support cough as an adjunct to mucociliary
30     movement in the removal of particles from the lungs of individuals with COPD. However, some
31      recent evidence suggests that, like mucociliary function, cough-induced clearance may become

        March 2001                                7-41         DRAFT-DO NOT QUOTE OR CITE

-------
 1      depressed with worsening airway disease. Noone et al. (1999) found that the efficacy of
 2      clearance via cough in patients with primary ciliary dyskinesia, who rely on coughing for
 3      clearance because of immotile cilia, correlated with lung function (FEV1), in that decreased
 4      cough clearance was associated with decreased percentage of predicted FEV1.
 5           Earlier reported studies (U.S. Environmental Protection Agency, 1996) indicated that rates
 6      of A region particle clearance were reduced in humans with chronic obstructive lung disease and
 7      in laboratory animals with viral infections, whereas the viability and functional activity of
 8      macrophages was impaired in human asthmatics and in animals with viral induced lung
 9      infections. However, any modification of functional properties of macrophages appears to be
10      injury specific, in that they reflect the nature and anatomic pattern of disease.
11           A factor that may affect clearance of particles is the integrity of the epithelial surface lining
12      of the lungs. Damage or injury to the epithelium may result from disease or from the inhalation
13      of chemical irritants.  Earlier studies performed with particle instillation had shown that alveolar
14      epithelial damage at the time of deposition in mice resulted in increased translocation of inert
15      carbon to pulmonary interstitial macrophages (Adamson and Hedgecock, 1995). A similar
16      response was observed in a more recent assessment (Adamson and  Prieditis,  1998), whereby
17      silica (<0.3 /urn) was instilled into a lung having alveolar epithelial  damage, as evidenced by
18      increased permeability, and particles were noted to reach the interstitium and lymph nodes.
19
20
21      7.4 PARTICLE OVERLOAD
22           Experimental  studies using some laboratory rodents have employed high exposure
23      concentrations of relatively nontoxic, poorly soluble particles.  These particle loads interfered
24      with normal clearance mechanisms, producing clearance rates different from those that would
25      occur at lower exposure levels.  Prolonged exposure to high particle concentrations is associated
26      with a phenomenon that has been termed particle "overload", defined as the overwhelming of
27      macrophage-mediated clearance by the deposition of particles at a rate that exceeds the capacity
28      of that clearance pathway. It has been hypothesized that in the rat, overload will begin when
29      deposition approaches 1 mg particles/g lung tissue (Morrow, 1988). Overload is a nonspecific
30      effect noted in experimental studies using many different kinds of poorly soluble particles and
31      results in A region clearance slowing or stasis, with an associated chronic inflammation and
        March 2001                               7-42       DRAFT-DO NOT QUOTE OR CITE

-------
  1      aggregation of macrophages in the lungs and increased translocation of particles into the
  2      interstitium.
  3           The relevance of lung overload to humans exposed to poorly soluble, nonfibrous particles
  4      remains unclear.  Although it is likely to be of little relevance for most "real world" ambient
  5      exposures, it may be of concern in interpreting some long-term experimental exposure data and,
  6      perhaps, also for occupational exposures. For example, it has been suggested that a condition
  7      called progressive massive fibrosis, which is unique to humans, has features indicating that dust
  8      overload is a factor in its pathogenesis (Green, 2000).  This condition is associated with
  9      cumulative dust exposure and impaired clearance, and can occur following high exposure
 10      concentrations associated with occupational situations. In addition, the relevance to humans is
 11      clouded by the suggestion that macrophage-mediated clearance is normally slower and perhaps of
 12      less relative importance in overall clearance in humans than in rats (Morrow, 1994), and that
 13      there can be significant differences in macrophage loading between species.  On the other hand,
 14      overload may be a factor in individuals with compromised lungs under normal exposure
 15      conditions. Thus, it has been hypothesized (Miller et al., 1995) that localized overload of particle
 16      clearance mechanisms in people with compromised lung status may occur, whereby these
 17      mechanisms are overwhelmed, resulting in morbidity or mortality from particle exposure.
 18
 19
 20      7.5 COMPARISON OF DEPOSITION AND CLEARANCE PATTERNS OF
 21          PARTICLES ADMINISTERED BY INHALATION AND
 22          INTRATRACHEAL INSTILLATION
 23           The most relevant exposure route to evaluate the toxicity of particulate matter is inhalation.
 24      However, many studies delivered particles by intratracheal instillation.  This latter technique has
 25      been used because it is easy to perform; requires significantly less effort, cost, and amount of test
 26      material than does inhalation; and can deliver a known, exact dose of a toxicant to the lungs.
 27      Because particle disposition is a determinant of dose, it is important to compare deposition and
28      clearance of particles delivered by these two routes. However, in most instillation studies, the
29      effect of this route of administration on particle deposition and clearance per se was not
30      examined. Although these parameters were evaluated in some studies, it has been very difficult
31      to compare particle deposition/clearance between  different inhalation and instillation studies

        March 2001                               7-43        DRAFT-DO NOT QUOTE OR CITE

-------
 1      because of differences in experimental procedures and in the manner by which particle
 2      deposition/clearance was quantitated. A recent paper provided a detailed evaluation of the role
 3      of instillation in respiratory tract dosimetry and toxicology studies (Driscoll et al., 2000), and a
 4      short summary derived from this paper is provided in this section.
 5           The pattern of initial regional deposition is strongly influenced by the exposure technique
 6      used.  Furthermore, the patterns within specific respiratory tract regions also are influenced in
 7      this regard. Depending on particle size, inhalation results in varying degrees of deposition within
 8      the ET airways, a region that is completely bypassed by instillation.  Thus, differences in amount
 9      of particles deposited in the lower airways will occur between the two procedures.
10           The exposure technique also influences the intrapulmonary distribution of particles, which
11      potentially would affect routes and rates of ultimate clearance from the lungs and dose delivered
12      to specific sites within the respiratory tract or to extrapulmonary organs.  Intratracheal instillation
13      tends to disperse particles fairly evenly within the tracheobronchial tree, but can result in
14      heterogeneous distribution in the alveolar region, whereas inhalation tends to produce a more
15      homogeneous distribution throughout the major conducting airways as well as the alveolar region
16      for the same particles. Thus, inhalation results in a randomized distribution of particles within
17      the lungs, whereas intratracheal instillation produces an heterogeneous distribution in that the
18      periphery of the lung receives little particle load and most of the instilled particles are found in
19      regions that have a short path length from the major airways. Furthermore, inhalation results in
20      greater deposition in apical areas of the lungs and less in basal areas, whereas intratracheal
21      instillation results in less apical than basal deposition.
22           Comparison of the kinetics of clearance of particles administered by instillation or
23      inhalation have shown similarities, as well as differences, in rates for different clearance phases,
24      dependent on the exposure technique used. However, some of the differences in kinetics may be
25      explained by differences in the  initial sites of deposition.
26           One  of the major pathways of clearance involves particle uptake and removal via
27      pulmonary macrophages.  Domes and Valberg (1992) noted that inhalation resulted in a lower
28      percentage of particles recovered in lavaged cells and a more even distribution of particles among
29      macrophages. More individual cells received measurable amounts of particles via inhalation than
30      via intratracheal instillation, whereas with the latter, many cells received little or no particles,
31      although others received very high burdens. Furthermore,  with intratracheal instillation,

        March 2001                                7-44         DRAFT-DO NOT QUOTE OR CITE

-------
  1      macrophages at the lung periphery contained few, if any, particles, whereas cells in the regions of
  2      highest deposition were overloaded, reflecting the heterogeneity of particle distribution when
  3      particles are administered via instillation.  Thus, the route of exposure influences the particle
  4      distribution in the macrophage population and could, by assumption, influence clearance
  5      pathways and clearance kinetics.
  6           In conclusion, inhalation may result in deposition within the ET region, the extent of which
  7      depends on the size of the particles used. Of course, intratracheal instillation bypasses this
  8      portion of the respiratory tract and delivers particles directly to the tracheobronchial tree.
  9      Although some  studies indicate that short (0 to 2 days) and long (100 to 300 days postexposure)
 10      phases of clearance of insoluble particles delivered either by inhalation or intratracheal
 11      instillation are similar, other studies indicate that the percentage retention of particles delivered
 12      by instillation is greater than that for inhalation, at least up to 30 days postexposure. Thus, there
 13      is some inconsistency. Perhaps the most consistent conclusion regarding differences between
 14      inhalation and intratracheal instillation is related to the intrapulmonary distribution of particles.
 15      Inhalation generally results in a fairly homogeneous distribution of particles throughout the
 16      lungs. On the other hand, instillation results in a heterogeneous distribution, especially within
 17      the alveolar region, and focally high concentrations of particles. The bulk of instilled material
 18      penetrates beyond the major tracheobronchial airways, but the lung periphery is often virtually
 19      devoid of particles. This difference is reflected in particle burdens within macrophages, with
20      those from animals inhaling particles being burdened more homogeneously and those from
21      animals with instilled particles showing some populations of cells with no particles and others
22      with heavy burdens. This difference reflects on clearance pathways, dose to cells and tissues,
23      and systemic absorption. Exposure method, thus, clearly influences dose distribution.
24
25
26      7.6 MODELING THE DISPOSITION OF PARTICLES IN THE
27          RESPIRATORY TRACT
28      7.6.1  Modeling Deposition and Clearance
29          The biologic effects of inhaled particles are a function of their disposition. This, in turn,
30      depends on their patterns of both deposition and clearance. Removal of deposited materials
31      involves the competing processes of macrophage-mediated clearance and dissolution-absorption.
        March 2001                                7-45         DRAFT-DO NOT QUOTE OR CITE

-------
 1      Over the years, mathematical models for predicting deposition, clearance and, ultimately,
 2      retention of particles in the respiratory tract have been developed. Such models help interpret
 3      experimental data and can be used to make predictions of deposition for cases where data are not
 4      available.
 5           A review of various mathematical deposition models was given by Morrow and Yu (1993)
 6      and in U.S. Environmental Protection Agency (1996). There are three major elements involved
 7      in mathematical modeling. First, a structural model of the airways must be specified in
 8      mathematical terms.  Second, deposition efficiency in each airway must be derived for each of
 9      the various deposition mechanisms.  Finally, a computational procedure must be developed to
10      account for the transport and deposition of the particles in the airways.  As noted earlier, most
11      models are deterministic, in that particle deposition probabilities are calculated using anatomical
12      and airflow information on an airway generation by airway generation basis.  Other models are
13      stochastic, whereby modeling is performed using individual particle trajectories and finite
14      element simulations of airflow.
15           Recent reports involve modeling the deposition of ultrafine particles and deposition at
16      airway bifurcations.  Zhang and Martonen (1997) used a mathematical model to simulate
17      diffusion deposition of ultrafine particles in the human upper tracheobronchial tree and compared
18      the results to those in a hollow cast obtained by Cohen et al. (1990). The model was in good
19      agreement with experimental data. Zhang and Martonen (1997) studied the inertial deposition of
20      particles in symmetric three-dimensional models of airway bifurcations, mathematically
21      examining effects of geometry and flow. They developed equations for use in predicting
22      deposition based on Stokes numbers, Reynolds numbers, and bifurcation angles for specific
23      inflows.
24           Models for deposition, clearance, and dosimetry of the respiratory tract of humans have
25      been available for the past four decades. The International Commission on Radiological
26      Protection (ICRP) has recommended three different mathematical models during this time period
27      (International Commission on Radiological Protection, 1960, 1979,  1994). These models make
28      it possible to calculate the mass deposition and retention in different parts of the respiratory tract
29      and provide, if needed, mathematical descriptions of the translocation of portions of the
30      deposited material to other organs and tissues beyond the respiratory tract.


        March 2001                               7-46        DRAFT-DO NOT QUOTE OR CITE

-------
  1          A morphological model based on laboratory data from planar gamma camera and single-
  2     photon emission tomography images has been developed (Martonen et al., 2000; Segal et al.,
  3     2000b). This model defines the parenchymal wall in mathematical terms, divides the lung into
  4     distinct left and right components, derives a set of branching angles from experimental
  5     measurements, and confines the branching network within the left and right components (so there
  6     is no overlapping of airways).  The authors conclude that this more physiologically realistic
  7     model can be used to calculate PM deposition patterns for risk assessment.
  8          Musante and Martonen (2000c) developed an age-dependent theoretical model to predict
  9     dosimetry in the lungs of children. The model comprises dimensions of individual airways and
 10     geometry of branching airway networks within developing lungs and breathing parameters as a
 11     function of age. The model suggests that particle size, age, and activity level markedly affect
 12     deposition patterns of inhaled particles. Simulations thus far predict a lung deposition fraction of
 13     38% in an adult and 73%, nearly twice as high, in a 7-mo-old for 2/2-particles inhaled during
 14     heavy breathing. The authors conclude that use of this model will be useful for estimating dose
 15     delivered to sensitive subpopulations, such as children.
 16          Segal et al. (2000a) developed a computer model for airflow and particle motion in the
 17     lungs of children to study how airway disease, specifically cancer, affects inhaled PM deposition.
 18     The model considers how tumor characteristics (size and location) and ventilatory parameters
 19     (breathing rates and tidal volumes) influence particle trajectories and deposition patterns.  The
 20     findings indicate that PM may be deposited on the upstream surfaces of tumors because of
 21     enhanced efficiency of inertial impaction. Also,  submicron particles and larger particles,
 22     respectively, may be deposited on the downstream surfaces of tumors because of enhanced
 23     efficiency of diffusion and sedimentation. The mechanisms of diffusion and sedimentation are
 24     functions of the particle residence times in airways.  Eddies downstream of tumors would trap
 25     particles and allow more time for deposition to occur by diffusion and sedimentation. The
 26     authors conclude that particle deposition is complicated by the presence  of airway disease but
27     that the effects are systematic and predictable.
28           Broday and Georgopoulos (2000) recently have presented a model that solves a variant of
29      the general dynamic equation for size evolution of respirable particles within human
30      tracheobronchial airways.  The model considers polydisperse aerosols with respect to size and
31      heterosperse with respect to thermodynamic state and chemical composition. The  aerosols  have

        March 2001                               7-47        DRAFT-DO NOT QUOTE OR CITE

-------
  1      an initial bimodal lognormal size distribution that evolves with time in response to condensation-
  2      evaporation and deposition processes.  Simulations reveal that submicron size particles grow
  3      rapidly and cause increased number and mass fractions of the particle population to be found in
  4      the intermediate size range. Because deposition by diffusion decreases with increasing size, fine
  5      hygroscopic particles persist longer in the inspired air than nonhygroscopic particles of
  6      comparable initial size distribution. In contrast, the enhanced deposition fraction of hygroscopic
  7      particles, initially from the intermediate size range, increases their deposition fraction in the
  8      airways. The model demonstrates that the combined effect of growth and deposition tends to
  9      decrease the size nonuniformiry of persistent particles in the airways and form an aerosol that is
10      characterized by a smaller variance; these factors also alter the deposition profile along airways.
11           Another respiratory tract dosimetry model was developed, concurrently with the new ICRP
12      model, by the National Council on Radiation Protection and Measurements (NCRP) (1997).
13      As with the ICRP model (International Commission on Radiological Protection, 1994), the  new
14      NCRP model addresses inhalability of particles, revised subregions of the respiratory tract,
15      dissolution-absorption as an important aspect of the model, and body size and age.  The NCRP
16      model defines the respiratory tract in terms of a naso-oro-pharyngo-laryngeal (NOPL) region, a
17      tracheobronchial (TB) region, a pulmonary (P) region, and lung-associated lymph nodes (LN).
18      Deposition and clearance are calculated separately for each of these regions. As with the 1994
19      ICRP model, inhalability of aerosol particles is considered, and deposition in the various regions
20      of the respiratory tract is modeled using methods that relate to mechanisms of inertial impaction,
21      sedimentation, and diffusion.
22           Fractional deposition in the NOPL region was developed from empirical relationships
23      between particle diameter and air flow rate.  Deposition in the TB and P regions were projected
24      from model calculations based on geometric or aerodynamic particle diameter and physical
25      deposition mechanisms such as impaction, sedimentation, diffusion, and interception.
26      Deposition in the TB and P regions used the lung model of Yeh and Schum (1980), with a
27      method of calculation similar to that of Findeisen (1935) and Landahl (1950). This method was
28      modified to accomodate an adjustment of lung volume and substitution of realistic deposition
29      equations.  These calculations were based on air flow information and idealized morphometry,
30      using a typical pathway model. Comparison of regional deposition fraction predictions between
31      the NCRP and ICRP models was provided in U.S. Environmental Protection Agency (1996).

        March 2001                               7-48        DRAFT-DO NOT QUOTE OR CITE

-------
  1     Inhalability was defined as per the American Conference of Governmental Industrial Hygenists
  2     (1985) definition. Breathing frequency, tidal volume, and functional residual capacity are the
  3     ventilatory factors used to model deposition. These were related to body weight and to three
  4     levels of physical activity, namely low activity, light exertion and heavy exertion.
  5          Clearance from all regions of the respiratory tract was considered to result from
  6     competitive mechanical and absorptive mechanisms. Mechanical clearance in the NOPL and TB
  7     regions was considered to result from mucociliary transport. This was represented in the model
  8     as a series of escalators moving towards the glottis and where each airway had an effective
  9     clearance velocity.  Clearance from the P region was represented by fractional daily clearance
 10     rates to the TB region, the pulmonary LN region, and the blood. A fundamental assumption in
 11     the model was that the rates for absorption into blood were the same in all regions of the
 12     respiratory tract; the rates of dissolution-absorption of particles and their constituents were
 13     derived from clearance data primarily from laboratory animals. The effect of body growth on
 14     particle deposition also was considered in the model, but particle clearance rates were assumed to
 15     be independent of age. Some consideration for compromised individuals was incorporated into
 16     the model by altering rates (compared to normal) for the NOPL and TB regions.
 17          Mathematical deposition models for deposition in a number of nonhuman species have
 18     been developed and discussed previously (U.S. Environmental Protection Agency, 1996).
 19     Despite difficulties, modeling studies in laboratory animals remain a useful step in extrapolating
 20     exposure-dose-response relationships from laboratory animals to human. Some additional work
 21      on modeling deposition in animals has been reported, but it merely expands on work and
 22     approaches already noted in the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996).
 23           Respiratory-tract clearance begins immediately on deposition of inhaled particles. Given
 24      sufficient time, the deposited particles may be removed completely by these clearance processes.
 25      However, single inhalation exposures may be the exception rather than the rule.  It generally is
 26      accepted that repeated or chronic exposures are common for environmental aerosols. As a result
 27      of such exposures, accumulation of particles may occur.  Chronic exposures produce respiratory
28      tract burdens of inhaled particles that continue to increase with time until the rate of deposition is
29      balanced by the rate of clearance.  This is defined as the "equilibrium respiratory tract burden".
30           It is important to evaluate these accumulation patterns, especially when assessing ambient
31      chronic exposures, because they dictate what the equilibrium respiratory tract burdens of inhaled

        March 2001                               7-49        DRAFT-DO NOT QUOTE OR CITE

-------
  1      particles will be for a specified exposure atmosphere.  Equivalent concentrations can be defined
  2      as "species-dependent concentrations of airborne particles which, when chronically inhaled,
  3      produce equal lung deposits of inhaled particles per gram of lung during a specified exposure
  4      period" (Schlesinger et al., 1997). Available data and approaches to evaluate exposure
  5      atmospheres that produce similar respiratory tract burdens in laboratory animals and humans
  6      have been discussed in detail in the previous criteria document.
  7           Several laboratory animal models have been developed to help interpret results from
  8      specific studies that involved chronic inhalation exposures to nonradioactive particles (Wolff
  9      et al., 1987; Strom et al., 1988; Stober et al., 1994). These models were adapted to data from
10      studies involving high level chronic inhalation exposures in which massive lung burdens of low
11      toxicity, poorly soluble particles were accumulated, but the models have not been adapted to
12      chronic exposures to low concentrations of aerosols in which particle overload does not occur.
13           Asgharian et al. (2000) described a method for calculating a deposited fraction for a
14      specific size distribution based on a summary of published data on regional deposition of
15      different size particles. The method is based on constructing nomograms that are used to
16      estimate alveolar deposition fractions for three species (human, monkey, and rat).  The data is
17      then incorporated into a regression model that calculates more exact deposition fractions. The
18      model is somewhat constrained at present because of limitations in the underlying deposition
19      database.
20           Hofmann et al. (2000) used three different morphometric models of the rat lung to compute
21      particle deposition in the acinar airways: the multipath lung model (MPL), with a fixed airway
22      geometry; the stochastic lung (SL) model, with  a randomly selected branching structure; and a
23      hybrid of the MPL and SL models. They calculated total and regional deposition for  a range of
24      particle sizes during quiet and heavy breathing.  Although the total bronchial and acinar
25      deposition fractions were similar for the three models, the SL and the hybrid models predicted a
26      substantial variation in particle deposition among different acini. Acinar deposition variances in
27      the MPL model were consistently smaller than in the SL and the hybrid lung models. The
28      authors conclude that the similarity of acinar deposition variations in the latter two models and
29      their independence of the breathing pattern suggest the heterogeneity of the acinar airway
30      structure is primarily responsible for the heterogeneity of acinar particle deposition.
31

        March 2001                               7-50        DRAFT-DO NOT QUOTE OR CITE

-------
  1     7.6.2  Models To Estimate Retained Dose
  2          Models have been used routinely to express retained dose in terms of temporal patterns for
  3     alveolar retention of acutely inhaled materials. Available information for a variety of
  4     mammalian species and humans can be used to predict deposition patterns in the respiratory tract
  5     for inhalable aerosols with reasonable degrees of accuracy. Additionally, alveolar clearance data
  6     for mammalian species commonly used in inhalation studies are available from numerous
  7     experiments that involved small amounts of inhaled radioactive particles.
  8          An important factor in using models to predict retention patterns in laboratory animals or
  9     humans is the dissolution-absorption rate of the inhaled material. Factors that affect the
 10     dissolution of materials or the leaching of their constituents in physiological  fluids, and the
 11     subsequent absorption of these constituents, are not fully understood.  Solubility is known to be
 12     influenced by the surface-to-volume ratio and other surface properties of particles (Mercer, 1967;
 13     Morrow, 1973).  The rates at which dissolution and absorption processes occur are influenced by
 14     factors that include the chemical composition of the material. Temperature history of materials is
 15     an important consideration for some metal oxides. For example, in controlled laboratory
 16     environments, the solubility of oxides usually decreases when the oxides are produced at high
 17     temperatures, which generally results in compact particles having small surface-to-volume ratios.
 18     It is sometimes possible to accurately predict dissolution-absorption characteristics of materials
 19     based on physical/chemical considerations; but, predictions for in vivo dissolution-absorption
 20     rates for most materials, especially if they contain multivalent cations or anions, should be
 21     confirmed experimentally.
 22          Phagocytic cells, primarily macrophages, clearly play a role in dissolution-absorption of
 23     particles retained in the respiratory tract (Kreyling, 1992). Some particles dissolve within the
 24     phagosomes because of the acidic milieu in those organelles (Lundborg et al., 1984, 1985), but
 25     the dissolved material may remain associated with the phagosomes or other organelles in the
 26     macrophage rather than diffuse out of the macrophage to be absorbed and transported elsewhere
 27     (Cuddihy, 1984). This same phenomenon has been reported for organic materials. For example,
28     covalent binding of benzo[a]pyrene or metabolites to cellular macromolecules resulted in an
29     increased alveolar retention time for that compound after inhalation exposures of rats  (Medinsky
30     and Kampcik, 1985). Understanding these phenomena and recognizing species similarities and

        March 2001                                7-51         DRAFT-DO NOT QUOTE OR CITE

-------
  1      differences are important for evaluating alveolar retention and clearance processes and
 2      interpreting results of inhalation studies.
 3           Dissolution-absorption of materials in the respiratory tract is clearly dependent on the
 4      chemical and physical attributes of the material. Although it is possible to predict rates of
 5      dissolution-absorption, it is prudent to experimentally determine this important clearance
 6      parameter.  It is important to understand the impact of this clearance process for the lung, TLNs,
 7      and other body organs that might receive particles, or their constituents that enter the circulatory
 8      system from the lung.
 9           Insufficient data were available to adequately model long-term retention of particles
10      deposited in the conducting airways of any mammalian species at the time of the previous
11      document, and this remains the case. Additional research must be done to provide the
12      information needed to properly evaluate retention  of particles in conducting airways.
13           However, a number of earlier studies discussed in the previous document and in
14      Section 7.2.2.2 herein noted that some particles were retained for relatively long times in the
15      upper respiratory tract and tracheobronchial regions, effectively contradicting the  general
16      conclusion that almost all inhaled particles that deposit in the TB region clear within hours or
17      days. These studies have demonstrated that variable portions of the particles that deposit in, or
18      are cleared through, the TB region are retained with half times on the order of weeks or months.
19      Long-term retention and clearance patterns for particles that deposit in the head airways and TB
20      region must continue to be thoroughly evaluated because of the implications of this information
21      for respiratory tract dosimetry and risk assessment.
22           Model projections are possible for the A region using the cumulative information in the
23      scientific literature relevant to deposition, retention, and clearance of inhaled particles.
24      Clearance parameters for six laboratory animal species were summarized in U.S. Environmental
25      Protection Agency (1996).  Recently, Nikula et al. (1997) evaluated results in rats exposed to
26      high levels of either diesel soot or coal dust.  Although the amount of retained material was
27      similar in both species, the rats retained a greater portion in the lumens of the alveolar ducts and
28      alveoli  than did monkeys, whereas the monkeys retained a greater portion of the material in the
29      interstitium than did rats. The investigators concluded that intrapulmonary retention patterns in
30      one species may not be predictive of those in another species at high levels of exposure, but this
31      may not be the case at lower levels.

        March 2001                               7-52        DRAFT-DO NOT QUOTE OR CITE

-------
   1      REFERENCES

   2      Adamson, I. Y. R.; Bowden, D. H. (1981) Dose response of the pulmonary macrophagic system to various
   3            particulates and its relationship to transepithelial passage of free particles. Exp. Lung Res. 2:  165-175.
   4      Adamson, I. Y. R.; Hedgecock, C. (1995) Patterns of particle deposition and retention after instillation to mouse
   5            lung during acute injury and fibrotic repair. Exp. Lung Res. 21: 695-709.
   6      Adamson, I. Y. R.; Prieditis, H. (1998) Silica deposition in the lung during epithelial injury potentiates fibrosis and
   7            increases particle translocation to lymph nodes. Exp. Lung Res. 24: 293-306.
   8      American Conference of Governmental Industrial Hygienists (ACG1H). (1985) Particle size-selective sampling in
   9            the workplace: report of the ACGIH technical committee on air sampling procedures. Cincinnati, OH:
 10            American Conference of Governmental Industrial Hygienists.
 11      Asgharian, B.; Wood, R.; Schlesinger, R. B. (1995) Empirical modeling of particle deposition in the alveolar region
 12            of the lungs: a basis for interspecies extrapolation. Fundam. Appl. Toxicol. 27: 232-238.
 13      Bennett, W. D.; Ilowite, J. S. (1989) Dual pathway clearance of 9"Tc-DTPA from the bronchial mucosa. Am. Rev.
 14            Respir. Dis. 139:  1132-1138.
 15      Bennett, W. D.; Zeman,  K. L. (1998) Deposition of fine particles in children spontaneously breathing at rest.
 16            Inhalation Toxicol. 10: 831-842.
 17      Bennett, W. D.; Zeman, K. L.; Kim, C. (1996) Variability of fine particle deposition in healthy adults: effect of age
 18            and gender. Am. J. Respir. Crit. Care Med. 153: 1641-1647.
 19      Bennett, W. D.; Zeman, K. L.; Kang, C. W.; Schechter, M. S. (1997a) Extrathoracic deposition of inhaled, coarse
 20            particles (4.Sum) in children vs adults. Ann. Occup.  Hyg. 41(suppl.l): 497-502.
 21       Bennett, W. D.; Zeman, K. L.; Kim, C.; Mascarella, J. (1997b) Enhanced deposition of fine particles in COPD
 22             patients spontaneously breathing at rest. Inhalation Toxicol. 9:  1-14.
 23       Bennett, W. D.; Scheuch, G.; Zeman, K. L.; Brown, J. S.; Kim, C.; Heyder, J.; Stahlhofen, W. (1998) Bronchial
 24             airway deposition and retention of particles in inhaled boluses: effect of anatomic dead space. J. Appl.
 25             Physiol. 85: 685-694.
 26       Bennett, W. D.; Scheuch, G.; Zeman, K. L.; Brown, J. S.; Kim, C.; Heyder, J.; Stahlhofen, W. (1999) Regional
 27             deposition and retention of particles in shallow, inhaled boluses: effect if lung volume. J. Appl. Physiol.
 28             86: 168-173.
 29       Broday, D. M.; Georgopoulos, P. G. (2000) Growth and deposition of hygroscopic particulate matter in the human
 30             lungs. Aerosol Sci. Technol.: submitted.
 31       Camner, P.; Anderson, M.; Philipson, K.; Bailey, A.; Hashish, A.; Jarvis, N.; Bailey, M.; Svartengren, M. (1997)
 32             Human bronchiolar deposition and retention of 6-, 8-, and 10-u particles. Exp. Lung Res. 23: 517-535.
 33       Cheng, Y.-S.; Smith, S. M.; Yeh, H.-C.; Kim, D.-B.; Cheng, K.-H.; Swift, D. L. (1995) Deposition of ultrafine
 34             aerosols and thoron progeny in replicas of nasal airways of young children.  Aerosol. Sci. Technol.
 35             23:541-552.
 36       Cheng, K.-H.; Cheng, Y.-S.; Yeh, H.-C.; Guilmette, R. A.; Simpson, S. Q.; Yang, Y.-H.; Swift, D. L. (1996) In vivo
 37             measurements of nasal airway dimensions  and ultrafine aerosol deposition in the human nasal and oral
 38             airways. J. Aerosol Sci. 27: 785-801.
 39       Cheng, K.-H.; Cheng, Y.-S.; Yeh, H.-C.; Swift, D. L. (1997) An experimental method for measuring aerosol
 40            deposition efficiency in the human oral airway. Am. Ind. Hyg. Assoc. J. 58: 207-213.
 41       Churg, A.; Brauer, M. (1997) Human lung parenchyma retains PM25. Am. J. Respir. Crit. Care Med.
 42             155:2109-2111.
 43       Churg, A.; Vedal, S. (1996) Carinal and tubular airway particle concentrations in the large airways of non-smokers
 44            in the general population: evidence for high particle concentration at airway carinas. Occup. Environ. Med.
 45            53:553-558.
 46      Cohen, B. S.; Harley, N. H.; Schlesinger, R. B.; Lippmann, M. (1988) Nonuniform particle deposition on
 47            tracheobronchial airways: implications for  lung dosimetry. In: Dodgson, J.; McCallum, R. I.; Bailey, M. R.;
48            Fisher, D. R., eds. Inhaled particles VI: proceedings of an international symposium and workshop on lung
49            dosimetry; September 1985; Cambridge, United Kingdom. Ann. Occup. Hyg. 32 (suppl. 1): 1045-1053.
 50      Cohen, B. S.; Sussman, R. G.; Lippmann, M. (1990) Ultrafine particle deposition in a human tracheobronchial cast.
 51            Aerosol Sci. Technol. 12: 1082-1091.
52      Cohen, B. S.; Xiong, J. Q.; Fang, C.-P.; Li, W. (1998) Deposition of charged particles on lung airways. Health
53            Phys. 74: 554-560.
         March 2001                                    7-53         DRAFT-DO NOT QUOTE OR CITE

-------
 1       Comer, J.K., Kleinstreuer, C., Hyun, S. and Kim, C. S. (2000) Aerosol transport and deposition in sequentially
 2             bifurcating airways. J. Biomech. Eng. 122:152-158.
 3       Crystal, R. G., West, J. B.; Barnes, P. J.; Weibel, E. R., eds. (1997) The lung: scientific foundations. Volume 1.
 4             Section III: major components. 2nd ed. Philadelphia, PA: Lippincott-Raven; chapters 30-67, pp. 445-991.
 5       Cuddihy, R. G. (1984) Mathematical models for predicting clearance of inhaled radioactive materials. In: Smith, H.;
 6             Gerber, G., eds. Lung modelling for inhalation of radioactive materials: proceedings of a meeting jointly
 7             organized by the Commission of the European Communities and the National Radiological Protection Board;
 8             March; Oxford, United Kingdom. Luxembourg: Commission of the European Communities; pp.  167-179;
 9             report no. EUR 9384 EN.
10       Cuddihy, R. G.; Yeh, H. C. (1988) Respiratory tract clearance of particles and substances dissociated from particles.
11             In: Mohr, U.; Dungworth, D.; Kimmerle, G.; Lewkowski, J.; McClellan, R.; Stober, W., eds. Inhalation
12             toxicology: the design and interpretation of inhalation studies and their use in risk assessment. New York,
13             NY: Springer-Verlag; pp. 169-193.
14       Dorries, A. M.; Valberg, P. A. (1992) Heterogeneity of phagocytosis for inhaled versus instilled material. Am. Rev.
15             Respir. Dis. 146:831-837.
16       Driscoll, K. E.; Costa, D. L.; Hatch, G.; Henderson, R.; Oberdorster, G.; Salem, H.; Schlesinger, R. B. (2000)
17             Intratracheal instillation as an exposure technique for the evaluation  of respiratory tract toxicity: uses and
18             limitations. Toxicol. Sci. 55: 24-35.
19       Falk, R.; Philipson, K.; Svartengren, M.; Jarvis, N.; Bailey, M.; Camner, P. (1997) Clearance of particles from small
20             ciliated airways. Exp. Lung Res. 23: 495-515.
21       Falk, R.; Philipson, K.; Svartengren, M.; Bergmann, R.; Hofmann, W.; Jarvis, N.; Bailey, M.; Camner, P. (1999)
22             Assessment of long-term bronchiolar clearance of particles from measurements of lung retention and
23             theoretical estimates of regional deposition. Exp. Lung Res. 25: 495-516.
24       Ferin, J. (1977) Effect of particle content of lung on clearance pathways. In: Sanders, C. L.;  Schneider,  R. P.; Dagle,
25             G. E.; Ragan, H. A., eds. Pulmonary macrophages and epithelial cells: proceedings of the sixteenth annual
26             Hanford biology symposium; September 1976; Richland, WA. Oak Ridge, TN: Energy Research and
27             Development Administration; pp. 414-423. Available from: NTIS, Springfield, VA; CONF-760927.
28             (ERDA symposium series 43).
29       Ferin, J.; Feldstein, M. L. (1978) Pulmonary clearance and hilar lymph node content in rats after particle exposure.
30             Environ. Res. 16: 342-352.
31       Ferin, J.; Oberdorster, G.; Penney, D. P. (1992) Pulmonary retention of ultrafine and fine particles in rats. Am. J.
32             Respir. Cell Mol. Biol. 6: 535-542.
33       Findeisen, W. (1935) Ober das Absetzen kleiner, in der Luft suspendierter Teilchen in der menschlichen Lunge bei
34             der Atmung [The deposition of small airborne particles in the human lung during respiration]. Pfluegers
35             Arch. Gesamte Physiol. Menschen Tiere 236: 367-379.
36       Foster, W. M.; Langenback, E.; Bergofsky, E. H. (1980) Measurement of tracheal and bronchial mucus velocities in
37             man: relation to lung clearance. J. Appl. Physiol.: Respir. Environ. Exercise Physiol. 48: 965-971.
38       Frampton, M. W.; Chalupa, D.; Morrow, P. E.; Gibb, F. R.; Oberdorster, G.; Speers, D. M.;  Zareba, W.; Utell, M. J.
39             (2000)  Deposition and  effects of inhaled ultrafine carbon particles in healthy subjects at rest. Presented at:
40             PM2000: particulate matter and health—the scientific basis for regulatory decision-making, specialty
41             conference & exhibition; January; Charleston, SC. Pittsburgh, PA: Air & Waste Management Association.
42       Fry, F. A.; Black, A. (1973) Regional deposition and clearance of particles in the human nose. J. Aerosol Sci.
43             4:113-124.
44       Gehr, P.; Schiirch, S.; Berthaiume, Y.; Im Hof, V.; Geiser, M. (1990) Particle retention in airways by surfactant.
45             J. Aerosol Med. 3: 27-43.
46       Gehr, P.; Im Hof, V.; Geiser, M.; Schurch, S. (1991) The fate of particles deposited in the intrapulmonary
47             conducting airways. J. Aerosol Med. 4: 349-362.
48       Green, F. H. Y. (2000) Pulmonary responses to inhaled poorly soluble particulate in the human. In: Gardner, D. E.,
49             ed. ILSI Risk Science Institute Workshop: The Relevance of the Rat Lung  Response to Particle Overload for
50             Human Risk Assessment; March, 1998. Inhalation Toxicol. 12: 59-95.
51       Groth, M. L.;  Macri, K.; Foster, W. M. (1997) Cough and mucociliary transport of airway particulate in chronic
52             obstructive lung disease. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII: proceedings of an
53             international symposium on inhaled particles organised by the British Occupational Hygiene Society;
54             August 1996; Cambridge, UK. Ann. Occup. Hyg. 41(suppl.): 515-521.
         March 2001                                     7-54         DRAFT-DO NOT QUOTE OR CITE

-------
  1      Guilmette, R. A.; Cheng, Y. S.; Griffith, W. C. (1997) Characterising the variability in adult human nasal airway
  2            dimensions. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII: proceedings of an international
  3            symposium on inhaled particles organised by the British Occupational Hygiene Society; August 1996;
  4            Cambridge, UK. Ann. Occup. Hyg. 41(suppl. 1): 491-496.
  5      Harmsen, A. G.; Muggenburg, B. A.; Snipes,  M. B.; Bice, D. E. (1985) The role of macrophages in particle
  6            translocation from lungs to lymph nodes. Science (Washington, DC) 230: 1277-1280.
  7      Heistracher, T.; Hofmann, W. (1997) Flow and deposition patterns in successive airway bifurcations. In: Cherry, N.;
  8            Ogden, T., eds. Inhaled Particles VIII:  proceedings of an international symposium on inhaled particles
  9            organised by the British Occupational Hygiene Society; August 1996; Cambridge, UK. Ann. Occup. Hyg.
 10            41(suppl.): 537-542.
 11      Henshaw, D. L.; Fews, A. P. (1984) The microdistribution of alpha emitting particles in the human lung.
 12            In: Smith, H.; Gerber, G., eds. Lung modelling for inhalation of radioactive materials: proceedings of a
 13            meeting jointly organized by the Commission of the European Communities and the National Radiological
 14            Protection Board; March; Oxford, United Kingdom. Luxembourg: Commission of the European
 15            Communities; pp.  199-218; report no. EUR 9384 EN.
 16      Heyder, J.; Rudolf, G. (1977) Deposition of aerosol particles in the human nose. In: Walton, W. H.; McGovern, B.,
 17            eds. Inhaled particles IV: proceedings of an international symposium, part 1; September 1975; Edinburgh,
 18            United Kingdom. Oxford, United Kingdom: Pergamon Press, Ltd.; pp.  107-126.
 19      Killer, F. C. (1991) Health implications of hygroscopic particle growth in the human respiratory tract. J. Aerosol
20            Med. 4:  1-23.
21      Hofmann, W.;  Bergmann, R. (1998) Predictions of particle deposition patterns in human and rat airways. Inhalation
22            Toxicol. 10: 557-583.
23      Hofmann, W.;  Martonen, T. B.; Graham, R. C. (1989a) Predicted deposition of nonhygroscopic aerosols in the
24            human lung as a function of subject age. J. Aerosol Med. 2: 49-68.
25      Hofmann, W.;  Koblmger, L.; Martonen, T. B. (1989b) Structural differences between human and rat lungs:
26            implications for Monte Carlo modeling of aerosol deposition. In: Mahaffey, J. A., ed. 26th Hanford life
27            sciences symposium, modeling for scaling to man: biology, dosimetry, and response. Health Phys.
28            57(suppl. 1): 41-47.
29      Hofmann, W.;  Balashazy, I.; Heistracher, T.; Koblinger, L. (1996) The significance of particle deposition patterns in
30            bronchial airway bifurcations for extrapolation modeling. Aerosol Sci. Technol. 25: 305-327.
31      Hofmann, W.;  Bergmann, R.; Koblinger, L. (1999) Characterization of local particle deposition patterns in human
32            and rat lungs by different morphometric parameters. J. Aerosol Sci. 30: 651-667.
33      Hofmann, W.;  Asgharian, B.; Bergmann, R.; Anjilvel, S.; Miller, F. J. (2000) The  effect of heterogeneity of lung
34            structure on particle deposition in the rat lung. Toxicol. Sci. 53: 430-437.
35      Hsieh, T. H.; Yu, C. P. (1998) Two-phase  pulmonary clearance of insoluble particles in mammalian species.
36            Inhalation Toxicol. 10: 121-130.
37      International Commission on Radiological Protection. (1960) Report of Committee II on permissible dose for
38            internal  radiation (1959). Health Phys.  3.
39      International Commission on Radiological Protection. (1979) Limits for intakes of radionuclides by workers.
40            Oxford,  United Kingdom: Pergamon Press; ICRP publication 30, part 1.
41      International Commission on Radiological Protection. (1994) Human respiratory tract model for radiological
42            protection: a report of a task group of the International Commission on Radiological Protection. Oxford,
43            United Kingdom: Elsevier Science Ltd. (ICRP publication 66; Annals of the ICRP:  v. 24, nos. 1-3).
44      John, J.; Wollmer, P.; Dahlback, M.; Luts, A.; Jonson, B. (1994) Tidal volume and alveolar clearance of insoluble
45            particles. J. Appl. Physiol. 76: 584-588.
46      Jaques, P. A.; Kim, C. S.  (2000) Measurement of total lung deposition of inhaled ultrafine particles in healthy men
47            and women. Inhalation Toxicol. 12: 715-731.
48      Katz, I. M.; Davis, B. M.; Martonen, T. B. (1999) A numerical study of particle motion within the human larynx
49            and trachea. J. Aerosol Sci. 30: 173-183.
50      Kesavanathan,  J.; Swift, D. L. (1998) Human  nasal passage particle deposition: the effect of particle size, flow rate,
51            and anatomical factors. Aerosol Sci. Technol. 28: 457-463.
52      Kim C. S. (1989) Aerosol deposition in the lung with obstructed airways. J. Aerosol Med. 2:111-120.
53      Kim, C.  S. (2000) Methods of calculating lung delivery and deposition of aerosol particles. Respir. Care
54            45:695-711.
         March 2001                                     7-55          DRAFT-DO NOT QUOTE OR CITE

-------
  1       Kim C. S. and Fisher, D.M. (1999) Deposition characteristics of aerosol particles in sequentially bifurcating airway
  2             models. Aerosol Sci. Technol. 31:198-220.
  3       Kim, C. S.; Hu, 5. C. (1998) Regional deposition of inhaled particles in human lungs: comparison between men and
  4             women. J. Appl. Physiol. 84:  1834-1844.
  5       Kim, C. S. and Jaques, P. A. (2000)  Respiratory dose of inhaled ultrafine particles in healthy adults. Phil. Trans.
  6             Roy. Soc. London A 358:2693-2705.
  7       Kim, C. S.; Kang, T. C. (1997) Comparative measurement of lung deposition of inhaled fine particles in normal
  8             subjects and patients with obstructive airway disease. Am. J. Respir. Crit.  Care Med. 155: 899-905.
  9       Kim, C. S., Iglesias, A. J., and Sackner, M.A. (1987) Mucus clearance by two-phase gas-liquid flow mechanism:
 10             Asymmetric periodic flow model. J. Appl. Physiol. 62:959-971.
 11       Kim C. S., Eldridge, M. A., Garcia, L. and Wanner A. (1989) Aerosol deposition in the lung with asymmetric
 12             airways obstruction: in vivo observation. J. Appl. Physiol. 67:2579-2585.
 13       Kim, C. S.; Hu, S. C.; DeWitt, P.; Gerrity, T. R. (1996)  Assessment of regional deposition of inhaled particles in
 14             human lungs by serial bolus delivery method. J. Appl. Physiol. 81: 2203-2213.
 15       Kim, C. S.; Hu, S. C.; DeWitt, P. (2000) Variation of lung deposition dose of inhaled particles with breathing
 16             pattern at rest and during moderate  exercise. Presented at: PM2000: particulate matter and health—the
 17             scientific basis for regulatory decision-making, specialty conference & exhibition; January; Charleston, SC.
 18             Pittsburgh, PA: Air & Waste Management Association.
 19       Koblinger, L.; Hofmann, W. (1985) Analysis of human  lung morphometric data for stochastic aerosol deposition
 20             calculations. Phys. Med. Biol. 30: 541-556.
 21       Koblinger, L.; Hofmann, W. (1988) Stochastic morphological model of the rat lung. Anat. Rec. 221: 533-539.
 22       Kohlhaufl, M.; Brand, P.; Scheuch, G.; Meyer, T. S.; Schulz, H.; Haussinger, K.; Heyder, J. (1999) Increased fine
 23             particle deposition in women with asymptomatic  nonspecific airway hyperresponsiveness. Am. J. Respir.
 24             Crit. Care Med. 159: 902-906.
 25       Kreyling, W. G. (1992) Intracellular particle dissolution in alveolar macrophages. Environ. Health Perspect.
 26             97:121-126.
 27       Kreyling, W. G.; Blanchard, J. D.; Godleski, J. J.; Haeussermann, S.; Heyder, J.; Hutzler, P.; Schulz, H.; Sweeney,
 28             T. D.; Takenaka, S.; Ziesenis, A. (1999) Anatomic localization of 24- and 96-h particle retention in canine
 29             airways. J. Appl. Physiol. 87: 269-284.
 30       LaBelle, C. W.; Brieger, H. (1961) Patterns and mechanisms in the elimination of dust from the lung. In: Davies,
 31             C. N., ed. Inhaled particles and vapours: proceedings of an international symposium; March-April 1960;
 32             Oxford, United Kingdom. New York, NY: Pergamon Press; pp. 356-368.
 33       Landahl, H. D. (1950) On the removal of air-borne droplets by the human respiratory tract: I. the lung. Bull. Math.
 34             Biophys. 12: 43-56.
 35       Lay, J. C.; Berry, C. R.; Chong, S. K.; Bennett, W. D. (1995) Retention of insoluble particles after local
 36             intrabronchial deposition in dogs. J. Appl. Physiol. 79: 1921-1929.
 37       Lay, J. C.; Bennett, W. D.; Kim, C. S.; Devlin, R. B.; Bromberg, P. A. (1998) Retention and intracellular
 38             distribution of instilled iron oxide particles in human alveolar macrophages.  Am. J. Respir. Cell Mol. Biol.
 39             18:687-695.
 40       Lehnert, B. E.; Morrow, P. E. (1985) Association of 59iron oxide with alveolar macrophages during alveolar
 41             clearance. Exp. Lung Res. 9: 1-16.
 42       Lehnert, B. E.; Valdez, Y. E.; Bomalaski, S. H. (1988) Analyses of particles in the lung free cell, tracheobronchial
 43             lymph nodal, and pleural space compartments following their deposition in the lung as related to lung
44             clearance mechanisms. In: Dodgson, J.; McCallum, R. I.; Bailey, M. R.; Fisher, D. R., eds. Inhaled particles
45             VI: proceedings of an international  symposium and workshop on lung dosimetry; September 1985;
46             Cambridge, United Kingdom. Ann.  Occup. Hyg.  32 (suppl. 1): 125-140.
47       Leikauf, G.; Yeates, D. B.; Wales, K. A.; Spektor, D.; Albert, R. E.; Lippmann, M.  (1981) Effects of sulfuric acid
 48             aerosol on respiratory mechanics and mucociliary particle clearance in healthy nonsmoking adults. Am. Ind.
49             Hyg. Assoc. J. 42: 273-282.
 50       Leikauf, G. D.; Spektor, D. M.; Albert, R.  E.; Lippmann, M. (1984) Dose-dependent effects of submicrometer
 51             sulfuric acid aerosol on particle clearance from ciliated human lung airways. Am. Ind. Hyg. Assoc. J.
 52             45:285-292.
 53       Lennon, S.; Zhang, Z.; Lessmann, R.; Webster, S. (1998) Experiments on particle deposition in the human upper
 54             respiratory system. Aerosol Sci. Technol. 28: 464-474.
         March 2001                                     7-56          DRAFT-DO NOT QUOTE OR CITE

-------
  1       Lundborg, M.; Lind, B.; Camner, P. (1984) Ability of rabbit alveolar macrophages to dissolve metals. Exp. Lung
  2             Res. 7:11-22.
  3       Lundborg, M.; Eklund, A.; Lind, B.; Camner, P. (1985) Dissolution of metals by human and rabbit alveolar
  4             macrophages. Br. J. Ind. Med. 42: 642-645.
  5       Madl, A. K.; Wilson, D. W.; Segall, H. J.;  Pinkerton, K. E. (1998) Alteration in lung particle translocation,
  6             macrophage function, and microfilament arrangement in monocrotaline-treated rats. Toxicol. Appl.
  7             Pharmacol. 153:28-38.
  8       Martonen, T. B.; Schroeter, J. D.; Hwang,  D.; Fleming, J.  S.; Conway, J. H. (2000) Human lung morphology
  9             models for particle deposition studies. In: Grant, L. D., ed. PM2000: particulate matter and health.
 10             Inhalation Toxicol. 12(suppl. 4): 109-121.
 11       Matsui, H.; Randell, S. H.; Peretti, S. W.; Davis, C. W.; Boucher, R. C. (1998) Coordinated clearance of periciliary
 12             liquid and mucus from airway surfaces. J. Clin. Invest. 102: 1125-1131.
 13       Medinsky, M. A.; Kampcik, S. J. (1985) Pulmonary retention of [14C]benzo/a7pyrene in rats as influenced by the
 14             amount instilled. Toxicology 35: 327-336.
 15       Mercer, T. T. (1967) On the role of particle size in the dissolution of lung burdens. Health Phys. 13: 1211-1221.
 16       Miller, F.  J.; Anjilvel, S.; Menache, M. G.; Asgharian, B.; Gerrity, T. R.. (1995) Dosimetric issues relating to
 17             particulate toxicity. Inhalation Toxicol. 7: 615-632.
 18       Morrow, P. E. (1973) Alveolar clearance of aerosols. Arch. Intern. Med.  131: 101-108.
 19       Morrow, P. E. (1977) Clearance kinetics of inhaled particles. In: Brain, J. D.; Proctor, D. F.; Reid, L. M., eds.
 20             Respiratory defense mechanisms (in two parts), part II. New York, NY: Marcel Dekker, Inc.; pp. 491-543.
 21             (Lenfant, C.,  ed. Lung biology in health and disease: v. 5).
 22       Morrow, P. E. (1986) Factors determining  hygroscopic aerosol deposition in airways. Physiol. Rev. 66: 330-376.
 23       Morrow, P. E. (1988) Possible mechanisms to explain dust overloading of the lungs. Fundam. Appl. Toxicol.
 24             10:369-384.
 25       Morrow, P. E. (1994) Mechanisms and significance of "particle overload." In: Mohr, U.; Dungworth, D. L.;
 26             Mauderly, J. L.; Oberdorster, G., eds. Toxic and carcinogenic effects of solid  particles in the respiratory tract:
 27             [proceedings  of the 4th international inhalation symposium]; March 1993; Hannover, Germany. Washington,
 28             DC: International Life Sciences Institute Press; pp.  17-25.
 29       Morrow, P. E.; Yu, C. P. (1993) Models of aerosol behavior in airways and alveoli. In: Moren, F.; Dolovich, M. B.;
 30             Newhouse, M. T.; Newman, S. P., eds. Aerosols in medicine: principles, diagnosis and therapy. 2nd rev. ed.
 31             Amsterdam, The Netherlands: Elsevier; pp.  157-193.
 32       Morrow, P. E.; Gibb, F. R.; Gazioglu, K. M. (1967) A study of particulate clearance  from the human lungs.
 33             Am. Rev. Respir. Dis. 96: 1209-1221.
 34       Musante, C. J.; Martonen, T. B. (1999) Predicted deposition patterns of ambient particulate air pollutants in
 35             children's lungs under resting conditions. In: Proceedings of the third colloquium on particulate air pollution
 36             and human health; June; Durham, NC. Irvine, CA: University of California, Air Pollution Health Effects
 3 7             Laboratory, p. 7-15 - 7-20.
 38       Musante, C. J.; Martonen, T. B. (2000a) Computer simulations of particle deposition in  the developing human lung.
 39             J. Air Waste Manage. Assoc. 50: 1426-1432.
 40       Musante, C.  J.; Martonen, T. B. (2000b) An extrapolation model to aid in toxicological studies of particulate air
 41             pollutants. Presented at: PM2000: particulate matter and health—the scientific basis for regulatory
 42             decision-making, specialty conference & exhibition; January; Charleston, SC. Pittsburgh, PA: Air & Waste
 43             Management Association.
 44      Musante, C. J.; Martonen, T. B. (2000c) Particulate matter deposition in the lungs of children and adults:
 45            predictions of an age-dependent computer model. Presented at: PM2000: particulate matter and health—the
 46            scientific basis for regulatory decision-making, specialty conference & exhibition; January; Charleston, SC.
47            Pittsburgh, PA: Air & Waste Management Association.
48      National Council on Radiation Protection and Measurements. (1997) Deposition, retention and dosimetry of inhaled
49            radioactive substances. Bethesda, MD: National Council on Radiation Protection and Measurements; report
 50            no.  125.
 51       Naumann,  B. D.; Schlesinger, R. B. (1986) Assessment of early alveolar particle clearance and macrophage
 52            function following an acute inhalation of sulfuric acid mist. Exp. Lung Res. 11: 13-33.
 53      Nikula, K. J.; Avila,  K. J.; Griffith, W. C.; Mauderly, J. L.  (1997) Lung tissue responses and sites of particle
54            retention differ between rats and cynomolgus monkeys exposed chronically to diesel exhaust and coal dust.
55             Fundam. Appl. Toxicol. 37: 37-53.


         March 2001                                     7-57         DRAFT-DO NOT QUOTE OR CITE

-------
  1       Nikula, K. J.; Vallyathan, V.; Green, F. H. Y.; Hahn, F. F. (2000) Influence of dose on the distribution of retained
  2             particulate material in rat and human lungs. Presented at: PM2000: particulate matter and health—the
  3             scientific basis for regulatory decision-making, specialty conference & exhibition; January; Charleston, SC.
  4             Pittsburgh, PA: Air & Waste Management Association.
  5       Noone, P. G.; Bennett, W. D.; Regnis, J. A.; Zeman, K. L; Carson, J. L.; King, M.; Boucher, R. C.; Knowles, M. R.
  6             (1999) Effect of aerosolized uridine-5'-triphosphate on airway clearance with cough in patients with primary
  7             ciliary dyskinesia. Am.  J. Respir. Crit. Care Med.: 160: 144-149.
  8       Oberdorster, G. (1993) Lung dosimerry: pulmonary  clearance of inhaled particles. Aerosol Sci. Technol.
  9              18:279-289.
10       Oberdorster, G.; Ferin, J.; Gelein, R.; Soderholm, S. C.; Finkelstein, J. (1992) Role of the alveolar macrophage in
11             lung injury: studies with ultrafine particles. Environ. Health Perspect. 97: 193-199.
12       Oberdorster, G.; Cox, C.; Gelein, R. (1997) Intratracheal instillation versus intratracheal inhalation of tracer
13             particles for measuring  lung clearance function. Exp. Lung Res. 23: 17-34.
14       Oldham, M. J.; Mannix, R. C.; Phalen, R. F. (1997) Deposition of monodisperse particles in hollow models
15             representing adult and child-size tracheobronchial airways. Health Phys. 72: 827-834.
16       Passali, D.; Bianchini Ciampoli, M. (1985) Normal  values of mucociliary transport time in young subjects. Int. J.
17             Pediatr. Otorhinolaryngol. 9: 151-156.
18       Patton, J. S. (1996) Mechanisms of macromolecule  absorption by the lungs. Adv. Drug  Delivery Rev. 19: 3-36.
19       Pritchard, J. N.; Jefferies, S. J.; Black, A. (1986) Sex differences in the regional deposition of inhaled particles in
20             the 2.5—7.5 urn size range. J. Aerosol Sci. 17:  385-389.
21       Radford, E. P.; Martell, E. A. (1977) Polonium-210: lead-210 ratios as an index of residence time of insoluble
22             particles from cigarette smoke in bronchial epithelium. In: Walton, W. H.; McGovern,  B., eds. Inhaled
23             particles IV: proceedings of an international symposium, part 2; September 1975; Edinburgh, United
24             Kingdom. Oxford, United Kingdom: Pergamon Press, Ltd.; pp. 567-581.
25       Roy, M. (3989) Lung clearance modeling on the basis of physiological and biological parameters. Health Phys.
26             57(suppl. 1): 255-262.
27       Rutland, J.; Cole, P.  J. (1981) Nasal mucociliary clearance and ciliary beat frequency in cystic fibrosis compared
28             with sinusitis  and bronchiectasis. Thorax 36:  654-658.
29       Scheuch, G.; Stahlhofen, W. (1988) Particle deposition of inhaled aerosol boluses in the upper human airways.
30             J. Aerosol Med. 1:29-36.
31       Scheuch, G.; Philipson, K.; Falk, R.; Svartengren, M.; Stahlhofen, W.; Camner, P. (1995) Retention of particles
32             inhaled in boli with and without induced bronchoconstriction. Exp. Lung Res. 21: 901-916.
33       Schlesinger, R. B. (1985) Clearance from the respiratory tract. Fundam. Appl. Toxicol. 5:  435-450.
34       Schlesinger, R. B. (1988) Biological disposition of airborne particles: basic principles and application to vehicular
35             emissions. In: Watson, A. Y.; Bates, R. R.; Kennedy, D., eds. Air pollution, the automobile, and public
36             health. Washington, DC: National Academy Press; pp. 239-298.
37       Schlesinger, R. B. (1995) Deposition and clearance  of inhaled particles. In: McClellan, R. O.; Henderson, R. F.,
38             eds. Concepts in inhalation toxicology. 2nd ed. Washington,  DC: Taylor & Francis; pp. 191-224.
39       Schlesinger, R. B.; Ben-Jebria, A.; Dahl, A. R.; Snipes, M. B.; Ultman, J. (1997) Disposition of inhaled toxicants.
40             In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press; pp. 493-550.
41       Segal, R. A.; Martonen, T. B.; Shearer,  M.; (2000a) Particle trajectories in the cancerous lungs of children.
42             Presented at: PM2000: particulate matter and health—the scientific basis for regulatory decision-making,
43             specialty conference &  exhibition;  January; Charleston, SC. Pittsburgh,  PA: Air & Waste Management
44             Association.
45       Segal, R. A.; Martonen, T. B.; Kim, C. S.  (2000b) Comparison of computer simulations to total lung deposition to
46             human subject data in healthy test subjects. J. Air Waste Manage. Assoc. 50: 1262-1268.
47       Smaldone, G. C.; Perry, R. J.; Bennett, W. D.; Messina, M. S.; Zwang, J.; Ilowite, J. (1988) Interpretation of
48             "24 hour lung retention" in studies  of mucociliary clearance.  J. Aerosol Med. 1:11 -20.
49       Snipes, M. B. (1989) Long-term retention and clearance of particles inhaled by mammalian species. CRC Crit. Rev.
50             Toxicol. 20: 175-211.
51       Snipes, M. B.; Clem, M.  F. (1981) Retention of microspheres in the rat lung after intratracheal instillation.
52             Environ. Res. 24:33-41.
53       Snipes, M. B.; McClellan, R. O.; Mauderly, J. L.; Wolff, R. K. (1989) Retention patterns for inhaled particles in the
54             lung: comparisons between laboratory  animals and humans for chronic exposures. Health Phys.
55             57(suppl. 1): 69-78.

         March 2001                                    7-58          DRAFT-DO NOT QUOTE OR CITE

-------
  1       Stahlhofen, W.; Gebhart, J.; Rudolf, G.; Scheuch, G.; Philipson, K. (1986a) Clearance from the human airways of
  2            particles of different sizes deposited from inhaled aerosol boli. In: Aerosols: formation and reactivity,
  3            proceedings of the second international aerosol conference; September; Berlin, Federal Republic of
  4            Germany. Oxford, United Kingdom: Pergamon Press; pp. 192-196.
  5       Stahlhofen, W.; Gebhart, J.; Rudolf, G.; Scheuch, G. (1986b) Measurement of lung clearance with pulses of
  6            radioactively-labelled aerosols. J. Aerosol Sci. 17: 333-336.
  7       Stanley, P. J.; Wilson, R.; Greenstone, M. A.; Mackay, I. S.; Cole, P. J. (1985) Abnormal nasal mucociliary
  8            clearance in patients with rhinitis and its relationship to concomitant chest disease. Br. J. Dis. Chest
  9            79:77-82.
 10       Stober, W.; Morrow, P. E.; Koch, W.; Morawietz, G. (1994) Alveolar clearance and retention of inhaled insoluble
 11            particles in rats simulated by a model inferring macrophage particle load distributions. J. Aerosol Sci.
 12            25:975-1002.
 13       Strom, K. A.; Chan, T. L.; Johnson, J. T. (1988) Pulmonary retention of inhaled submicron particles in rats: diesel
 14            exhaust exposures and lung retention model. In: Dodgson, J.; McCallum, R. I.; Bailey, M. R.; Fischer, D. R.,
 15            eds. Inhaled particles VI: proceedings of an international symposium and workshop on lung dosimetry;
 16            September 1985; Cambridge, United Kingdom. Ann. Occup. Hyg. 32(suppl. 1): 645-657.
 17       Svartengren, K.; Lindestad, P.;  Svartengren, M.; Philipson, K.; Bylin, G.; Camner, P. (1995) Added external
 18            resistance reduces oropharyngeal deposition and increases lung deposition of aerosol particles in asthmatics.
 19            Am. J. Respir. Crit. Care Med.  152: 32-37.
 20       Svartengren, K.; Philipson, K.;  Svartengren, M.; Anderson, M.; Camner, P. (1996a) Tracheobronchial deposition
 21            and clearance in small airways in asthmatic subjects. Eur. Respir. J. 9: 1123-1129.
 22       Svartengren, K.; Ericsson, C. H.; Svartengren, M.; Mossberg, B.; Philipson, K.;  Camner, P. (1996b)  Deposition and
 23            clearance in large and small airways in chronic bronchitis.  Exp. Lung Res. 22: 555-576.
 24       Svartengren, K.; Philipson, K.;  Svartengren, M.; Camner, P. (1998) Effect of adrenergic stimulation  on clearance
 25            from small ciliated airways in healthy subjects. Exp. Lung  Res. 24:  149-158.
 26       Svartengren, M.; Svartengren, K.; Aghaie, F.; Philipson, K.; Camner, P. (1999) Lung deposition and extremely slow
 27            inhalations of particles. Limited effect of induced airway obstruction. Exp. Lung Res. 25: 353-366.
 28       Swift, D. L.; Strong, J. C. (1996) Nasal deposition of ultrafine 2l8Po aerosols in human subjects. J. Aerosol Sci.
 29            27:1125-1132.
 30       Takahashi, S.; Asaho, S.; Kubota, Y.; Sato, H.; Matsuoka, O. (1987) Distribution of l98Au and 133Ba m thoracic and
 31            cervical lymph nodes of the rat  following the intratracheal instillation of mAu-colloid and 1MBaSO4.
 32            J. Radiat. Res. 28: 227-231.
 33       Takahashi, S.; Kubota, Y.; Hatsuno, H. (1992) Effect of size on the movement of latex particles in the respiratory
 34            tract following local administration. Inhalation Toxicol. 4:  113-123.
 35       Toms, N.; Hasani, A.; Pavia, D.; Clarke, S. W.; Agnew, J. E. (1997) Effect of mucus hypersecretion  on initial time
 36            course of inert particle clearance from the lung. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII:
 37            proceedings of an international  symposium on inhaled particles organised by the British Occupational
 38            Hygiene Society; August 1996; Cambridge, UK. Ann. Occup. Hyg. 41(suppl.): 509-514.
 39       U.S. Environmental Protection Agency. (1996) Air quality criteria for particulate matter. Research Triangle Park,
 40            NC: National Center for  Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
 41       Venkataraman, C.; Kao, A. S. (1999) Comparison of particle lung doses from the fine and coarse fractions of urban
 42            PM-10 aerosols. Inhalation Toxicol. 11:  151-169.
 43       Wolff, R. K.; Henderson, R. F.; Snipes, M. B.; Griffith, W. C.;  Mauderly, J. L.; Cuddihy, R. G.; McClellan, R. O.
 44            (1987) Alterations in particle accumulation and clearance in lungs of rats chronically exposed to diesel
45            exhaust. Fundam. Appl. Toxicol. 9: 154-166.
46       Xu, G. B.; Yu, C. P. (1986) Effects of age on deposition of inhaled aerosols in the human lung. Aerosol Sci.
47            Technol. 5: 349-357.
48      Yeates, D. B.; Aspin, M. (1978) A mathematical description of the airways of the human lungs, Respir. Physiol.
49            32:91-104.
 50      Yeates, D. B.; Aspin, N.; Levison, H.;  Jones, M. T.; Bryan, A. C.  (1975) Mucociliary tracheal transport rates in
 51             man. J. Appl. Physiol. 19: 487-495.
 52      Yeates, D. B.; Pitt, B. R.; Spektor, D. M.; Karron, G. A.; Albert, R. E. (1981) Coordination of mucociliary transport
 53            in human trachea and intrapulmonary airways.  J. Appl. Physiol.: Respir. Environ. Exercise Physiol.
 54            51:1057-1064.
         March 2001                                     7-59         DRAFT-DO NOT QUOTE OR CITE

-------
1      Yeh, H.-C.; Schum, G. M. (1980) Models of human lung airways and their application to inhaled particle
2            deposition. Bull. Math. Biol. 42: 461-480.
3      Yu, G.; Zhang, Z.; Lessmann, R. (1998) Fluid flow and particle diffusion in the human upper respiratory system.
4            Aerosol Sci. Technol. 28: 146-158.
5      Zhang, Z.; Martonen, T. (1997) Deposition of ultrafine aerosols in human tracheobronchial airways. Inhalation
6            Toxicol.9:99-110.
        March 2001                                   7-60         DRAFT-DO NOT QUOTE OR CITE

-------
  i             8.  TOXICOLOGY OF PARTICULATE MATTER
  2
  3
  4      8.1  INTRODUCTION
  5           Toxicological research on ambient particulate matter (PM) is used to address several
  6      related questions, including (1) what causal mechanisms may be involved in the toxicological
  7      response to PM exposures, (2) what factors affect individual or subpopulation susceptibility to
  8      the effects of PM exposures, (3) what characteristics of PM (e.g., size, composition) are
  9      producing observed toxicity, and (4) what are the combined effects of PM and gaseous
 10      co-pollutants in producing toxic responses? A variety of research approaches are used to address
 11      these questions, including in vivo studies of human volunteers to controlled exposures; in vivo
 12      studies of animals such as nonhuman primates, dogs and rodent species; and in vitro studies of
 13      tissue, cellular, genetic, and biochemical systems. Similarly, a variety of exposure conditions are
 14      employed, including whole body and nose-only inhalation exposures to artificially generated PM
 15      or concentrated ambient air, pulmonary instillation, and in vitro exposure to test materials in
 16      solution.  The various research designs are targeted to test hypotheses and, ultimately, provide a
 17      scientific basis for an improved understanding of the role of PM in producing health effects
 18      identified by epidemiological studies.
 19           Because of the sparsity of toxicological data on ambient PM at the time the previous PM
 20      Air Quality Criteria Document or "PM AQCD" (U.S. Environmental Protection Agency, 1996a)
 21      was completed, the discussion of respiratory effects of PM were organized into specific chemical
 22      components of ambient PM or model "surrogate" particles (e.g., acid aerosols, metals, ultrafine
 23      particles, bioaerosols, "other particle matter"). In this chapter, the conclusions of the 1996 PM
 24      AQCD are summarized for each of these components. Since completion of the previous
 25      document, there are many new studies demonstrating the potentially adverse effects of
 26      combustion-related particles. The main reason for this increased interest in combustion particles
 27      is that these particles are typically the dominant sources represented in the fine fraction of
 28      ambient PM.
29          This chapter is organized as follows. The respiratory effects of specific components of
30      ambient PM or surrogate particles delivered by in vivo exposures of both humans and laboratory

        March 2001                               8-1        DRAFT-DO NOT QUOTE OR CITE

-------
 1     animals are discussed first (Section 8.2), followed by systemic and cardiovascular effects of
 2     particles (Section 8.3) and effects in laboratory animal models that mimic human disease
 3     (Section 8.4). The in vitro exposure studies are discussed next (Section 8.5) because they
 4     provide valuable information on potential hazardous constituents and mechanisms of PM injury.
 5     The remaining section on exposure studies examines the health effects of mixtures of ambient
 6     PM or PM surrogates with gaseous pollutants (Section 8.6). This organization provides the
 7     underlying data for evaluation in the final section of this chapter (Section 8.7), but it may fail to
 8     adequately convey the extensive and intricate linkages among the pulmonary, cardiac, and
 9     nervous systems, all of which may be involved individually and in concert to represent the effects
10     of exposure to PM.
11
12
13     8.2  RESPIRATORY EFFECTS OF PARTICULATE MATTER IN
14          HEALTHY  HUMANS AND LABORATORY ANIMALS: IN VIVO
15          EXPOSURES
16          The following sections assess the results of human exposure to various types of PM and
17     also discuss controlled animal toxicology studies, as well as in vitro studies using  animal or
18     human respiratory cells. The discussion focuses on those studies published since the previous
19     1996 PM AQCD (U.S. Environmental Protection Agency,  1996a).
20          The biological responses occurring in the respiratory tract following controlled PM
21     inhalation encompass a continuum of changes, including changes in pulmonary inflammation,
22     pulmonary function,  and systemic effects.  The observed responses are dependent  on the
23     physicochemical characteristics of the PM, the total exposure, and the health status of the host.
24     Many of the responses are usually seen only at higher level exposures characteristic of
25     occupational and laboratory animal studies and not at (typically much lower) ambient particle
26     concentrations; however, there are substantial differences in the inhalability  and deposition
27     profiles of PM in humans and rodents (see Chapter 7 for details). Observed  responses and
28     dose-response relationships also are very dependant on the variable being measured.
29          Paniculate matter is a broad term that encompasses thousands of chemical species, many
30     of which have not been investigated in controlled laboratory animal or human studies.  However,
31     a full discussion of all types of particles that have been studied is beyond the scope of this

       March 2001                               8-2        DRAFT-DO NOT QUOTE OR CITE

-------
  1      chapter.  Thus, specific criteria were used to select topics for presentation.  High priority was
  2      placed on studies that may (1) elucidate health effects of major common constituents of ambient
  3      PM or (2) contribute to enhanced understanding of the epidemiological studies (e.g., use of
  4      ambient particles, "surrogate" particles, or particles with low inherent toxicity that may cause
  5      effects because of their physicochemical characteristics, such as their size and composition).
  6      Most studies, therefore, have been designed to address the question of biologic plausibility,
  7      rather than providing dose-response or risk assessment quantitation.
  8           Diesel exhaust particles (DPM) generally fit the criteria; but, because they are described
  9      elsewhere in great detail (U. S. Environmental Protection Agency, 1999; Health Effects  Institute,
 10      1995), they are not covered extensively in this chapter except in the discussions of their
 11      immunological effects. Particles with high inherent toxicity, such as silica and asbestos, that are
 12      of concern primarily because of occupational exposure, also are excluded from this chapter and
 13      are discussed in detail elsewhere  (U.S. Environmental Protection Agency, 1996b; Gift and Faust,
 14      1997). Most of the laboratory animal studies summarized here have used high particulate mass
 15      concentrations administered by inhalation, compared to ambient levels,  even when laboratory
 16      animal-to-human dosimetric differences or high doses by intratracheal instillation are considered.
 17      More research on particle dose extrapolation is needed, therefore, to determine species
 18      differences and the importance of exercise and other factors influencing particle deposition in
 19     humans that together can account for a 50-fold or more difference in dose.
 20          As mentioned earlier, the data available in the previous 1996 PM AQCD were from studies
 21      that investigated  the respiratory effects of specific components of ambient PM or surrogate
 22     particles. More recently, pulmonary effects of controlled exposures to ambient PM have been
 23      investigated by the use of aerosol concentrators (Sioutas  et al., 1995; Gordon et al., 1998).  These
 24      concentrators are capable of exposing animals or humans to PM concentrations that are up to
 25      90-fold higher than ambient PM levels and have been used to investigate the effects of ambient
 26      PM in normal and compromised animals  and humans.
 27
 28      8.2.1 Acid Aerosols
29           There have  been extensive studies of the effects of  controlled exposures to aqueous acid
30      aerosols on various aspects of lung function in humans and laboratory animals. Many of these
31      studies were reviewed in the previous  criteria  document (U.S. Environmental Protection Agency
        March 2001                               8-3         DRAFT-DO NOT QUOTE OR CITE

-------
 1      1996a) and in the Acid Aerosol Issue Paper (U.S. Environmental Protection Agency, 1989); more
 2      recent studies are summarized in the present document (see Table 8-1).  Methodology and
 3      measurement methods for controlled human exposure studies have been reviewed elsewhere
 4      (Folinsbeeetal., 1997).
 5           These studies illustrate that aqueous acidic aerosols have minimal effects on symptoms and
 6      mechanical lung function in young healthy adult volunteers at concentrations as high as
 7      2000 Aig/m3.  The findings include minimal changes in lung function accompanied by only mild
 8      lower respiratory symptoms. However at concentrations as low as 100 /wg/m3, acid aerosols can
 9      alter mucociliary clearance.  Brief exposures (< 1 h) to low concentrations (=100 /wg/m3) may
10      accelerate clearance while longer (multihour) exposures to higher concentrations  (>100 /ug/rn3)
11      can depress clearance. Asthmatic subjects appear to be more sensitive to the effects of acidic
12      aerosols on mechanical lung function. Responses have been reported in adolescent asthmatics at
13      concentrations as low as 68 /wg/m3 and modest bronchoconstriction has been seen in adult
14      asthmatics exposed to concentrations >400 //g/m3, but the available data are not consistent.
15           A previously described acid aerosol exposure in humans (1000 Mg/m3) did not result in
16      airway inflammation (Frampton et al., 1992), and there was no evidence of altered macrophage
17      host defenses. More recently, Zelikoff et al. (1997) compared the responses of rabbits and
18      humans exposed to similar concentrations of acid aerosol. For both rabbits and humans, there
19      was no evidence of PMN infiltration into the lung and no change in BAL protein  level, although
20      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. No respiratory effects of long-term
24      exposure to acid aerosol were found in dogs (Heyder et al.,  1999).
25
26      8.2.2 Metal Particles, Fumes, and  Smoke
27           Data from occupational and laboratory animal studies reviewed in the previous criteria
28      document (U. S. Environmental Protection Agency, 1996a) indicated that acute exposures to very
29      high levels (hundreds of /ug/m3 or more) or chronic exposures to lower levels (up to 15 /ug/m3,
30      albeit high compared to ambient levels) of metallic particles could have an effect on the
31      respiratory tract. However, it was concluded on the basis of available data that the metals at
        March 2001                               8-4         DRAFT-DO NOT QUOTE OR CITE

-------
§
tr
O
O
               TABLE 8-1.  RESPIRATORY EFFECTS OF ACID AEROSOLS IN HUMANS AND LABORATORY ANIMALS
         Species, Gender, Strain
              Age, etc.
Particle
Exposure
Technique
                           Concentration
                                           Particle Size
                                                         Exposure
                                                         Duration
                                                                               Effects of Particles
                                                                                                                                           Reference
oo
T1
H
6
o
o
H
O
c
o
         Healthy beagle dogs;     Neutral sulfite       Inhalation      1.5 mg/m3
         n = 16                aerosol
                                          1.0/^mMMAD   165h/day    Long-term exposure to particle-associated sulfur   Heyderet al (1999)
                                          6g=22        for 13 mo     and hydrogen ions at concentrations close to
                                                                    ambient levels caused only subtle respiratory
                                                                    responses and no change in lung pathology.

Asthmatic subjects;
13 M, 11 F
Female Fischer 344 rats
Female Hartley Guinea
Pigs
Healthy nonsmokers;
10M, 2 F, 2 1-37 years
old
Acidic sulfate
aerosol
H,SO4 aerosol
NHVSO'5
aerosol
H2SO4 aerosol
H,SO4 aerosol
Inhalation 5 7 mg/m3
Inhalation by 500 ^g/m3
face mask
Inhalation 94 mg/m3
43 mg/m3
Inhalation 1 ,000 f^g/m3
1.1 jumMMAD
6g = 2.0
9 nm MMAD
7 urn MMAD
0.80 ± 1 89 6g
0.93 ±2 11 8g
0.8-0.9 nm
MMAD
6 h/day for
13 mo
1 h
4h
3h

Exposure to simulated natural acid fog did not
induce bronchoconstnction and did not change
bronchial responsiveness in asthmatics
Acid aerosol increased surfactant film
compressibility in guinea pigs
No inflammatory responses, LDH activity in BAL
was elevated Effect on bacterial killing by
macrophages was inconclusive; latex particle
phagocytosis was reduced 28%

Leducetai (1995)
Leeetal (1999)
Zehkoffetal. (1997)
         BAL - Bronchoalveolar lavage
         LDH - Lactate dehydrogenase
         MMAD - Mass median aerodynamic diameter
         MMD - Mass median diameter
         5g - Geometnc standard deviation
 n
 t— H
 H
 W

-------
 1     concentrations present in the ambient atmosphere (1 to 14 /ug/m3) were not likely to have a
 2     significant acute effect in healthy individuals.  These metals include arsenic, cadmium, copper,
 3     vanadium, iron, and zinc.  Other metals found at concentrations less than 0.5 //g/m3 were not
 4     reviewed in the previous criteria document.  However, published data added to the existing PM
 5     data base demonstrate that particle-associated metals are plausible causal components of PM.
 6          Only limited controlled human exposure studies have been performed with particles other
 7     than acid aerosols (see Table 8-2).  Controlled inhalation exposure studies to high concentrations
 8     of two different metal fumes, MgO and ZnO, demonstrate the differences in response based on
 9     particle metal composition (Kuschner et al., 1997). Up to 6400 mg/m3* min cumulative dose of
10     MgO had no effect on lung function (spirometry, DLCO), symptoms of metal fume fever, or
11     changes in inflammatory mediators or cells recovered by BAL.  However, lower concentrations
12     of ZnO fume (165 to 1110 mg/m3* min) induced a neutrophilic inflammatory response in the
13     airways 20 h postexposure.  Lavage fluid PMNs, TNF-cc, and IL-8 were increased by ZnO
14     exposure. However, the concentrations used in these exposure studies exceed ambient levels by
15     more than 1000-fold. The absence of a response to an almost 10-fold higher concentration of
16     MgO compared with ZnO indicates that metal composition may be more important than particle
17     size (ultrafine/fine) when considering observed health  responses to inhaled PM.  Fine et al.
18     (1997) have shown elevated body temperature (metal fume fever) and increased levels of plasma
19     IL-6 (from 2.9 to 6.4 pg/mL) in naive subjects exposed to the 8-h TLV concentration of ZnO of
20     5 mg/m3 for 2 h.
21          Several metals have been shown to stimulate cytokine release in cultured human pulmonary
22     cells including zinc, chromium, cobalt,  and vanadium. Boiler makers, exposed occupationally to
23     approximately 400  to 500 yUg/m3 of fuel oil ash, showed acute nasal inflammatory responses
24     characterized by increased PMNs and elevated IL-8 that  were associated with vanadium levels
25     (increased about ninefold) in the upper airway (Woodin et al., 1998). Irsigler et al. (1999)
26     reported that V2O5 can induce asthma and bronchial hyperreactiviry in exposed workers.
27     A comparison of autopsy cases in Mexico City from the  1950s versus the 1980s indicated
28     substantially higher levels of (5- to 20-fold) Cd, Co, Cu,  Ni, and Pb in lung tissue from the 1980s
29     (Fortoul et al., 1996). Similar studies have examined metal content in human blood and lung
30     tissue (Tsuchiyama et al.,  1997; Osman et al.,  1998). The autopsy data suggest that chronic
31     exposures to urban air pollution leads to an increased deposition of metals in human tissues.

       March 2001                               8-6        DRAFT-DO NOT QUOTE OR CITE

-------
              TABLE 8-2. RESPIRATORY EFFECTS OF METAL PARTICLES, FUMES, AND SMOKE IN HUMANS AND
o
cr
to
O



Species, Gender,
Strain, Age, etc. Particle
Naive subjects; ZnO
8 M, 5 F; 30.8±
7.7 years old
Healthy Colloidal iron
nonsmokers; oxide
12 M, 4F; 18-35
years old
Vanadium plant V2OS
workers; 40 M,
19-60 years old

Exposure
Technique
Inhalation by
face mask
Bronchial
instillation
Ambient air
JLABUKAJH
Concentration Particle Size
2.5 mg/m3 0.3 Mm MMD
5 mg/m3
SmgmlOmL 2. 6 Mm
<0.05-1.53 N/A
mg/m3
JKY APMMA
Exposure
Duration
2h
1, 2, and 4 days
after instillation
Variable
LS
Effect of Particles
Increased oral temperature after 2.5- and 5.0-mg/m3
exposure. Elevated IL-6 after exposure to 5 mg/m3.
Symptoms (myalgia, cough, fatigue) peaked 9 h after
5 mg/m3 exposure.
L-femtm increased after iron oxide particle exposure;
transfemn was decreased Both lactoferrin and
transfemn receptors were increased.
1 2/40 workers had bronchial hyperreactivity that
persisted in some for up to 23 mo.

Reference
Fine etal. (1997)
Ohio et al
(1998a)
Irsigler et al.
(1999)
oo
Healthy         MgO
nonsmokers; 4 M,  ZnO
2 F; 21-43 years
old
                                  Inhalation       100-200 mg/m3
99% < 1.8 Mm
29% < 0.1 ^m
                                                                  45 mm
No significant differences in BAL inflammatory cell      Kuschner et al.
concentrations, BAL mterleukms (IL-1, IL-6, IL-8),       (1997)
tumor necrosis factor, pulmonary function, or peripheral
blood neutrophils.




O
5
>
T1
1
o

2,
O
H
0
G
O
H
W
O
Hrl
?3
O
H— 1
H
M

Healthy Fe,O3
nonsmokers;
27 M, 7 F, 20-36
years old

Fischer 344 rats Fe2O,
(250 g)


NMRI mice; MnO2
Mouse peritoneal
macrophage

7-week-old CdO Fume
WISTAR Furth
rats;
C57BL6 and
DBA3NCR mice
Rat, M, F344, TiO2
1 75-225 g




Intrapulmonary
instillation



Intratracheal
instillation


Intratracheal
instillation;
in vitro

Nose-only
Inhalation



Intratracheal
inhalation and
Intratrachea)
instillation


3x 10s
microspheres in
10 mL saline.


7.7 x 107
microspheres in
5 mL saline

0.037,0.12,
0.75,
2.5 mg/amma!

1 04 mg/m3
Rats dose =
18.72Mg
Mouse dose =
4.59 Mg
Inhalation at
125 Mg/m3
Instillation at
500 Mg for fine,
750 Mg for
ultrafine
2.6 Mm




2. 6 Mm



surface area of
0.16,05, 17,
62 m2/g

CMD = 0 008 Mm
8g= 1.1



Fine: 250 nm
Ultrafine: 2 1 nm




N/A




N/A



Sacrificed at
5 days


1 x3h




Inhalation
exposure, 2 h;
sacrificed at 0,
1, 3, and 7 days
postexposure for
both techniques
Transient inflammation induced initially (neutrophils,
protein, LDH, IL-8) was resolved by 4 days
postmstillation.


Transient inflammation at I day postmstillation.



LDH, protein and cellular recruitment increased with
increasing surface area; freshly ground particles had
enhanced cytotoxicity.

Mice created more metallothionem than rats, which may
be protective of tumor formation



Inflammation produced by mtratracheal inhalation (both
seventy and persistence) was less than that produced by
instillation; ultrafine particles produced greater
inflammatory response than fine particles for both dosing
methods.

Lay etal (1998)




Lay etal. (1998)



Lison et al.
(1997)


McKenna et al.
(1998)



Osier and
Oberdorster
(1997)




-------
fu
s*
oo
6
o
2
o
H
O
O
H
m
O
w
o
h—H
H
 TABLE 8-2 (cont'd).  RESPIRATORY EFFECTS OF METAL PARTICLES, FUMES, AND SMOKE IN HUMANS AND
                                              LABORATORY ANIMALS
Species, Gender,
Strain, Age, etc. Particle
Rat, M. F344, TiO2
1 75-225 g




Rats NaVO,
VOSO,
VA



Boilermakers VA
(18 M), 26-61
years old, and
utility worker
controls (1 1 M),
30-55 years old
Exposure
Technique
Intratracheal
inhalation and
Intratracheal
instillation


Intratracheal
instillation




Inhalation of
fuel-oil ash




Concentration Particle Size
Inhalation at Fine. 250 nm
125,ug/m' Ultrafine: 21 nm
Instillation at
500 us for fine,
750 jj.% for
ultrafine
21or210//g N/A
V/kg (NaVOj,
VOSO, soluble)
42 or 420 ,ug
V/kg (V A) 'ess
soluble
0.4-0.47 mg/m3 lO^m

0 l-O.I3mg/m3



Exposure
Duration
Inhalation
exposure, 2 h;
sacrificed at 0,
1, 3, and 7 days
postexposure for
both techniques
1 h or 1 0 days
following
instillation



6 weeks





Effect of Particles
MIP-2 increased m lavage cells but not in supernatant m
those groups with increased PMN (more in instillation
than in inhalation, more in ultrafine than in fine); TNF-ct
levels had no correlation with either particle size or
dosing methods

PMN influx was greatest following VOSO4, lowest for
VA, VOSO4 induced inflammation persisted longest;
MIP-2 and KC (CXC chemokmes) were rapidly induced
as early as I h postmstillation and persisted for 48 h;
Soluble V induced greater chemokine mRNA expression
than insoluble V; AMs have the highest expression level
Exposure to fuel-oil ash resulted in acute upper airway
inflammation, possibly mediated by increased IL-8 and
PMNs.



Reference
Osier et al.
(1997)




Pierce et al.
(1996)




Woodm et al.
(1998)




BAL - Bronchoalveolar lavage
CMD - Count median diameter
1L - Interleukin
LDH - Lactate dehydrogenase
MIP-2 - Macrophage inflammatory protem-2
mRNA - Messenger RNA (nbonucleic acid)
N/A - Data not available

-------
  1           Iron is the most abundant of the elements that are capable of catalyzing oxidant generation
  2      and also present in ambient urban particles.  Lay et al. (1998) and Ohio et al. (1998a) tested the
  3      hypothesis that the human respiratory tract will attempt to diminish the added, iron-generated
  4      oxidative stress. They examined the cellular and biochemical response of human subjects
  5      instilled with iron (III) oxide via the intrapulmonary route. Saline alone and iron-containing
  6      particles suspended in saline were instilled into separate lung segments of human subjects.
  7      Subjects underwent bronchoalveolar lavage at 1 to 91 days after instillation of 2.6-yUm diameter
  8      iron oxide agglomerates.  Lay and colleagues found the greatest iron oxide-induced inflammatory
  9      response in the alveolar fraction of the lavage fluid, although a significant increase in
 10      macrophages also was observed in the bronchial  fraction. The peak response for all cellular and
 11      biochemical changes occurred at 1 day postinstillation. Lung lavage within 1 day of exposure
 12      revealed decreased transferrin concentrations and increased ferritin  and lactoferrin
 13      concentrations, consistent with a host-generated decrease in the availability of catalytically
 14      reactive iron (Ohio et al.,  1998a). Normal iron homeostasis returned within 4 days of the iron
 15      particle instillation. The same iron oxide preparation, which contained a small amount of soluble
 16      iron, produced similar pulmonary changes in rats. Instillation of rats with two  iron oxide
 17      preparations that contained no soluble iron did not produce injury or inflammation (see Section
 18      8.2.2), thus suggesting that soluble iron was responsible for the observed intrapulmonary
 19      changes. Although only a small amount of the iron instilled in human subjects was "active", it is
20      not clear if the total dose of iron oxide delivered acutely to the lung segments (approximately
21      5 mg or 2.1 x 10s particles) would be relevant to  deposition of iron oxide particles at the
22      concentrations of iron present in ambient urban air (generally less than 1 yUg/m3).
23
24      8.2.3 Ambient Combustion-Related and Surrogate Particulate  Matter
25           The majority of the in vivo exposures to ambient particles have utilized intratracheal
26      instillation techniques in laboratory animals. Discussions on the pros and cons of this technique
27      are covered in Chapter 7 (Section 7.5), and these  issues have also been reviewed elsewhere
28      (Driscoll et al., 2000; Oberdorster et al., 1997; Osier and Oberdorster, 1997). In most of these
29      studies, PM samples were collected on filters, resuspended in a vehicle (usually saline), and a
30      small volume of the suspension was instilled intratracheally into the animals. The
31      physiochemcial characteristics of PM are altered by deposition on a filter and resuspension in an
        March 2001                                 8-9         DRAFT-DO NOT QUOTE OR CITE

-------
  1      aqueous medium. In addition, the doses used in these instillation studies are generally high
  2      compared to ambient concentrations, even when laboratory animal-to-human dosimetric
  3      differences are considered. Therefore, in terms of direct extrapolation to humans in ambient
  4      exposure scenarios, greater importance should be place on inhalation studies. However, delivery
  5      of PM by instillation has the advantage that much less material is needed and the delivered dose
  6      can be determined directly without extrapolating from estimates of lung deposition. Instillation
  7      studies have proven valuable in comparing the effects of different types of PM and for
  8      investigating the mechanisms by which particulates cause lung injury and inflammation.
  9      Tables 8-3, 8-4, and 8-5 outline studies in which various biological endpoints were measured
 10      following exposures to ambient PM, complex combustion-related PM, or laboratory-derived
 11      surrogate PM, respectively.
 12           There were only limited data available from human studies or laboratory animal studies on
 13      ultrafme aerosols at the time of the release of the previous criteria document (U.S. Environmental
 14      Protection Agency, 1996a). In vitro studies have shown that ultrafme particles have the capacity
 15      to cause injury to cells of the respiratory tract. High levels of ultrafme particles, as metal or
 16      polymer "fume," are associated with toxic respiratory responses in humans and other mammals.
 17      Such exposures are associated with cough, dyspnea, pulmonary edema, and acute inflammation.
 18      At concentrations less than 50 ywg/m3, freshly generated insoluble ultrafme teflon polymer fume
 19      particles can be severely toxic to the lung. However, it was not clear what role in the observed
20      effects was played by fume gases which adhered to the particles.  Thus, it  was not clear at the
21      time of the previous review what role, if any, ambient ultrafine particles may play in PM-induced
22      mortality and morbidity.  Newer data from clinical exposures have demonstrated that
23      composition and not particle size was responsible for the adverse health effects associated with
24      exposures to metal fumes containing both ultrafine and fine particles (Kuschner et al., 1997).
25           Toxicologic studies of other particulate matter species also were discussed in the previous
26      criteria document. These studies included exposures to fly ash, volcanic ash, coal dust, carbon
27      black, TiO2, and miscellaneous other particles, either alone or in mixture.  Some of the particles
28      discussed were considered to be models of "nuisance" or "inert" dusts (i.e., those having low
29      intrinsic toxicity) and were used in instillation studies to delineate nonspecific particle effects
30      from effects of known toxicants.  A number of studies on "other PM" examined effects of up to
31      50,000 /ug/m3 of respirable particles with inherently low toxicity. Although there was no

        March 2001                                8-10       DRAFT-DO NOT QUOTE OR CITE

-------
                                   TABLE 8-3.  RESPIRATORY EFFECTS OF AMBIENT PARTICULATE MATTER
3
tr
o
o









Species, Gender,
Strain, Age, etc

Male S-D rats
200-225 g,
control and
SCytreated

S-D rats
60 days



Particle

Concentrated
ambient particles
(Boston) (CAP)


Provo, UT,
TSP filters
(10 years old),
soluble and
insoluble extracts
Exposure
Technique

Harvard/EPA
fine particle
concentrator
animals restrained
in chamber
Intratracheal
instillation



Concentration Particle Size Exposure Duration

206,733, 607 vg/m3 for 0 1 8 ^m 5 h/day for 3 days
Days 1-3; 29 °C, 6g = 2.9
59% RH


100-1000/^gofPM N/A 24 h
extract in 0.5 mL saline



Effect of Particles

PEF and TV increased in CAPS exposed animals.
Increased protein and % neutrophlls and
lymphocytes in lavage fluid after CAPS exposure.
Responses were greater in SO2-bronchitis animals.
No changes in LDH. No deaths occurred.
Inflammation (PMN) and lavage fluid protein was
greater with the soluble fraction containing more
metal (Zn, Fe, Cu).


Reference

Clarke et al.
(1999)



Ohio et al
(1999a)



oo
Healthy
nonsmokers;
18 to 40 yr old

Mongral dogs,
some with balloon
occluded LAD
coronary artery
n= 14
                         CAP
                         CAP
          Male F 344 rats,
          monocrotahne
          treated
                         CAP
                               Inhalation        23.1 to 311.1 ^g/m3      0.65 ^m          2 h; analysis at 18 h      Increased BAL neutrophils m both bronchial and   Ghioetal
                                                                     6g = 2 35                                alveolar fractions                             (2000a)


                               Inhalation via     69-828 ,ug/m3           0 23 to 0 34 ^m    6 h/day * 3 days         Decreased respiratory rate and increased lavage    Godleski et al.
                               tracheostomy                           6g = 0 2 to 2 9                           fluid neutrophlls in normal dogs                (2000)
                                Inhalation       132 to 919 /^g/m3        0.2 to 1.2 ,um      I*3hor3x6h         No inflammatory responses, no cell damage       Gordon et al.
                                                                     6g = 0 2  to 3 9                            responses, no PFT changes.                    (2000)
 H
 b
 o
 2
 o
 H
O
 M
 O
 ?o
 o
 HH
 H
 m
8-mo-old Bi TO-2
male hamsters

Rats
                CAP
                                Nose-only
                                inhalation
S-D rats         TSP collected in   Intratracheal
Human Bronchial Provo           instillation
Epithelial
(BEAS-2B) cells
                                               11 0-350 Mg/m3
                                               TSP filter samples (36 5
                                               mg/mL) agitated m
                                               deionized H,O2 for 96 h,
                                               centnfuged at 1200g for
                                               30 mm, lyophyhzed and
                                               resuspended m deionized
                                               H202 or saline
N/A              3 h                     Increased penpheral blood neutrophlls and        Gordon et al.
                                        decreased lymphocytes.                       (1998)

N/A (TSP samples,  Sacrificed at 24 h         Provo particles caused cytokme-mduced         Kennedy et al.
comprised 50 to                           neutrophil-chemoattractant-dependent           (1998)
60% PMK))                               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 m primary cultures of
                                        BEAS-2B cells; cytokine secretion was preceded
                                        by activation of NF-KB and was reduced by SOD,
                                        DEF, or NAC; quantities of Cu2+ found in Provo
                                        particles replicated the effects

-------








|
|
H
NT PARTICULA

S
5

0
i/5
H
U
U
fa
W
X
0
H

K
^^
PH

fT1

.
?

C
O
u


f)
00
td
-1
ca
•^
H









o
CD
W





"o
Effect of Part


« c
3 2
|1
w


S
c/5
JD
£
a.

_.
0
2
*^
g
g
0
U


S S
3 CT*
o c

X gj





ju

«
Q-

w"
*O 4)
W u"
O w>

|.s
8.S
t« VI


cd
w
c
ra ''reT

iS §.
s |
l|
h; (S
;ased BALF protein and neul
chitic rats; responses were \;
isure regimens.

lit

x
tt
IS 17
vo <~J









m
I

O

sO


.2
'•^
—

w
£





cu
U
rt -*-•
>-Q g_
-S . -£ ^


o 2 1 gj
p^
0\
~-
g;
c.

ra
^
"5
'1 1 1 1
X m » t-o
;ased PMN, protein, and LD
ter response with ultrafme C
eased GSH level in BAL; fre
lete supercoil DNA), leukoc:
lals produced greater NO am
C a g a. c
£ ob-o 2- §
M
•a
o
0 j-
t/2 ^0

E
s?s
=> II c cQ
SmgS
S 0 S D



c
00

CN) C
"T
0 ^
iri o
S1""*
c
-c o
03 -^
*3 — -"
2 s
i: vi
£ S

oa"
S u u
&• -fi >~
§ 3 ^ - — •
— -^ *2 CO
•^ PJ O ^— '
~^r

X

!/)
It
f* t8
March 2001
8-12
DRAFT-DO NOT QUOTE OR CITE

-------
ta
§•
K>
o
o
               TABLE 8-4.  RESPIRATORY EFFECTS OF COMPLEX COMBUSTION-RELATED PARTICULATE MATTER
OO
 I

U)
H
6
o
2
o
H
O
d
o
H
W
Species, Gender, Strain,
Age, etc.
M Syrian golden
hamsters 90- 125g




NMRI mouse

















Rats, male, S-D, 60 days
old
MCT (60 mg/kg), ip






Exposure
Particle Technique
Kuwaiti oil fire Intratracheal
particles instillation
Urban particles
from St. Louis,
MO

CFA Intratracheal
CMP instillation
WC















Emission Intratracheal
source PM instillation
Ambient
airshed PM
ROFA




Concentration
0.15, 0.75, and 3.75
mg/lOOg




CMP: 20 //g
arsenic/kg, or CMP:
lOOmg
particles/kg,
WC alone
(100 mg/kg), CFA
alone (100 mg/kg
[i.e., 20 Mg
arsenic/kg]), CMP
mixed with WC
(CMP, 13.6 mg/kg
[(i.e., 20 ng
arsenic/kg]), WC
(86.4 mg/kg) and
Ca3(AsO4)2 mixed
with WC (20 ^g
arsenic/kg), WC
(100 mg/kg)
Total mass:
2.5 mg/rat

Total transition
metal: 46 ^g/rat




Particle Size
Oil fire particles:
<3.5 /J.m, 10 days of
24-h samples (April
30 to May 9, 1991),
in Ahmadi, Kuwait

N/A

















Emission PM:
1 78-4.1 7 ^m

Ambient PM-
3.27-4 09 ^m




Exposure
Duration
Sacrificed 1 and
7 days post
instillation



1,5, 30 days
posttreatment,
lavage for total
protein content,
inflammatory
cell number and
type, and TNF-
a production
particle
retention








Analysis at 24
and 96 h
following
instillation





Effect of Particles
Increases in PMN, AM, albumin, LDH,
myeloperoxidase, and
(3-N-acetylglucosaminidase;
acute toxicity of the particles found in the
smoke from the Kuwaiti oil fires is
comparable to that of urban particles.
Mild inflammation for WC; Ca3(AsO,)2
caused significant inflammation;
CMP caused severe but transient
inflammation; CFA caused persistent
alveolitis; cytokine production was
upregulated in WC- and Ca3(AsO4) treated
animals after 6 and 30 days, respectively;
a 90% inhibition of TNF-a production still
was still observed at Day 30 after
administration of CMP and CFA;
a significant fraction persisted (10-15% of
the arsenic administered) in the lung of
CMP- and CFA-treated mice at Day 30
Suppression of TNF-a production is
dependent on the slow elimination of the
particles and their metal content from the
lung

Increases in PMNs, albumin, LDH, PMN,
and eosmophils following exposure to
emission and ambient particles;
induction of injury by emission and
ambient PM samples is determined
primarily by constituent metals and their
bioavailability;
MCT-ROFA show enhanced neutrophihc
inflammation and an increase in mortality.
Reference
Brain et al. (1998)





Broeckaert et al.
(1997)
















Costa and Dreher
(1997)







         WISTAR male rats
         Bor: WISW strain
Coal oil fly ash  Inhalation
             (chamber)
0,11,32, and      1.9-2.6 ^m         6 h/day,        At the highest concentration, type II cell
103mg/m3        6g=1.6-18         5days/week,    proliferation and mild fibrosis occurred and
                                 4 weeks        increased penvascular lymphocytes were
                                              seen. The mam changes at the lowest
                                              concentration were particle accumulation in
                                              AM and mediastmal lymph nodes.
                                              Lymphoid hyperplasia observed at all
                                              concentrations. Effects increased with
                                              exposure duration
Dormans et al.
(1999)

-------
TABLE 8-4 (cont'd). RESPIRATORY EFFECTS OF COMPLEX COMBUSTION-RELATED PARTICULATE MATTER
g.
K>
O
O















oo

*•



a

j>
3
>— j
i
o
o
o
H
O
c

w
n
NV/
o
H
m

Species, Gender, Strain,
Age, etc.

Rat, male, S-D, 60 days
old
Male S-D rats 60 days
old

5-day-old male S-D rats




Rat, male, S-D, 60 days
old







Female
Balb/cJ
mice 7- 1 5 weeks

7-week-oId Female
Balb/cJ mice (16-21 g)




Rat, male, S-D


Mice, normal and Hp,
105 days old




2-day-old BALB/C mice
sensitized to ovalbumin
(OVA)
Particle

ROFA

#6 ROFA,
volcanic ash

LowS
#6 ROFA,

volcanic ash
saline
Two ROFA
samples
Rl had 2*
salme-
leachable
sulfate, Ni, and
V and 40* Fe
as R2; R2 had
3 1 * higher Zn
ROFA



ROFA lo-S
residual oil




ROFA


ROFA





Aerosolized
ROFA leachate

Exposure
Technique

Intratracheal
instillation
Instillation in NaCl
solution

Intratracheal




Intratracheal
instillation







Intratracheal
instillation


Inhalation and
instillation
challenge with
OVA


Intratracheal
instillation

Intratracheal
instillation




Nose-only
inhalation

Concentration

8 33 mg/mL
0.3 mL/rat
03, 1.7
8 3 mg/mL
8 3 mg/mL
0 3, 1 7,
8 3 mg/mL
in saline
8 3 mg/kg BW
1 mL/kg BW
2 5 mg in 0 3 mL








60 Mg in 50 ML
(dose 3mg/kg)


1 58 ± 3 mg/m1





500 Mg/ammal


50 Mg





50 mg/mL


Particle Size

1 95 Mm MMAD

1 .95 Mm
6g = 2.19
1 4 Mm
1 .95 Mm
6g = 1 95
1 4 Mm


Rl: 1 88 Mm,
MMAD
R2. 2.03 Mm,
MMAD





<25



PM,5 sample





3. 6 Mm


1 95 Mm








Exposure
Duration

Analysis at
24 and 96 h
24 h


24 h




Analysis at
4 days







N/A



1,3,8, 15 days
after
instillation



Analyzed
4 and 96 h
postexposure
Analysis at
24 h




30 mm


Effect of Particles

Increased PMNs, protein, LDH at both time
points.
Plasma flbnnogen elevated after ROFA
instillation but not volcanic ash

Increased WBC count in ROFA-exposed rats
plasma flbnnogen increased 86% in ROFA
rats at highest concentration.


Four of the 24 animals treated with R2 or R2s
(supernatant) died; none in R 1 s treated
animals; more AM, PMN, eosmophils 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 hyperreactwity were
worse in R2 and R2s groups
ROFA caused increases in eosmophils, IL-4
and IL-5 and airway responsiveness in
ovalbumm-sensitized and challenged mice.

Increased BAL protein and LDH at 1 and
3 days but not at 1 5 days postexposure.
Combined OVA and ROFA challenge
increased all damage markers and enhanced
allergen sensitization. Increased methacholme
response after ROFA.
Femtm and transfemn were elevated, greatest
increase in femtm, lactofemn, transfemn
occurred 24 h postexposure.
Diminished lung injury (e g., decreased lavage
fluid ascorbate, protein, lactate
dehydrogenase, inflammatory cells, cytokmes)
in Hp mice lacking transfemn; associated
with increased metal storage and transport
proteins
Increased airway response to methylchohne
and to OVA in ROFA exposed mice,
increased airway inflammation also.
Reference

Dreheretal. (1997)

Gardner et al. (2000)


Gardner et al. (2000)




Gavettetal (1997)








Gavettetal (1999)



Gavettetal (1999)





Ghioetal. (1998b)


Ghioetal (2000b)





Hamadaetal (1999)



-------
s
to
3
cr
K)
O
O
TABLE 8-4 (cont'd).  RESPIRATORY EFFECTS OF COMPLEX COMBUSTION-RELATED PARTICULATE MATTER
oo
H
6
o
2
o
H
0
G
O
H
M
O
5*
n
H
Species, Gender, Strain,
Age, etc. Particle
Rat, male, S-D, ROFA
60 days old
Rats FOFA
MCT
Exposure
Technique
Intratracheal
instillation
Inhalation
Concentration
1 0 mg in 0.5 mL
saline
580 ± 110^g/m3
Particle Size
1 .95 nm
2.06 //m MMAD
8g=157
Exposure
Duration
Analysis at
24 h
6 h/day for
3 days
Effect of Particles
Increased PMNs, protein.
Death occurred only in MCT rats exposed to
ROFA. Neutrophils in lavage fluid were
increased significantly in MCT rats exposed
to ROFA versus filtered air. MIP-2 mRNA
expression in lavage cells was induced in
normal animals exposed to fly ash.
Reference
Kadiiska et al.
(1997)
Killingsworth et al.
(1997)
          Male S-D and F-344 rats   ROFA
          (60 days old)
          Male S-D, WIS, and
          F-344 rats (60 days old)
                     ROFA
Intratracheal        8.3 mg/kg         1 95 /j.m            Sacrificed at     Increase in neutrophils in both S-D and F-344
instillation                          6g = 2.14           24 h           rats; a time-dependent increase in eosmophils
                                                                  occurred in S-D rats but not in F-344 rats.

Intratracheal        8 3 mg/kg         1.95 t*m            Sacrificed at 6,   Inflammatory cell infiltration, as well as
instillation                          6g = 2.14           24,48, and      alveolar, airway, and interstitial thickening in
                                                    72 h; 1,3, and   all three rat strains; a sporadic incidence of
                                                    12 weeks       focal alveolar fibrosis in S-D rats, but not in
                                                                  WIS and F-344 rats; cellular fibronectm (cFn)
                                                                  mRNA isoforms EIIIA(+) were up-regulated
                                                                  in S-D and WIS rats but not in F-344 rats. Fn
                                                                  mRNA expression by macrophage and
                                                                  alveolar and airway epithelium and within
                                                                  fibrotic areas m S-D rats; increased presence
                                                                  of Fn EIIIA(+) protein in the areas of fibrotic
                                                                  injury and basally to the airway epithelium.
Kodavanti et al.
(1996)


Kodavanti et al
(1997a)
Male S-D Rats,
60 days old






Male S-D rats,
60 days old






ROFA Intratracheal
instillation
Fe2(S04)3,
VSO4,
NiSO4



10 ROFA Intratracheal
compositionally instillation
different
particles from a
Boston power
plant


8.33 mg/kg 1.95^m
6g = 2.14
ROFA-equivalent
dose of metals




0.833,333,8.3 1. 99-2.59 laa
mg/kg MMAD






Analysis at 3,
24, and 96 h,
postmstillation





Sacrificed at
24 h






ROFA-induced pathology lesions were as
severe as those caused by Ni. Metal mixture
caused less injury than ROFA or Ni alone; Fe
was less pathogenic. Cytokme and adhesion
molecule gene expression occurred as early as
3 h after exposure V-mduced gene expression
was transient but Ni caused persistent
expression and injury.
ROFA-induced increases in BAL protein and
LDH, but not PMN, were associated with
water-leachable total metal, Ni, Fe, and S;
BALF neutrophilic inflammation was
correlated with V but not Ni or S.
Chemiluminescence signals m vitro (AM)
were greatest with ROFA containing soluble
V and less with Ni plus V.
Kodavanti et al
(1997b)






Kodavanti et al.
(1998a)







-------



td
H


^
^?
W
H
ON-RELATED PARTICULA
H
OJ
^^
CQ
5
O
U
X
td
•J
o.

O
U
ts.
O
C/5
H
U
W
ta
fe
U
OS
o
H
S
NH
0.
!/3
S
^"H
a
^
"^^
G
O
Q

^
X
N
Cfl
H



u
c
3J
(H
&



Effect of Particles

1 I

x 5
uj D


u
N
"2
•^
£
£


c
o
S
Concent


S 3
§ 5
tl




£
-H
rt
a.



c"
i
C/3
i."
•o
C
G
S s





«
4J

i_
w c^
1§

Both IT and IN rats showed inflammatory
responses (1L-6, MIP-2, 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) in MCT rats.
Z
L.~JS
^ 1 —

S 5 i?Q
^ ^- ^j
fN ^O m 03




^.
E — •
in II
°! oo
— . ^
x: 5 ^ §
2 Js 9 "S
S 3 g j3
£ £ Z £




.
u.
O





.
— U
rt ^
111?
V A S
T Q « Q
O i P i^




«
^
—
Iff
« 2
•gs
£ ci

Increased BALF protein and LDH alveotis
with macrophage accumulation m alveoli;
increased neutrophils in BAL. Increased
S

„ cfl
a1 Z
~° -3 j=
2 § s




•a-
E -
m II
o^ M
— «o



si
oo ~oi)
o £

S™
c
IS
2.2
2 3
£ ^




^
O
a!.

,
S
l-
tf\
•a 3
c o
M  2
"c5 j_











pulmonary protein leakage and inflammation
in SH rats. Effects of metal constituents of
ROFA were strain specific, vanadium caused
pulmonary injury only in SH rats; nickel was
toxic in both SH and WKY rats.
















00
"o
~~ '









o
- 4
£°s
> Z S













"ra
a

!„
M """
|I












More puhnonary injury in SH rats. Increased
RBCsm BALF of SH rats. ROFA increased
airway reactivity to Ach in both SH and
WKY rats. Increased protein, albumin, and
LDH in BALF after ROFA exposure
(SH>WKY). Increased oxidative stress in
SH rats. SH rats failed to increase
glutathlone.
& 0

X «
a Z

^ c °c




^.
£ ~
m II
<*! oo
— <0



"E
E
in
~~"



~e .3
o •£





+
u,
O


„
1fl
•-
£fi
T3 2
ni o
^" "S
^ u
i?
Ji -
"(5 '
s -









































^^

ON


"rt
4J
•g
_C
1

ROFA enhanced the response to house dust
mite (HDM) antigen challenge. Eosmophil
numbers, LDH, BAL protein, and IL-10 were
increased with ROFA + HDM versus HDM
alone.




„(.
Z






<
Z



OO 00
a. =i
§8



S n
11
2 "5
i: u5
_c c




.,
O
oi




15
£•
td
S
1
g
m
^


2
^r
"c3
tj
g
-o
T3
cd

Production of acetaldehyde increased at 2 h
postinstillation


0
e
1 .c
— (N


E
OO

O
-H
Ov
— •

U OJQ
ji e
|l-if
o^2 —



"S
1

ROFA increased PGE2 via cycloxygenase
expression

H
^j
O
G.

r-j






<
z

E
^
o
o
o
"*


S c
•5 2
si
2 ~
£ .S




^
tu
O
OL

tn
e
Q
or)
jj
1
2
3
1
O
0
o
o
CM
*e3
^,
(U
C
O

,0

No inflammation until 24 h.

^=
^"
o

— :







r-i

H
^4
§1
— V3


S c
"S 2
2 i
2 "5
J=5 H



^
O
"» S
Jj ^

i2
2
9
(/)
,u
1
"3
ra
-o
0
March 2001
8-16
DRAFT-DO NOT QUOTE OR CITE

-------
         TABLE 8-4 (cont'd).  RESPIRATORY EFFECTS OF COMPLEX COMBUSTION-RELATED PARTICULATE MATTER
o
to
 I
-J
"fl
H
6
o
1
o
H
O
H
W
O
&
n
si
Species, Gender, Strain,
Age, etc.
Particle
Exposure
Technique
Concentration
Particle Size
Exposure
Duration
Effect of Particles
Reference
        60-day-old male S-D rats  Ottawa dust,    Intratracheal       Dose  0, 0.25,     1.95 /j.m          6 h/day for     IT ROFA caused acute and dose-related      Watkmson et al
        and 60-day male        ROFA, and     instillation, nose-    1.0, and                         3-day         increase in pulmonary inflammation; no effect  (2000)
        Wistar-Kyoto rats,       volcanic ash    only inhalation     2.5 mg/rat                       inhalation;     of volcanic ash
        60-day-old male SH rats,                                                            instillation -
        some cold-stressed,                                                                96 h post-IT
        some ozone-exposed,
        some MCT-treated

-------
                  TABLE 8-5. RESPIRATORY EFFECTS OF SURROGATE PARTICULATE MATTER
1-1
o
O
O
oo
I

oo
H

6
o

2
o
H

O


O
H
W

O
?=»
O
H^
H
m
Species, Gender, Strain,
Age, etc
Syrian golden hamsters
900 M
900 F

C57B1/6J mice







Rat





Mice,C57BL/6J,
8 weeks and 8 mo old



MCT-treated S-D rats


Male S-D rat
(200g)




Male mice, 6-8 weeks
old (AJ, AKRJ, Sulfate,
B6C3F1), BALBcJ
strains raised in a
pathogen free laboratory
Swiss- Webster mice
Particle
Toner

TiO2
Silica
PTFE
TiO2






PTFE Fumes





PTFE Fumes




Fluorescent
microspheres

Diesel,
SiO2,
carbon black



Carbon black
Regal 660

SO4"


Exposure
Technique
Nose-only
inhalation


Inhalation







Whole body
inhalation




Whole body
inhalation



Inhalation


Intratracheal
instillation




Nose only
inhalation




Concentration
1 5, 6.0, or
24 mg/m3
40 mg/m3
3 mg/m3
PTFE:
1.25, 25, or
5x 105
particles/cc
TiO2-F 10 mg/m3
NiO. 5 mg/m3
Ni3S,: 0 5 mg/m3

1,2.5, or 5 x 10s
particles/cm3




1,2.5, or 5 * 10s
particles/cm3



3.85 ±0.81
Mg/1"3

1 mg in 0 4 mL.





10 Mg/m3


285 Mg/m3


Particle Size
4.0 Mm

1 1 Mm
1.4 Mm
PTFE: 18nm
TiCyF: 200 nm
TiO2-D. lOnm





18nm





18nm




1.0 Mm
1 38 ±0.10 Mm

DEP Collected as
TSP - disaggregated
in solution by
somcation (20 nm);
SiO2 (7 nm),
carbon black
0.29 Mm
±2. 7 Mm




Exposure
Duration
3,9, 15 mo
6h/day
5days/week

30 min or
6 h/day,
5days/week,
6 mo




1 5 mm,
analysis 4 h
postexposure



30-mm
exposure,
analysis 6 h
following
exposure
3 h/day
x 3 days

Sacrificed at
2,7,21,42,
and 84 days
postmstillation


4h





Effect of Particles
Retention increased with increased exposure.



Effects on the epithelium caused by direct
interactions with particles, not a result of
macrophage-derived mediators, and suggest
a more significant role in the overall
pulmonary response than previously
suspected; type II cell growth factor
production may be significant in the
pathogenesis of pulmonary fibrosis.
Increased PMN, mRNA of MnSOD and MT,
IL-la, IL-1P, IL-6, MIP-2, TNF-a mRNA of
MT and IL-6 expressed around all airways and
interstitial regions; PMN expressed IL-6, MT,
and TNF-a; AM and epithelial cells were
actively involved.
Increased PMN, lymphocytes, and protein
levels in old mice over young mice; increased
TNF-a mRNA in old mice over young mice;
no difference in LDH and P-Glucuronidase

Monocrotahne-treated animals contained
fewer microspheres in their macrophages,
probably because of impaired chemotaxis.
Amorphous SiO, increased permeability, and
neutrophilhc inflammation. Carbon black
and DEP translocated to mterstltum and
lymph nodes by 1 2 weeks.


Differences in inflammatory responses
(PMN) across strains Appears to be genetic
component to the susceptibility



Reference
Creutzenberg et al
(1998)


Fmkelstem et al.
(1997)






Johnston et al. (1996)





Johnston et al. (1998)




Madletal (1998)


Murphy et al. (1998)





Ohtsuka et al. 2000






-------
  1     mortality, some mild pulmonary function changes after exposure to 5,000 to 10,000 /ug/m3 of
  2     inert particles were observed in rats and guinea pigs.  Lung morphology studies revealed focal
  3     inflammatory responses, some epithelial hyperplasia, and fibrotic responses after exposure to
  4     >5,000 /ug/m3. Changes in macrophage clearance after exposure to >10,000 /ug/m3 were
  5     equivocal (no infectivity effects). In studies of mixtures of particles and other pollutants, effects
  6     were variable depending on the toxicity of the associated pollutant, hi humans, co-exposure to
  7     carbon particles appeared to increase responses to formaldehyde but not to acid aerosol. None of
  8     the "other" particles mentioned above are present in ambient air in more than trace quantities.
  9     Thus, it was concluded that the relevance of any of these studies to standard setting for ambient
 10     PM may be extremely limited.
 11          Recent studies that examined the acute effects of intratracheal instillation of ambient PM
 12     have shown clearly that PM obtained from various sources can cause lung inflammation and
 13     injury. Costa and Dreher (1997) showed that instillation of PM samples from three emission
 14     sources (two oil and one coal fly ash) and four ambient airsheds (St Louis, MO; Washington,
 15     DC; Dusseldorf, Germany; and Ottawa, Canada) resulted in increases in lung PMN and
 16     eosinophils in rats 24 h after instillation. Biomarkers of permeability (total protein and albumin)
 17     and cellular injury (LDH) also were increased. This study demonstrated that the lung dose of
 18     bioavailable transition metal, not instilled PM mass, was the primary determinant of the acute
 19     inflammatory response. Kennedy et al. (1998) reported a similar dose-dependent inflammation
 20     (i.e., increase in protein and PMN in lavage  fluid, proliferation of bronchiolar epithelium, and
 21      intraalveolar hemorrhage) in rats instilled with water-extracted particles (TSP) collected in
 22     Provo, UT. This study also  indicated that the metal constituent, in this case PM-associated Cu,
 23      was a plausible cause of the outcome.  Likewise, instillation of ambient PM10 collected in
 24      Edinburgh, Scotland, also caused pulmonary injury and inflammation in rats (Li et al., 1996,
 25      1997).  Brain et al. (1998) examined the effects of instillation of particles that resulted from the
 26      Kuwaiti oil fires in 1991  compared to urban particulate matter collected in St. Louis (NITS SRM
 27      1648, collected in a bag house in early 1980s) and showed that on an equal mass basis, the acute
 28      toxicity of the Kuwaiti oil fire particles was  similar to that of urban particles collected in the
29      United States.
30           The fact that instillation of ambient PM collected from different geographical areas and
31      from a variety of emission sources consistently caused pulmonary inflammation and injury tends

        March 2001                               8-19       DRAFT-DO NOT QUOTE OR CITE

-------
 1     to corroborate epidemiological studies that report increased respiratory morbidity and mortality
 2     associated with PM in many different geographical areas and climates. However, high-dose
 3     instillation studies may produce different effects on the lung than inhalation exposures done at
 4     more relevant concentrations. This concern is somewhat diminished by the results of an
 5     inhalation study of concentrated PM in healthy nonsmokers. Ghio et al. (2000a) exposed 38
 6     volunteers exercising intermittently at moderate levels of exertion for 2 h to either filtered air or
 7     particles concentrated from the air in Chapel Hill, NC (23 to 311 ug/m2). Analysis of cells and
 8     fluid obtained 18 h after exposure showed a mild increase in neutrophils in the bronchial and
 9     alveolar fractions.  No respiratory symptoms or decrements in pulmonary function were found
10     after exposure to CAP.
11           Because emission sources contribute to the overall ambient air particulate burden (Spengler
12     and Thurston, 1983), many studies investigating the response of laboratory animals to particle
13     exposures have used fly ash as a useful source of particle for exposure (see Table 8-3).  The
14     residual oil fly ash (ROFA) samples used in toxicological studies have been collected from a
15     variety of sources such as boilers, bag houses used to control emissions from power plants, and
16     from the fine particles that are emitted downstream of the collection devices.
17           ROFA has a high content of water soluble sulfate and metals, accounting for 82 to 92% of
18     water-soluble mass, while the water-soluble mass fraction in ambient air varies from  low teens to
19     more than 60% (Costa and Dreher, 1997; Prahalad et al., 1999). More than 90% of the metals in
20     ROFA are transition metals, whereas these metals are only a small subfraction of the  total
21     ambient PM mass. Thus, the dose of bioavailable metal that is delivered to the lung when ROFA
22     is instilled into a laboratory animal can be orders of magnitude greater than a ambient PM dose,
23     even under a worst-case scenario.
24           Intratracheal instillation of various doses of ROFA suspension has been shown  to produce
25     severe inflammation, an indicator of pulmonary injury that includes recruitment of neutrophils,
26     eosinophils, and monocytes into the airway. The biological effects of ROFA in rats have been
27     shown to depend on aqueous leachable chemical constituents of the particles (Dreher et al., 1997;
28     Kodavanti et al., 1997b).  A leachate prepared from ROFA, containing predominantly Fe, Ni, V,
29     Ca, Mg, and sulfate, produced similar lung injury to that induced by the complete ROFA
30     suspension (Dreher et al., 1997). Depletion of Fe, Ni, and V from the ROFA leachate eliminated
31     its pulmonary toxicity.  Correspondingly, minimal lung injury was observed in animals exposed

       March 2001                               8-20        DRAFT-DO NOT QUOTE OR CITE

-------
  1     to saline-washed ROFA particles. A surrogate transition metal sulfate solution containing Fe, V,
  2     and Ni largely reproduced the lung injury induced by ROFA. Interestingly, ferric sulfate and
  3     vanadium sulfate antogonized the pulmonary toxicity of nickel sulfate. Interactions between
  4     different metals and the acidity of PM were found to influence the severity and kinetics of lung
  5     injury induced by ROFA and its soluble transition metals.
  6          To further investigate the response to ROFA with differing metal and sulfate composition,
  7     male Sprague Dawley rats (60 days old) were exposed to ROFA or metal sulfates (iron,
  8     vanadium, and nickel, individually or in combination) (Kodavanti et al., 1997b).  Transition
  9     metal sulfate mixtures caused less injury than ROFA or Ni alone, suggesting metal  interactions.
 10     In addition, this study showed that V-induced effects were less severe than that of Ni and were
 11      transient. Ferric sulfate was least pathogenic.  Cytokine gene expression was induced prior to the
 12     pathology changes in  the lung and the kinetics of gene expression suggested persistent injury by
 13     nickel sulfate. Another study by the same investigators was performed using 10 different ROFA
 14     samples collected at various sites within a power plant burning residual oil firing chamber
 15     (Kodavanti et al., 1998a). Animals received intratracheal instillations of either saline, or a saline
 16     suspension of whole ROFA (<3.0 yum MMAD) at three concentrations (0.833,  3.33, or
 17     8.33 mg/kg). This study showed that ROFA-induced PMN influx appeared to be associated with
 18     its water-leachable V  content; however, protein leakage appeared to be associated with water-
 19     leachable Ni content.  ROFA-induced in vitro activation of AM was highest with ROFA
 20     containing leachable V but not with Ni plus V, suggesting that the potency and the mechanism of
 21      pulmonary injury may differ between emissions containing bioavailable V and  Ni.
 22          Other studies have shown that soluble metal components play an important role in the
 23      toxicity of emission source particles.  Gavett et al. (1997) investigated the  effects of two ROFA
 24      samples of equivalent diameters, but having different metal and sulfate content, on pulmonary
 25      responses in Sprague-Dawley rats. ROFA sample 1  (Rl) (the same emission particles used by
 26      Dreher et al. [1997]) had approximately twice as  much saline-leachable sulfate, nickel, and
27      vanadium, and 40 times as much iron as ROFA sample 2 (R2); whereas R2 had a 31 -fold higher
28      zinc content. Rats were instilled with suspensions of 2.5 mg R2 in 0.3 mL saline, the supernatant
29      of R2 (R2s), the  supernatant of 2.5 mg Rl (Rls), or saline only. By 4 days after instillation, 4 of
30      24 rats treated with R2s or R2 had died. None of those treated with Rls or saline died.
31      Pathological indices, such as alveolitis, early fibrotic changes and perivascular edema, were

        March 2001                               8-21        DRAFT-DO NOT QUOTE OR CITE

-------
  1      greater in both R2 groups, hi surviving rats, baseline pulmonary function parameters and airway
  2      hyperreactivity to acetylcholine were significantly worse in R2 and R2s groups than in the Rls
  3      groups.  Other than BAL neutrophils, which were significantly higher in the R2 and R2s groups,
  4      no other inflammatory cells (macrophages, eosinophils, or lymphocytes) or biochemical
  5      parameters of lung injury were significantly different between the R2 and R2s groups and the
  6      Rls group.  Although soluble forms of zinc had been found in guinea pigs to produce a greater
  7      pulmonary response than other sulfated metals (Amdur et al., 1978), and, although the level of
  8      zinc was 30-fold greater in R2 than Rl, the precise mechanisms by which zinc may induce such
  9      responses are unknown. Nevertheless, these results show that the composition of soluble metals
10      and sulfate leached from ROFA, a type of emission source particle, is critical in the development
11      of airway hyperractivity and lung injury.
12           It has been shown that reactive oxygen species play an important role in the in vivo toxicity
13      of ROFA, Dye et al. (1997) pretreated rats with an intraperitoneal injection of saline or
14      dimethylthiourea (DMTU) (500 mg/kg), followed 30 min later by intratracheal instillation of
15      either acidic saline (pH = 3.3) or an acidified suspension of ROFA (500 /ug).  The systemic
16      administration of DMTU impeded development of the cellular inflammatory response to ROFA,
17      but did not ameliorate biochemical alterations in BAL fluid. In a subsequent study, these
18      investigators determined that oxidant generation, possibly induced by soluble vanadium
19      compounds in ROFA, are responsible for the subsequent rat tracheal epithelial cells gene
20      expression, inflammatory cytokine productions (MIP-2 and IL-6), and cytotoxicity (Dye et al.,
21      1999).
22           In addition to transition metals, other components in fly ash also may cause lung injury.
23      The effects of arsenic compound in coal fly ash or copper smelter dust on the lung integrity and
24      on the ex vivo release of TNFa by alveolar phagocytes were investigated by Broeckaert et al.
25      (1997). Female Naval Medical Research Institute (NMRI) mice were instilled with different
26      particles normalized for the arsenic content (20 Mg/kg body weight [i.e., 600 ng arsenic/mouse])
27      and the particle load (100 mg/kg body weight [i.e.,  3 mg/mouse]). Mice received tungsten
28      carbide (WC) alone, coal fly ash (CFA) alone,  copper smelter dust (CMP) mixed with WC, and
29      Ca3(AsO4)2 mixed with WC (see Table 8-2 for concentration details). Copper smelter dust
30      caused a severe but transient inflammatory reaction, whereas a persisting alveolitis (30 days
31      postexposure) was observed after treatment with coal fly ash. hi addition, TNFa production  in

        March 2001                               8-22        DRAFT-DO NOT QUOTE OR CITE

-------
  1      response to LPS by alveolar phagocytes were significantly inhibited at Day 1 and still was
  2      observed at 30 days after administration of CMP and CFA. Although arsenic was cleared from
  3      the lung tissue 6 days after Ca3(AsO4)2 administration, a significant fraction persisted (10 to 15%
  4      of the arsenic administered) in the lung of CMP- and CFA-treated mice at Day 30.  It is possible
  5      that suppression of TNF-a production is dependent upon the slow elimination of the particles and
  6      their metal content from the lung.
  7           In summary, intratracheally injected ROFA produced acute lung injury and inflammation.
  8      The water soluble metals in ROFA appear to play a key role in the acute effects of instilled
  9      ROFA.  Although studies done with ROFA clearly show that combustion generated particles
 10      with a high metal content can cause substantial lung injury, there is still insufficient data to
 11      extrapolate these effects to the low levels of particle associated transition metals in ambient PM.
 12
 13      8.2.4 Ambient Bioaerosols
 14           Ambient bioaerosols include fungal  spores, pollen, bacteria, viruses, endotoxins, and plant
 15      and animal debris. Such biological aerosols can produce various health effects including:
 16      infections, hypersensitivity, and toxicoses. Bioaerosols present in the ambient environment have
 17      the potential to cause disease in humans under certain conditions. However, it was concluded in
 18      the previous criteria document (U.S. Environmental Protection Agency, 1996a) that bioaerosols,
 19      at the concentrations present in the  ambient environment, would not account for the observed
20      effects of particulate matter on human mortality and morbidity reported in PM epidemiological
21      studies.  Moreover, bioaerosols generally represent a rather small fraction of the measured urban
22      ambient PM mass  and are typically present even at lower concentrations during the winter
23      months when notable ambient PM effects  have been demonstrated.  Bioaerosols tend to be in the
24      coarse fraction of PM, but some bioaerosols are found in the fine fraction.
25           More recent studies on ambient bioaerosols are summarized in Table 8-6.  Endotoxin
26      exposure in pig farmers is associated with a large annual decline in FEV,  (mean of 73 ml/year),
27      which is about 2 to 3 times more rapid than in healthy adults (Vogelzang et al., 1998).  Michel
28      et al. (1997) examined the dose-response relationship to inhaled lipopolysaccharide (LPS: the
29      purified derivative of endotoxin) in normal healthy volunteers exposed to 0, 0.5, 5, and 50 /ug of
30      LPS. Inhalation of 5 or 50 yUg of LPS resulted in increased PMNs in blood and sputum samples.
31      At the higher concentration, a slight (3%) but not significant decrease in FEV, was observed.
        March 2001                               8-23        DRAFT-DO NOT QUOTE OR CITE

-------













C/J
0
CNT BIOAEROS
HH
S
fe
0
t/3
H
U
U,
M
^,
o:
o
H
tf
£"H
OH
C/3
tt
vd
00
W
J
^
H














u
g
CU
a£









Z
•c
Effect of Pa;


u
1 -I
w Q


to
QJ
"0
(f

c
_0
"s
"c
c
o


a §
3 O1
0 J
UJ "

"y

c^
CU


CJ O
"O "Jj
D c^
O M
tn" <
-- C

0. -fa
CO CO

s
s ^^
c ^
o ^
U C.
fc-
0 (U

S 1 "°
-c c: in
^
.g ^ "o
|^ |
111
Rcantly reduced FEV,, incr
ho-responsiveness, and inc
ror all exposures. Effects w
of exposure.
5 c « £
a g = 3



In




<
Z




<
Z
J= S
^ *- c
ills
.3 5-sl
5 -o § "3
H § 1 8

CJ)
0) £
S 2
S 3
00 .0
S'S
0 £
IJ

•^ i ^
•^ vi O
H O f*"l DO
-C C (N CQ
•§
U4
|
ion of LPS and O, on mflamm:
Significant interact
o
S
(N
CO
CL.
J
> s
o"
o
rats. Oj and UF-C interacted •
o produce greater PMN respon
t on lung inflammatory respon:
responses in young
"priming" by LPS t
has a priming effec
and UF-C
u
B m PMN lummol-enhanced
1
•o
1
1C
1
o^
0
C
1
0
c
oo'
cu
00
o
_c
1
chemllummescence

id increase in sputum PMNs,
CRP and PMNs, an
e
£
1
'O with 5.0 ,ug LPS, increase n
PMNs, blood and unne CRP, i
monocytes, and MF
'1
lymphocytes, TNFa, and ECP
temperature, blood
PMNs, monocytes,
50 Mg LPS.
w
0>
o
at
o
s of pool-associated granulom!
Recurring outbreak
oo
Ov
£
"3
f- <~
3); case patients had higher cui
sis indicated increased levels o
pneumonitis (n = 3.
work hours. Analyi

ir and water
endotoxin m pool a
                               E
                               o
                                Q
                     f~ UI 00
                     O (N eO
                     S
                     "Ss
                     3.
                      o •£

                      c u.
                               , o
                                o
     <
     Z
                                            E
                                            3.
                                                       00
                                                       I;
                                                       U. O
                                                       'B'S
                                                       -o S
                                                                       vi
                                                                    S S •*
                                                                    IIS
                                                                    S u u
                                                                    CJ = o
                                                                    >l 8.
                                                                    Q -S '
                                                                         UJ es
                                                                         r~ II
                                                          «   £
                                                          e   2
                                                         , O   o,
                                   M
                                   O
                                   -
                                                       •2 ~ I,
                                                           -
March 2001
8-24
DRAFT-DO NOT QUOTE OR CITE

-------
  1      Cormier et al. (1998) reported an approximate 10% decline in FEV, and an increase in
  2      methacholine airway responsiveness after a 5-h exposure inside a swine containment building.
  3      This exposure induced significant neutrophilic inflammation in both the nose and the lung.
  4      Although these exposures are massive compared to endotoxin levels in ambient PM in U.S.
  5      cities, these studies serve to illustrate the effects of endotoxin and associated bioaerosol material
  6      in healthy nonsensitized individuals.
  7           Some health effects have been observed after occupational exposure to complex aerosols
  8      containing endotoxin at concentrations relevant to ambient levels. Zock et al. (1998) reported a
  9      decline in FEV, (=3%) across  a shift in a potato processing plant with up to 56 endotoxin units
 10      (EU)/m3 in the air.  Rose et al. (1998) reported a high incidence (65%) of BAL lymphocytes in
 11      lifeguards working at a swimming pool where endotoxin levels in the air were on the order of
 12      28 EU/m3. Although these latter two studies may point towards pulmonary changes at low
 13      concentrations of airborne endotoxin, it is not possible to rule out the contribution of other agents
 14      in these complex organic aerosols.
 15
 16
 17      8.3   SYSTEMIC EFFECTS OF PARTICULATE MATTER IN HEALTHY
 18           HUMANS AND LABORATORY ANIMALS:  IN VIVO EXPOSURES
 19           A small number of epidemiology studies have demonstrated that increases in cardiac-
20      related deaths are associated with exposure to PM (U.S. Environmental Protection Agency,
21      1996a), and that PM-related cardiac deaths appear to be as great or greater than those attributed
22      to respiratory causes (see Chapter 6). The toxicological consequences of inhaled particles on the
23      cardiovascular system had not  been extensively investigated prior to 1996. Since then (see
24      Table 8-7), Costa and colleagues (Costa and Dreher, 1997) have demonstrated that intratracheal
25      instillation of high levels of ambient particles can increase or accelerate death related to
26      monocrotaline administration in rats. These deaths did not occur with all types of ambient
27      particles tested.  Some dusts, such as volcanic ash from Mount Saint Helens, were relatively
28      inert, whereas other ambient dusts, including those  from urban sites, were toxic. These early
29      observations suggested that particle composition plays an important role in the adverse health
30      effects associated with episodic exposure to ambient PM, despite the "general particle" effect
31      attributed to the epidemiological associations of ambient PM exposure and increased mortality in

        March 2001                              8-25        DRAFT-DO NOT QUOTE OR CITE

-------
2
1
to
0
0
~













oo
KJ
ON



O
>
H
1
a
0
2
0
H
XD
r^^
O
H
W
O

0
HH
H- 1
D
W

TABLE 8-7. CARDIOVASCULAR EFFECTS AND OTHER SYSTEMIC
Species, Gender,
Strain Age, or
Body Weight Particle
Fischer 344 rats, OTT
male, 200-250 g



Rats, S-D, male, ROFA
60 days old,
MCT and healthy,
n = 64
Female mongrel CAP
dogs, 14 to 17 kg



Rats, male, S-D, Emission
60 days old, source PM
MCT (60 mg/kg), Ambient
ip and healthy airshed PM
ROFA

Rats, S-D, male, ROFA
60 days old



Healthy CAP
nonsmokers, 1 8 to
40 years old

Mongrel dogs, CAP
some with balloon
occluded LAD
coronary artery,
n= 14

F-344 rat, male, CAP
MCT-treated

Hamster, 6-8 mo
old, Bio TO-2
Exposure Mass
Technique Concentration
Nose-only 40 mg/m'
Inhalation



Instillation 0 0, 0.25, 1.0, and
2.5 mg/rat


Inhalation via 3-360 //g/m3
tracheostomy



Instillation Total mass:
2.5 mg/rat

Total transition
metal 46 Mg/rat

Instillation 0 3, 1 7, or
8 3 mg/kg



Inhalation 23 1 to
3111 Mg/m3


Inhalation via 69-828 Mg/m'
tracheostomy




Inhalation 132-91 9 Mg/m!




Particle
Size
4 to 5 Mm MMAD




1 95 Mm



0.2 to 0 3 Mm




Emission PM
1 78-4. 17 Mm

Ambient PM-
3. 27-4 09 Mm

1.95 Mm
6g = 219



065 Mm
8g = 235


0 23 to 0.34 Mm
6g = 02to2.9




02-1.2 Mm
5g = 0 2-3 9



Exposure
Duration
4h




Analysis at
96 h


6 h/day for 3 days




Analysis at 24
and 96 h
following
instillation


Analysis at
24 h



2 h, analysis at
18h


6 h/day for
3 days




3 h, evaluated at
3 and 24 h



EFFECTS OF PARTICULATE MATTER
Cardiovascular Effects
Increased plasma levels of endothehn- 1 .
No acute lung injury; however, lung NO
production decreased and macrophage
inflammatory protem-2 from lung lavage cells
increased after exposure.
Dose-related hypothermia and bradycardia in
healthy rats, potentiated by compromised models


Peripheral blood parameters were related to
specific particle constituents. Factor analysis
from paired and crossover expenments showed
that hematologic changes were not associated
with increases in total CAP mass concentration
ROFA alone induced some mild arrhythumas;
MCT-ROFA showed enhanced neutrophihc
inflammation;
MCT-ROFA animals showed more numerous
arrhythmias including S-T segment inversions
and A-V block.
Increased plasma fibrinogen at 8 3 mg/kg only.




Increased blood fibnnogen.



Decreased time to ST segment elevation and
increased magnitude in compromised dogs.
Decreased heart and respiratory rate and
increased lavage fluid neutrophils in normal
dogs.

No increase in cardiac arrhythmias;
PM associated increases in HR and blood cell
differential counts, and atnal conduction time
of rats were inconsistent. No adverse cardiac
or pulmonary effects in hamsters
Reference
Bouthilher et al
(1998)



Campen et al. (2000)



Clarke et al (2000)




Costa and Dreher
(1997)




Gardner et al. (2000)




Ghioetal (2000a)



Godleski et al. (2000)





Gordon et al (2000)





-------
00
H

6
O

2
O
H

/O

O
H
Species, Gender,
Strain Age, or
Body Weight Particle
Rats, S-D, MCT FOFA
(50 mg/kg SC),
250 g
12 to 13 -week-old ROFA
male WKY and SH
rats
Hartley guinea pig, DEP
male, 890 g
Healthy 10.5-year- ROFA
old beagles,
n = 4
Rabbit, New Colloidal
Zealand White, carbon
female, 1.8 to
2.4kg
Rat, S-D male, ROFA
MCT
Exposure Mass Particle Exposure
Technique Concentration Size Duration
Inhalation 580 ± 1 10 ^g/m' 2 06 ^m MMAD 6 h/day for 3 days
5g=157
Nose-only 1 5 mg/m' N/A 6 h/day for
inhalation 3 days
Intravenous 500 mg/mL 0 34 ,um 10% solution
solution every 5 mm
Oral inhalation 3 mg/m! 2 22 ^m MMAD 3 h/day for
6g = 27I 3 days
Instillation 2 mL of 1 % < 1 ^m Examined for 24
colloidal carbon to 1 92 h after
(20 mg) instillation
Instillation 0, 250, 1000, or 1 .95 ^m MMAD Monitored for
2500 Mg in 0.3 mL 8g = 2.19 96 h after
saline instillation of
ROFA particles
Cardiovascular Effects
Increased expression of the proinflammatory
chemokme MP-2 in the lung and heart of
MCT-treated rats, less in healthy rats
Significant mortality only in MCT-treated rats.
Cardiomyopathy and monocytic cell infiltration,
along with increased cytokine expression, was
found in left ventricle of SH rats because of
underlying cardiovascular disease ECG showed
exacerbated ST segment depression caused by
ROFA
DMSO extract of DEP solution induced
arrhythmias and deaths by AV block; thus,
water-soluble fractions of DEP may be
responsible for cardiotoxicity
No consistent changes in ST segment, the form
or amplitude of the T wave, or arrhythmias,
slight bradycardia during exposure.
Colloidal carbon stimulated the release of
BRDU-labeled PMNs from the bone marrow
The supernatant of alveolar macrophages treated
with colloidal carbon in vitro also stimulated the
release of PMNs from bone marrow, likely via
cytokmes
Dose-related increases in the incidence and
duration of serious arrhythmic events in normal
rats. Incidence and seventy of arrythmias were
increased greatly in the MCT rats. Deaths were
seen at each instillation level in MCT rats only
(6/12 died after MCT + ROFA)
Reference
Killmgsworth et al.
(1997)
Kodavanti et al.
(2000b)
Minami et al.
(1999)
Muggenberg et al.
(2000)
Terashima et al
(1997)
Watkmson et al.
(1998)
O
HH
H
m

-------
      TABLE 8-7 (cont'd). CARDIOVASCULAR EFFECTS AND OTHER SYSTEMIC EFFECTS OF PARTICULATE MATTER
*-«
•-(
o
to
o
1—
















oo
to
00
Species, Gender,
Strain Age, or
Body Weight
(1) Healthy S-D
rats and cold-
stressed, ozone-
treated, and
MCT-treated rats

(2) Heathy and
MCT-treated S-D
rats, SH rats,
WKY rats

(3) 15-mo-oldSH
rats


(4) MCT-treated
S-D rats



Particle
ROFA





ROFA




OTT
ROFA
MSH

Fe2(S04)3
VS04
NiSO,


Exposure
Technique
Intratracheal
instillation




Inhalation




Intratracheal
instillation


Intratracheal
instillation



Mass
Concentration Particle Size
00, 025, 10, or 1.95 ^m
2 5 mg/rat




15 mg/m3 1.95 f^m




2 5 mg I 95 Atm
0.5 mg
25mg

105/ug 1.95 Mm
245 Aig
262 5 A-g


Exposure
Duration
Monitored for
96 h after
instillation



6 h/day for
3 days



Monitored for
96 h after
instillation

Monitored for
96 h after
instillation


Cardiovascular Effects
(I) Healthy rats exposed IT to ROFA
demonstrated dose-related hypothermia,
bradycardia, and increased arrhythmias.
Compromised rats demonstrated exaggerated
hypothermia and cardiac responses to IT ROFA
Mortality was seen only in the MCT-treated rats
exposed to ROFA by IT. (2) Pulmonary
hypertensive (MCT-treated S-D) and
systemically hypertensive (SH) rats exposed to
ROFA by inhalation demonstrated similar
effects, but of diminished amplitude. There were
no lethalities by the inhalation route. (3) Older
rats exposed IT to OTT showed a pronounced
biphasic hypothermia and a severe drop in HR
accompanied by increased arrhythmias; exposure
to ROFA caused less pronounced, but similar
effects No cardiac effects were seen with
exposure to MSH. (4) Ni and V showed the
greatest toxicity, Fe-exposed rats did not differ
from controls
Reference
Watkinson et al
(2000)


















Tl
H

6
o
2

3

O
d
o
H
w


g

o
H—I
H
W

-------
  1      many regions of the United States (i.e., regions with varying particle composition).  Work that
  2      examines the role of inherent susceptibility to the adverse effects of PM in compromised animal
  3      models provides a potentially important link to epidemiological observations.
  4           To date, studies examining the systemic and cardiovascular effects of particles have used a
  5      number of compromised animal models, largely rodents, to mimic human disease. Two studies
  6      in normal or compromised dogs (Godleski et al., 2000; Muggenberg et al., 2000) also have been
  7      published as well as the preliminary results from human exposure studies (see Section 8.4.1).
  8      The following discussion of the systemic effects of PM first describes studies performed using
  9      metal-laden ROFA as a source particle and then compares these findings with studies using
10      concentrated ambient PM.
11           Killingsworth and colleagues (1997) used a fuel oil fly ash to examine the adverse effects
12      of a model urban particle in an animal model (monocrotaline-MCT) of cardiorespiratory disease;
13      MCT causes pulmonary vascular inflammation and hypertension.  They observed 42% mortality
14      in MCT rats exposed to approximately 580 /ug/m3 fly ash for 6 h/day for 3 consecutive days.
15      Deaths did not occur in MCT rats exposed to filtered air or in saline-treated rats exposed to  fly
16      ash.  The increase in deaths in the MCT/fly ash group was accompanied by an increase in
17      neutrophils in lavage fluid and  an increased immunostaining of MIP-2 in the heart and lungs of
18      the MCT/fly ash animals. Cardiac immunohistochemical analysis indicated increased MIP-2 in
19      cardiac macrophages.  The fly ash-induced deaths did not result from a change in pulmonary
20      arterial pressure; the cause of death was not identified.
21           In a similar experimental  model, Watkinson et al. (1998) examined the effects of
22      intratracheally instilled ROFA (0.0, 0.25, 1.0, 2.5 mg in 3 mL saline) on ECG measurements in
23      control and MCT rats.  They observed a dose-related increase in the incidence and duration of
24      serious arrhythmic events in control animals exposed to ROFA particles and these effects were
25      clearly exacerbated in the MCT animals. Similar to the results of Killingsworth et al. (1997),
26      health animals treated with ROFA suffered no deaths, but MCT rats had 1, 2, and 3 deaths in the
27      low-, medium-, and high-dose groups, respectively.  This study suggests that ROFA PM may be
28      implicated in conductive and hypoxemic arrhythmias associated with the cardiac-related deaths.
29           Kodavanti et al. (1999) exposed MCT rats to ROFA by either intratracheal instillation
30      (0.83 or 3.33 mg/kg) or nose-only inhalation (15 mg/m3, 6 h/day for 3 consecutive days).  Similar
31      to Watkinson et al. (1998), intratracheal instillation of ROFA in MCT rats resulted in 58%

        March 2001                                8-29       DRAFT-DO NOT QUOTE OR  CITE

-------
 1     mortality, whereas no mortality occurred in MCT rats exposed to ROFA by inhalation exposure.
 2     No mortality occurred in healthy rats exposed to ROFA or in MCT rats exposed to clean air.
 3     Despite the fact that mortality was not associated with ROFA inhalation exposure of MCT rats,
 4     exacerbation of lung lesions and pulmonary inflammatory cytokine gene expression, as well as
 5     ECG abnormalities, clearly were evident.
 6          Watkinson and colleagues further examined the effect of instilled ROFA in two additional
 7     rodent models of compromised health (Watkinson et al., 2000; Campen et al., 2000). The effect
 8     of ozone-induced pulmonary inflammation (preexposure to 1 ppm ozone for 6 h) or housing in
 9     the cold (10 °C) on the response to ROFA in rats was similar to that produced by MCT.
10     Bradycardia, arrhythmias, and hypothermic changes were consistently observed in the ozone and
11     hypothermic animals treated with ROFA, although, unlike in the MCT animals, no deaths
12     occurred. Thus, in three rodent models of cardiopulmonary disease/stress,  instillation of 0.25  mg
13     or more of ROFA can produce systemic changes that can be considered adverse health effects
14     and address potential mechanisms of toxicity consistent with the epidemiology and panel studies
15     showing cardiac effects in humans.
16          Watkinson and colleagues (2000) also sought to examine the relative toxicity of different
17     particles on the cardiovascular system of spontaneously hypertensive rats.  They instilled 2.5 mg
18     of representative particles from ambient (Ottawa) or natural (Mount Saint Helens volcanic ash)
19     sources and compared the response to 0.5 mg ROFA. Instilled particles were either mass
20     equivalent dose or adjusted to produce equivalent metal dose.  They observed adverse changes in
21     ECG, heart rate, and arrhythmia incidence that were much greater in the Ottawa- and ROFA-
22     treated rats than in the Mount Saint Helens-treated rats. The cardiovascular changes observed
23     with the Ottawa particles were actually greater than with the ROFA particles.  These series of
24     experiments by Watkinson and colleagues clearly demonstrate that instillation of particles, albeit
25     at a very high concentration, can produce cardiovascular effects. They also demonstrate that PM
26     exposures of equal mass dose did not produce the same cardiovascular effects,  suggesting that
27     PM composition was responsible for the observed effects and that PM metal content was a better
28     indicator than PM mass.
29          Because of concerns regarding the relevance of particles administered by intratracheal
30     instillation,  investigators also have examined the cardiovascular effects of ROFA particles using
31     more realistic inhalation exposure protocols. Kodavanti et al. (2000b) found that exposure to  a
       March 2001                               8-30        DRAFT-DO NOT QUOTE OR CITE

-------
  1     high concentration of ROFA (15 mg/m3 for 6 h/day for 3 days) produced alterations in the ECG
  2     waveform of spontaneously hypertensive (SH) but not normotensive rats. Although the ST
  3     segment area of the ECG was depressed in the SH rats exposed to air, further depressions in the
  4     ST segment were observed at the end of the 6-h exposure to ROFA on Days  1 and 2.  The
  5     enhanced ST segment depression was not observed on the third day of exposure, suggesting that
  6     adaptation to the response had occurred.  Thus, exposure to a high concentration of ROFA
  7     exacerbated a defect in the electroconductivity pattern of the heart in an animal model of
  8     hypertension. This ROFA-induced alteration in the ECG waveform was not  accompanied by an
  9     enhancement in the monocytic cell infiltration and cardiomyopathy that also  develop in SH rats.
 10     Further work is necessary to determine the relevance of this ROFA study to PM at concentrations
 11     relevant to ambient exposures.
 12          Godleski and colleagues (2000) have performed a series of important experiments
 13     examining the cardiopulmonary effects of inhaled concentrated ambient PM on normal mongrel
 14     dogs and on dogs undergoing coronary artery occlusion. Dogs were exposed to concentrated
 15     ambient PM for 6 h/day for 3 consecutive days. The investigators found little evidence of
 16     pulmonary inflammation or injury in normal dogs exposed to PM (daily range of mean
 17     concentrations was approximately 100 to 1000 //g/m3).  A greater than twofold increase in
 18     percent neutrophils (p < 0.05) was the only lavage parameter that was significantly different from
 19     sham-exposed animals.  Despite the absence of major pulmonary effects, a significant increase in
 20     heart rate variability (an indice of cardiac autonomic activity), a decrease in heart rate, and an
 21      increase in T alternans (an indice of vulnerability to ventricular fibrillation) were observed.  The
 22     significance of these effects is not yet clear as the effects did not occur on all  exposure days.
 23      For example, the change in heart rate variability was observed on 10 of the 23 exposure days.
 24      In support of the "general particle" theory, exposure assessment of particle composition produced
 25      no specific components of the particles that were correlated with the day-to-day variability in
 26      response. Moreover, whereas the heart rate variability change suggests a proarrhythmic response,
 27      the increase in T alternans suggests an anti-arrhythmic effect of inhaled concentrated ambient
 28      PM.
29           The most important finding in the experiments of Godleski and colleagues (2000) was the
30      observation of a potential increase in ischemic stress of the cardiac tissue from repeated exposure
31      to concentrated ambient PM. During coronary occlusion in four dogs exposed to PM, they

        March 2001                               8-31        DRAFT-DO NOT QUOTE OR CITE

-------
 1     observed a significantly more rapid development of ST elevation of the ECG waveform.  In
 2     addition, the peak ST-segment elevation was greater after PM exposure. Together, these changes
 3     suggest that concentrated ambient PM can augment the ischemia associated with coronary artery
 4     occlusion in this dog model. Additional work in more dogs as well as other species is necessary
 5     to determine the significance of these findings to the human response to ambient PM.
 6          Contrary to the adverse effects of inhaled concentrated ambient PM reported by Godleski
 7     and colleagues in a peer-reviewed publication on ambient PM (Godleski et al., 2000).
 8     Muggenberg and colleagues (2000) have found that exposure to high concentrations of ROFA
 9     produces no consistent changes in amplitude of the ST-segment, form of the T wave, or
10     arrhythmias in dogs. In their studies, four beagle dogs were exposed to 3  mg/m3 ROFA particles
11     generated for 3 h/day for 3 consecutive days with a Wright dust feeder. They did note that there
12     was a slight but variable decrease in heart rate, but the changes were not statistically or
13     biologically significant. The ROFA was collected from the same power plant as the Godleski
14     study but at a later time point.  The transition metal content of the ROFA  used by Muggenberg
15     was approximately 15% by mass, a value that is on the order of a magnitude higher than that
16     found in ambient urban PM samples. Although the study did not specifically address the effect
17     of metals, it suggests that inhalation of high concentrations of metals may have little effect on the
18     cardiovascular system of a healthy individual. Therefore, the different findings between the dog
19     studies illustrate the difficulties in extrapolating animal toxicology data to human health effects.
20          In a series of studies, Gordon, Nadziejko, and colleagues examined the response of the
21     rodent cardiovascular system to concentrated ambient PM derived from New York city air
22     (Gordon et al., 2000).  Particles of 0.2 to 2.5  /^m in diameter were concentrated up to 10 times
23     their levels in ambient air (-150 to 900 ,ug/m3) to maximize possible differences in effects
24     between normal and cardiopulmonary-compromised laboratory animals. ECG changes were not
25     detected in normal Fischer 344 rats or hamsters exposed to concentrated ambient PM for 1 to 3
26     days. Similarly, no deaths or ECG changes were observed in MCT rats or cardiomyopathic
27     hamsters exposed to PM.  Contrary to the decrease in heart rate observed  in dogs exposed to
28     concentrated ambient PM (Godleski et al., 2000), heart rate was increased in both normal and
29     MCT rats exposed to PM.  The increase was approximately 5% and was not observed on all
30     exposure days.  Thus, extrapolation of the heart rate changes in these animal studies to human

       March 2001                               8-32        DRAFT-DO NOT QUOTE OR CITE

-------
  1      health effects is difficult, although the increase in heart rate in rats is similar to that observed in
  2      human population studies (see Chapter 6).
  3           Gordon and colleagues (1998) have reported other cardiovascular effects in animals
  4      exposed to inhaled CAP. Increases in peripheral blood platelets and neutrophils were observed
  5      in control and MCT rats at 3 h, but not 24 h, after exposure to 150 to 400 /ug/m3 concentrated
  6      ambient PM (CAP). This neutrophil effect, likely a result of vascular demargination, did not
  7      appear to be dose related and did not occur on all exposure days, thus, suggesting that day-to-day
  8      changes in particle composition may play an important role in the systemic effects of inhaled
  9      particles.  Terashima et al. (1997) also examined the effect of particles on circulating neutrophils.
 10      They instilled rabbits with 20 mg colloidal carbon, a relatively inert particle (<1 /urn), and
 11      observed a stimulation of the release of 5'-bromo-2'deoxyuridine (BrdU)-labeled PMNs from the
 12      bone marrow at 2 to 3 days after instillation. Because the instilled supernatant from rabbit AMs
 13      treated in vitro with colloidal carbon also stimulated the release of PMNs from the bone marrow,
 14      they hypothesized that cytokines released from activated macrophages could be responsible for
 15      this systemic effect.
 16           The results of epidemiology studies have suggested that homeostatic changes in the
 17      vascular system can occur after episodic exposure to ambient PM.  Ohio et al. (2000a) have
 18      shown that inhalation of concentrated PM in healthy nonsmokers causes increased levels of
 19      blood fibrinogen. They exposed 38 volunteers exercising intermittently at moderate levels of
20      exertion for 2 h to either filtered air or particles concentrated from the air in Chapel Hill, NC (23
21      to 311 ug/m2). Blood obtained 18 h after exposure contained significantly more fibrinogen than
22      blood obtained before exposure. The observed effects in blood may associated with the mild
23      inflammation also found 18 h after exposure to CAP (see Section 8.2.3).
24           Gardner et al. (2000) examined whether the instillation of particles would alter blood
25      coagulability factors in laboratory animals.  Sprague-Dawley rats were instilled with 0.3, 1.7, or
26      8.3 mg/kg of ROFA or 8.3  mg/kg Mount Saint Helens volcanic ash. They observed an increase
27      in plasma fibrinogen in healthy rats.  Because fibrinogen is a known risk factor for ischemic heart
28      disease and stroke, the authors suggested that this alteration in the coagulation pathway could
29      take part in the triggering of cardiovascular events in susceptible individuals. Elevations in
30      plasma fibrinogen, however, were observed in healthy rats only at the  highest treatment dose, and
31      no other changes in clotting function were noted. Because the lower treatment doses are known

        March 2001                                8-33        DRAFT-DO NOT QUOTE OR CITE

-------
 1     to cause pulmonary injury and inflammation, albeit to a lower extent, the absence of plasma
 2     fibrinogen changes at these lower doses suggests that only high levels of pulmonary injury are
 3     able to produce an effect in healthy test animals.
 4           In summary, controlled animal studies have provided initial evidence that high
 5     concentrations of inhaled or instilled particles can have systemic, especially cardiovascular,
 6     effects, hi the case of MCT rats, these effects can be lethal. Understanding the pathways by
 7     which very small concentrations of inhaled ambient PM can produce systemic, life-threatening
 8     changes, however, is far from clear. Among the hypotheses that have been proposed to account
 9     for the nonpulmonary effects of PM are activation of neural reflexes, cytokine effects on heart
10     tissue (Killingsworth et al., 1997), alterations in coagulability (Seaton et al., 1995; Sjogren,
11     1997), and perturbations in homeostatic processes such as heart rate or heart rate variability
12     (Watkinson et al., 1998).  A great deal of research using controlled exposures of animal and
13     human  subjects to PM will be necessary to test mechanistic hypotheses generated to date, as well
14     as those that are likely to be proposed in the future.
15
16
17     8.4  SUSCEPTIBILITY TO THE  EFFECTS OF PARTICIPATE
18           MATTER EXPOSURE
19           Susceptibility of an  individual to adverse health effects of PM can vary depending on a
20     variety  of host factors such as age, nutritional status, physiological activity profile, genetic
21     predisposition, or preexistent disease.  The potential for preexistent disease to alter adverse
22     response to toxicant exposure is widely acknowledged but poorly understood.  Because of
23     inherent variability (necessitating large numbers of subjects) and ethical concerns associated with
24     using diseased subjects in clinical research studies, a solid database on human susceptibilities is
25     lacking. For more control over both host and environmental variables, animal models often are
26     used. However, care must be taken in extrapolation from animal models of human disease to
27     humans. Rodent models of human disease, their use in toxicology and the criteria for judging
28     their appropriateness as well as their limitations must be considered (Kodavanti et al., 1998b;
29     Kodavanti and Costa, 1999).
30
        March 2001                               8-34        DRAFT-DO NOT QUOTE OR CITE

-------
  1      8.4.1 Effects of Particulate Matter on Cardiopulmonary Compromised Hosts
  2           Epidemiological studies suggest there may be subsegments of the population that are
  3      especially susceptible to effects from inhaled particles (see Chapter 6). The elderly with chronic
  4      cardiopulmonary disease, those with pneumonia and possibly other lung infections, and those
  5      with asthma (at any age) appear to be at higher risk than healthy people of similar age.
  6      Unfortunately, most toxicology studies have used healthy adult animals. An increasing number
  7      of newer studies have examined effects of ambient particles in compromised host models. Costa
  8      and Dreher (1997) used a rat model of cardiopulmonary disease to explore the question of
  9      susceptibility and the possible mechanisms by which PM  effects are potentiated. Rats with
 10      advanced monocrotaline (MCT)-induced pulmonary vasculitis/hypertension were given
 11      intratracheal instillations of ROFA (0, 0.25, 1.0, and 2.5 mg/rat). The MCT animals had a
 12      marked neutrophilic inflammation.  In the context of this inflammation, ROFA induced a four- to
 13      fivefold increase in BAL PMNs. There was increased mortality at 96 h that was ROFA-dose
 14      dependent. The results of this study indicate that particles, albeit at a high concentration,
 15      enhanced the neutrophilic inflammation and mortality in MCT animals.
 16           Kodavanti et al. (1999) also studied PM effects in the MCT rat model of pulmonary
 17      disease.  Rats treated with 60 mg/kg MCT were exposed to 0, 0.83. or 3.3 mg/kg ROFA by
 18      intratracheal instillation and to 15 mg/kg ROFA by inhalation.  Both methods of exposure caused
 19      inflammatory lung responses and ROFA exacerbated the lung lesions, as shown by increased
 20      lung edema, inflammatory cells, and alveolar thickening.
 21           The manner in which MCT can alter the response of rats to inhaled particles was examined
 22      by Madl and colleagues (1998). Rats were exposed to fluorescent colored microspheres (1 //m)
 23      2 weeks after treatment with MCT.  In vivo phagocytosis of the microspheres was altered in the
 24      MCT rats in comparison with control animals. Fewer microspheres were phagocytized in vivo
 25      by alveolar macrophages and there was a concomitant increase in free microspheres overlaying
26      the epithelium at airway bifurcations.  The decrease in in vivo phagocytosis was not accompanied
27      by a similar decrease in vitro.  Macrophage chemotaxis, however, was impaired significantly in
28      MCT rats compared with control rats.  Thus, MCT appeared to impair particle clearance from the
29      lungs via inhibition of macrophage chemotaxis.
30          The sulfur dioxide (SO2)-induced model of chronic bronchitis has also been used to
31      examine the potential interaction of PM with preexisting lung disease.  Clarke and colleagues
        March 2001                              8-35        DRAFT-DO NOT QUOTE OR CITE

-------
 1     pretreated Sprague Dawley rats for 6 weeks with air or 170 ppm SO2 for 5 h/day and 5 days/week
 2     (Clarke et al., 1999). Exposure to concentrated air particles for 5 h/day for 3 days at an average
 3     concentration of 515 Mg/m3 produced changes in pulmonary function as evidenced by significant
 4     increases in tidal volume in both air- and SO2-pretreated rats. Exposure to concentrated ambient
 5     PM also produced significant changes in both cellular and biochemical markers in lavage fluid.
 6     In comparison to control animal values, protein was increased approximately threefold in SO2-
 7     pretreated animals exposed to concentrated ambient PM. Lavage fluid neutrophils and
 8     lymphocytes were increased significantly in both pretreatment groups of rats exposed to
 9     concentrated ambient PM, with greater increases in both cell types in the SO2-pretreated rats.
10     Thus, exposure to concentrated ambient PM produced adverse changes in the respiratory system,
11     but no deaths, in both normal rats and in a rat model of chronic bronchitis.
12          Clarke et al. (2000) next examined the effect of concentrated ambient PM in normal rats of
13     different ages. Unlike the  earlier study that used Sprague-Dawley rats, 4- and 20-mo-old Fischer
14     344 were examined after 3 days of exposure to concentrated ambient PM.  They found that
15     exposure to daily mean concentrations of 80, 170, and 50 /ug/m3 PM produced statistically
16     significant increases in total neutrophil counts (up over 10-fold) in lavage fluid of the young, but
17     not the old, rats. Thus, repeated exposure to relatively low concentrations of ambient PM
18     produced an inflammatory response, although the actual percent neutrophils in the concentrated
19     ambient PM-exposed young rats was low (approximately 3%).  On the other hand, Gordon and
20     colleagues found no evidence of neutrophil influx in the lungs of normal and monocrotaline-
21     treated Fischer 344  rats exposed in nine separate experiments to concentrated ambient PM
22     (Gordon et al., 2000) as  high as 400 /u,g/m3 for a 6-h exposure or 192 /ug/m3 for three daily 6-h
23     exposures. Similarly, normal and cardiomyopathic hamsters showed no evidence of pulmonary
24     inflammation or injury after a single exposure to concentrated ambient PM. Gordon and
25     colleagues did report a statistically significant doubling in protein concentration  in lavage fluid in
26     monocrotaline-treated rats exposed for 6 h to 400 /ug/m3 concentrated ambient PM. Because of
27     the disparity in findings  in the response of normal Fischer 344 rats to concentrated ambient PM
28     between these two labs,  it is important that the reproducibility of these experiments be examined.
29          Kodavanti and colleagues (1998b) also have examined the effect of concentrated ambient
30     PM in normal rats and rats with sulfur dioxide-induced chronic bronchitis. In four separate
31     exposures to PM, there was a significant increase in lavage fluid protein in bronchitic rats from

       March 2001                               8-36        DRAFT-DO NOT QUOTE OR CITE

-------
  1      only one exposure protocol in which the rats were exposed to 444 and 843 /ug/m3 PM on
  2      2 consecutive days (6 h/day). Neutrophil counts were increased in bronchitic rats exposed to
  3      concentrated ambient PM in three of the four exposure protocols, but was decreased in the fourth
  4      protocol. No other changes in normal or bronchitic rats were observed, even in the exposure
  5      protocols with higher PM concentrations. Thus, rodent studies have demonstrated that
  6      inflammatory changes can be produced in normal and compromised animals exposed to
  7      concentrated ambient PM. These findings are important because only a limited number of
  8      studies have used real-time inhalation exposures to actual ambient urban PM.
  9          Pulmonary function measurements are often less invasive than other means to assess the
 10      effects of inhaled air pollutants on the mammalian lung.  Although the publication of the 1996
 11       PM AQCD, a number of investigators have examined the response of rodents and dogs to inhaled
 12       ambient particles. In general, these investigators have demonstrated that ambient PM has
 13       minimal effects on pulmonary function tests. Gordon et al. (2000) exposed normal and
 14       monocrotaline-treated rats to filtered air or 181 //g/m3 concentrated ambient PM for 3 h.
 15       For both normal and monocrotaline-treated rats, no differences in lung volume measurements or
 16       diffusion capacity for carbon monoxide were observed between the air or PM exposed animals at
 17       3 or 24 h after exposure.  Similarly, in cardiomyopathic hamsters, concentrated ambient PM had
 18      no effect on these same pulmonary function measurements.
 19           In an examination of the effect of concentrated ambient PM on airway responsiveness in
 20      mice, Goldsmith and colleagues (1999) exposed control and ovalbumin-sensitized mice to an
 21      average concentration of 787 /^g/m3 PM for 6 h/day for 3 days. Although ovalbumin
 22      sensitization itself produced an increase in the nonspecific airway responsiveness to inhaled
 23      methylcholine, concentrated ambient PM did not change the response to methylcholine in
 24      ovalbumin-sensitized or control mice. For comparison, these investigators examined the effect
 25      of inhalation of an aerosol of the active soluble fraction of ROFA on control and ovalbumin-
 26      sensitized mice and demonstrated that ROFA could produce nonspecific airway
 27      hyperresponsiveness to methylcholine in both control and ovalbumin-sensitized mice. Similar
 28      increases in airway responsiveness have been observed after exposure to ROFA in normal and
29      ovalbumin-sensitized rodents (Gavett et al., 1997, 1999; Hamada et al., 1999, 2000).  Other
30      pulmonary function endpoints have been studied in animals exposed to concentrated ambient
31      PM. Clarke et al. (1999) observed that tidal volume was increased slightly in both control rats

        March 2001                              8-37        DRAFT-DO NOT QUOTE OR CITE

-------
 1      and rats with sulfur dioxide-induced chronic bronchitis exposed to 206 to 733 /ug/m3 PM on
 2      3 consecutive days. No changes in peak expiratory flow, respiratory frequency, or minute
 3      volume were observed after exposure to concentrated ambient PM. In the series of dog studies
 4      by Godleski et al. (2000) (also see Section 8.3), no signficant changes in pulmonary functions
 5      were observed in normal mongrel dogs exposed to concentrated ambient PM, although a 20%
 6      decrease in respiratory frequency was observed in dogs that underwent coronary artery occlusion
 7      and were exposed to PM. Thus, studies using normal and compromised animal models exposed
 8      to concentrated ambient PM have found minimal biological effects of ambient PM on pulmonary
 9      function.
10          In studying the influence of age on susceptibility to PM, Johnston et al. (1998) exposed
11      8-week-old mice (young) and 18-mo-old mice (old) to polytetrafluoroethylene fumes (PTFE)
12      (0, 10, 25, and 50 yUg/m3) for 30 min. Lung lavage endpoints (PMN, protein, LDH, and
13      p-glucuronidase) as well as lung tissue mRNA levels for various cytokines, metallothionein and
14      for Mn superoxide dismutase were measured 6 h following exposure. Protein, lymphocyte,
15      PMN,  and TNF-a mRNA levels were increased in older mice when compared to younger mice.
16      These  findings suggest that the inflammatory response to PTFE fumes is altered with age, being
17      greater in the older animals. Although  Teflon particles are not  a valid surrogate for ambient
18      ultrafme particles (Oberdorster et al., 1992), this study did provide evidence to support the
19      hypothesis  that particle-induced pulmonary inflammation is different between young and old
20      organisms.
21          Kodavanti et al. (2000b; 2001) used genetically predisposed spontaneously hypertensive
22      (SH) rats as a model of cardiovascular disease to study PM-related susceptibility. The SH rats
23      were found to be more susceptible to acute pulmonary injury from intratracheal ROFA exposure
24      than normotensive control Wistar Kyoto (WKY) rats (Kodavanti et al., 2001).  The primary
25      metal constituents of ROFA, V and Ni, caused differential species-specific effects. Vanadium,
26      which  was  less toxic than Ni in both strains, caused inflammatory responses only in WKY rats,
27      whereas Ni was injurious to both WKY and SH rats (SH > WKY). This differential
28      responsiveness of V and Ni was  correlated with their specificity for airway and parenchymal
29      injury, discussed in another study (Kodavanti et al., 1998b). When exposed to the same ROFA
30      by inhalation, SH rats were more sensitive than WKY rats in regards to vascular leakage
31      (Kodavanti et al., 2000b).  The SH rats exhibited a hemorrhagic response to ROFA. Oxidative

        March 2001                               8-38        DRAFT-DO NOT QUOTE OR CITE

-------
  1      stress was much higher in ROFA exposed SH rats than matching WKY rats. Also, SH rats,
  2      unlike WKY rats, showed a compromised ability to increase BALF glutathione in response to
  3      ROFA, suggesting a potential link to increased susceptibility.  Cardiovascular effects were
  4      characterized by ST-segment area depression of the ECG in ROFA-exposed SH but not WKY
  5      rats.  These studies demonstrate the potential utility of cardiovascular disease models for the
  6      study of PM health effects and show that genetic predisposition to oxidative stress and
  7      cardiovascular disease may play a role in sensitivity to increased PM-related cardiopulmonary
  8      injury.
  9           In summary, although these studies are just emerging and are only now being replicated or
 10      followed more thoroughly to investigate the mechanisms, they do provide evidence of enhanced
 11       susceptibility to inhaled PM in "compromised" hosts.
 12
 13       8.4.2 Genetic Susceptibility to Inhaled Particles
 14            A key question in understanding the adverse health effects of inhaled PM is which
 15       individuals are susceptible to PM.  Although factors such as age and health status have been
 16       studied in both epidemiology and toxicology studies, a number of investigators have begun to
 17       examine the importance of genetic susceptibility in  the response to inhaled particles because of
 18       considerable evidence that genetic factors play a role in the response to  inhaled pollutant gases.
 19       To accomplish this goal, investigators typically have studied the interstrain response to particles
 20      in rodents.  The response to ROFA instillation in different strains of rats has been investigated by
 21      Kodavanti et al. (1996, 1997a). In the first study, male Sprague Dawley (SD) and Fischer-344
 22      (F-344) rats were instilled intratracheally with saline or ROFA particles. ROFA instillation
 23      produced an increase in lavage fluid neutrophils in both SD and F-344 rats, whereas a time-
 24      dependent increase in eosinophils occurred only in SD rats. In a subsequent study (Kodavanti
 25      et al., 1997a), SD, Wistar (WIS), and F-344 rats (60 days old) were exposed to saline or ROFA
 26      (8.3 mg/kg) by intratracheal instillation and examined for up to 12 weeks. Histology indicated
 27      focal areas of lung damage showing inflammatory cell infiltration as well as  alveolar, airway, and
 28      interstitial thickening in all three rat strains during the week following exposure. Trichrome
29      staining for fibrotic changes indicated a sporadic incidence of focal alveolar fibrosis at 1, 3, and
30      12 weeks in SD rats, whereas WIS and F-344 rats showed only a modest increase in trichrome
31      staining in the septal areas. One of the isoforms of fibronectin mRNA was upregulated in
        March 2001                                8-39        DRAFT-DO NOT QUOTE OR CITE

-------
  1      ROFA-exposed SD and WIS rats, but not in F-344 rats.  Thus, there appeared to be a rat strain-
  2      dependent variability in the fibrotic response to instilled ROFA.
  3           Kleeberger and colleagues have examined closely the role that genetic susceptibility plays
  4      in the effect of inhaled acid-coated particles on macrophage function. Nine inbred strains of
  5      mice were exposed nose-only to acid-coated particles (10 mg/m3 with 285 yUg/m3 sulfate) for 4 h
  6      (Yoshinori et al., 2000). Significant inter-strain differences in Fc-receptor-mediated macrophage
  7      phagocytosis were observed, with C57BL/6J mice being the most sensitive. Although neutrophil
  8      counts were increased more  in C3H/HeOuJ and C3H/HeJ strains of mice than in the other
  9      strains, the overall magnitude of change was small and not correlated with the changes in
10      macrophage phagocytosis. In follow-up studies, Ohtsuka et al. (2000a,b) performed a genome-
11      wide scan with a intercross cohort derived from C57BL/6J and C3H/HeJ mice.  Analyses of
12      macrophage dysfunction phenotypes of segregant  and nonsegregant populations derived from
13      these two strains indicate that two unlinked genes control susceptibility. They identified a
14      3-centiMorgan segment on mouse chromosome 17 that contains an acid-coated particle
15      susceptibility loci. Interestingly, this quantitative  trait loci overlaps with those described for
16      ozone-induced inflammation (Kleeberger et al., 1997) and acute lung injury (Prows et al.,  1997)
17      and contains several  promising candidate genes that may be responsible for the observed genetic
18      susceptibility for macrophage dysfunction in mice exposed to  acid-coated particles.
19           Only one study has examined the interstrain susceptibility to ambient particles. C57BL/6J
20      and C3H/HeJ mice were exposed to 250 /^g/m3 concentrated ambient PM for 6 h and examined
21      at 0 and 24 h after exposure for changes in lavage fluid parameters and cytokine mRNA
22      expression in lung tissue (Shukla et al., 2000). No interstrain differences in response were
23      observed.  Surprisingly, although no indices of pulmonary inflammation or injury were increased
24      over control values in the lavage fluid, increases in cytokine mRNA expression were observed in
25      both murine strains exposed  to PM.  Although the increase in cytokine mRNA expression was
26      generally small (approximately twofold), the effect on IL-6, TNF-cc, TGF-P2, and y-interferon
27      was consistent and replication of this study is necessary.
28           Thus, a handful of studies have begun to demonstrate that genetic susceptibility can play a
29      role in the response to inhaled particles. Similar strain differences in response to inhaled metal
30      particles have been observed by other investigators (McKenna et al., 1998; Wesselkamper et al.,
31      2000), although the concentration of metals used in these studies is more relevant to occupational

        March 2001                               8-40        DRAFT-DO NOT QUOTE OR CITE

-------
  1     rather than environmental exposure levels.  It remains to be determined whether genetic
  2     susceptibility plays as significant a role in the adverse effects of ambient PM as does age or
  3     health status.
  4
  5     8.4.3 Effect of Particulate Matter on Allergic Hosts
  6           Relatively little is known about the effects of inhaled particles on humoral (antibody) or
  7     cell-mediated immunity. Alterations in the response to a specific antigenic challenge have been
  8     observed in animal models at high concentrations of acid sulfate aerosols (above 1,000 Aig/m3)
  9     (Pinto et al., 1979; Kitabatake et al., 1979; Fujimaki et al.,  1992). Several studies have reported
 10     an enhanced response to nonspecific bronchoprovocation agents, such as acetylcholine and
 11      histamine, after exposure to inhaled particles. This nonspecific airway hyperresponsiveness,
 12     a central feature of asthma, occurs in animals and human subjects exposed to sulfuric acid under
 13      controlled conditions (Gearhart and Schlesinger,  1986; Utell et al.,  1983). Although, its
 14      relevance to specific allergic responses in the airways of atopic individuals is unclear, it
 15      demonstrates that the airways of asthmatics may become sensitized to either specific or
 16      nonspecific triggers that could result in increases  in asthma severity and asthma-related hospital
 17      admissions (Peters et al., 1997; Jacobs et al, 1997; Lipsett et al., 1997).
 18           Nel et al. (1998) have suggested that the rise in the U.S. prevalence rate for allergic rhinitis
 19      (5% in the 1950s to about 20% in the 1980s) may be related to increased diesel particulate matter
 20      (DPM),  in addition to other combustion related PM.  Combustion particles also may serve as
 21      carrier particles for allergens (Knox et al., 1997).
 22           A number of in vivo and in vitro studies have demonstrated that DPM can alter the immune
 23      response to challenge with specific antigens and suggest that DPM may act as an adjuvant.
 24      These studies have shown that treatment with DPM enhances the secretion of antigen-specific
 25      IgE in mice (Takano et al., 1997) and in the nasal cavity of human subjects (Diaz-Sanchez et al.,
 26      1996, 1997; Ohtoshi et al., 1998). Because IgE levels play a major role in allergic asthma
 27      (Wheatley and Platts-Mills, 1996), upregulation of its production could lead to an increased
28      response to inhaled antigen in particle-exposed individuals.
29           Only a small number of studies have examined the mechanisms underlying the
30      enhancement of allergic asthma by ambient urban particles.  Ohtoshi et al. (1998) reported that a
31      coarse size-fraction of resuspended ambient PM, collected in Tokyo, induced the production of
        March 2001                                8-41         DRAFT-DO NOT QUOTE OR CITE

-------
  1      granulocyte macrophage colony stimulating factor (GMCSF), an upregulator of dendritic cell
  2      maturation and lymphocyte function, in human airway epithelial cells in vitro.  In addition to
  3      increased GMCSF, epithelial cell supernatants contained increased IL-8 levels  when incubated
  4      with DPM, a principal component of ambient particles collected in Tokyo.  Although the sizes of
  5      the two types of particles used in this study were not comparable, the results suggest that ambient
  6      PM, or at least the DPM component of ambient PM, can upregulate the immune response to
  7      inhaled antigen through GMCSF production. Similarly, Takano et al. (1998) has reported airway
  8      inflammation, airway hyperresponsiveness, and increased GM-GSF and IL-5 in mice exposed to
  9      diesel exhaust.
 10           Gavett et al. (1999) have investigated the effects of ROFA (intratracheal instillation) in
 11      ovalbumin (OVA) sensitized and challenged mice. Instillation of 3 mg/kg (approximately 60 /ug)
 12      ROFA induced inflammatory and physiological responses in the OVA mice that were related to
 13      increases in Th2 cytokines (IL-4, IL-5). ROFA induced greater than additive increases in
 14      eosinophil numbers and in airway responsiveness to methylcholine.
 15           Hamada et al. (1999, 2000) have examined the effect of a ROFA leachate aerosol in a
 16      neonatal mouse model of allergic asthma. In the first study, neonatal mice sensitized by ip
 17      injection with OVA developed airway hyperresponsiveness, eosinophilia, and elevated serum
 18      anti-ovalbumin IgE after a challenge with inhaled OVA. Exposure to the ROFA leachate aerosol
 19      had no marked effect on the airway responsiveness to inhaled methacholine in nonsensitized
20      mice, but did enhance the airway hyperresponsiveness to methylcholine produced in
21      OVA-sensitized mice. No other interactive effects of ROFA exposure with OVA were observed.
22      In a subsequent study, Hamada et al. clearly demonstrated that, whereas inhaled OVA alone was
23      not sufficient to sensitize mice to a subsequent inhaled OVA challenge, pretreatment with a
24      ROFA leachate aerosol prior to the initial exposure to aerosolized OVA resulted in an allergic
25      response to the inhaled OVA challenge. Thus, exposure to a ROFA leachate aerosol can alter the
26      immune response to inhaled OVA both at the sensitization stage at an early age and at the
27      challenge stage.
28          Lambert et al. (1999) also examined the effect of ROFA on a rodent model of pulmonary
29      allergy. Rats were instilled intratracheally with 200 or 1,000 f^g ROFA 3 days prior to
30      sensitization with house dust mite antigen.  HDM sensitization after 1000 /^g ROFA produced
31      increased eosinophils, LDH, BAL protein, and IL-10 relative to HDM alone. The immediate

        March 2001                               8-42        DRAFT-DO NOT QUOTE OR CITE

-------
  1     bronchoconstrictive and associated antigen-specific IgE response to a subsequent antigen
  2     challenge was increased in the ROFA-treated group in comparison with the control group.
  3     Together, these studies suggest the components of ROFA can augment the immune response to
  4     antigen. Evidence that metals are responsible for the ROFA-enhancement of an allergic
  5     sensitization was demonstrated by Lambert et al. (2000).  In this follow-up study, Brown Norway
  6     rats were instilled with 1  mg ROFA or the three main metal components of ROFA (iron,
  7     vanadium, or nickel) prior to sensitization with instilled house dust mite.  The three individual
  8     metals were found to augment different aspects of the immune response to house dust mite.
  9     Nickel and vanadium produced an enhanced immune response to the antigen as seen by higher
 10     house dust mite-specific IgE serum levels after an antigen challenge at  14 days after sensitization.
 11     Nickel and vanadium also produced an increase in the lymphocyte proliferative response to
 12     antigen in vitro. In addition, the antigen-induced bronchoconstrictive response was greater only
 13     in nickel-treated rats.  Thus, instillation of metals at concentrations equivalent to those present in
 14     the ROFA leachate mimicked the  response to ROFA, suggesting that the metal components of
 15     ROFA are responsible for the increased allergic sensitization observed  in ROFA-treated animals.
 16          Goldsmith et al. (1999) have compared the effect of inhalation of concentrated ambient PM
 17     for 6 h/day for 3 days versus the effect of a single exposure to a ROFA  leachate aerosol on the
 18     airway responsiveness to methylcholine in OVA-sensitized mice. Daily exposure to ROFA
 19     leachate aerosols significantly enhanced the airway hyperresponsiveness in OVA-sensitized
20     mice. Importantly,  exposure to concentrated ambient PM (average concentration of 787 ,wg/m3)
21     had no effect on airway responsiveness in six separate experiments.  Thus, the effect of the
22     ROFA leachate aerosols on the induction of airway hyperresponsiveness in allergic mice was
23     significantly different than that of a high concentration of concentrated  ambient PM. Although
24     airway responsiveness was examined at only one postexposure time point, these findings do
25     suggest that a great  deal of caution should be used in interpreting the results of studies using
26     ROFA particles or leachates in the attempt to investigate the biologic plausibility of the adverse
27     health effects of PM.
28          Several other studies have examined in greater detail the contribution to allergic asthma of
29     the particle component and the organic fraction of DPM. Tsien et al.  (1997) treated transformed
30     IgE-producing human B lymphocytes in vitro with the organic extract of DPM. The organic
31      phase extraction had no effect on cytokine production but did increase IgE production.

        March 2001                               8-43        DRAFT-DO NOT QUOTE OR CITE

-------
 1     Moreover, these experiments determined that DPM appeared to be acting on cells already
 2     committed to IgE production, thus suggesting a mechanism by which the organic fraction of
 3     combustion particles can directly affect B cells and influence human allergic asthma.
 4           Cultured epithelial cells from atopic asthmatics show a greater response to DPM exposure
 5     when compared with cells from nonatopic nonasthmatics. IL-8, GM-CSF, and soluble ICAM-1
 6     increased in response to DPM at a concentration of 10 //g/mL DPM (Bayram et al., 1998a,b).
 7     This study suggests that particles could modulate airway disease through their actions on airway
 8     epithelial cells. This study also suggests that bronchial epithelial cells from asthmatics are
 9     different from those of nonasthmatics in regard to their mediator release in response to DPM.
10           Sagai and colleagues (1996) repeatedly instilled mice with DPM for up to 16 weeks and
11     found increased numbers of eosinophils, goblet cell hyperplasia, and nonspecific  airway
12     hyperresponsiveness, changes which are central features of chronic asthma (National Institutes of
13     Health, 1997). Takano et al. (1997) extended this line of research and examined the effect of
14     repeated instillation of DPM on the specific response to antigen (OVA) in mice.  They observed
15     that antigen-specific IgE and IgG levels were significantly greater in mice repeatedly instilled
16     with both DPM and OVA. Because this upregulation in antigen-specific immunoglobulin
17     production was not accompanied by an increase in inflammatory cells or cytokines in lavage
18     fluid, it would suggest that, in vivo, DPM may act directly on immune system cells, as described
19     in the work by Tsien et al. (1997). Animal studies have confirmed that the adjuvant activity of
20     DPM also applies to the sensitization of Brown Norway rats to timothy grass pollen (Steerenberg
21     etal., 1999).
22           Diaz-Sanchez and colleagues (1996) have continued to study the mechanism of DPM-
23     induced upregulation of allergic response in the nasal cavity of human subjects. In one study, a
24     200 juL aerosol bolus containing 0.15 mg of DPM was delivered into each naris of subjects with
25     or without seasonal allergies,  hi addition to increases in IgE in nasal lavage  fluid (NAL), they
26     found an enhanced production of IL-4, IL-6, and IL-13, cytokines known to be B  cell
27     proliferation factors.  The levels of several  other cytokines also were increased, suggesting a
28     general inflammatory response to a nasal challenge with DPM.  In a following study, these
29     investigators delivered ragweed antigen, alone or in combination with DPM, on two occasions, to
30     human subjects with both allergic rhinitis and positive skin tests to ragweed  (Diaz-Sanchez et al.
31     1997).  They found that the combined challenge with ragweed antigen and DPM produced

       March 2001                               8-44        DRAFT-DO NOT QUOTE OR CITE

-------
  1      significantly greater antigen-specific IgE and IgG4 in NAL. A peak response was seen at 96 h
  2      postexposure.  The combined treatment also induced expression of IL-4, IL-5, IL-10, and IL-13,
  3      with a concomitant decrease in expression of Thl-type cytokines.  Although the treatments were
  4      not randomized (antigen alone was given first to each subject), the investigators reported that
  5      pilot work showed no interactive effect of repeated antigen challenge on cellular and biochemical
  6      markers in NAL.  DPM also resulted in the nasal influx of eosinophils, granulocytes, monocytes,
  7      and lymphocytes,  as well as the production of various inflammatory mediators.  The combined
  8      DPM plus ragweed exposure did not increase the rhinitis symptoms beyond those of ragweed
  9      alone.
 10           Blomberg et al. (1998) observed a 10-fold increase in NAL fluid ascorbate concentration
 11      after a 1-h exposure to diluted diesel exhaust (300 /ug/m3 particles  and 1.6 ppm NO2). However,
 12      there were no effects on BAL ascorbate levels. Rudell et al. (1990) had previously shown
 13      increased BAL neutrophils in nonsmoking subjects exposed to 100 /^g/m3 of DPM in diesel
 14      exhaust (gases were present). Thus, diesel exhaust (particles and gases) can produce an enhanced
 15      response to antigenic material in the nasal cavity. Extrapolation of these findings, of enhanced
 16      allergic response in the nose, to the lung, would suggest that ambient combustion particles
 17      containing DPM may have significant  effects on allergic asthma. These studies provide
 18      biological plausibility for the exacerbation of allergic asthma associated with episodic exposure
 19      to PM. Although DPM may make up only a fraction of the mass of urban PM, because of their
 20      small size, DPM may represent a significant fraction of the ultrafine particle mode in urban air,
 21      especially in cities and countries that rely heavily on diesel-powered vehicles. It must be noted
 22      that the potential contribution of DPM to the rising prevalence in asthma is  complicated by the
 23      fact that DPM levels have been decreasing over the last decade (CALEPA report). The reported
 24      decrease in DPM levels is a result of the increased combustion efficiency of diesel engines.  This
 25      improvement in diesel engine design also has brought about a significant decrease in the particle
 26      size of diesel emissions. Thus, the balance between a decrease in diesel emissions and the
 27      production of a potentially more toxic particle size needs further exploration.
28
29      8.4.4 Resistance to Infectious Disease
30           The development of an infectious disease requires both the presence of the appropriate
31      pathogen, as well as host susceptibility to the pathogen. There are  numerous specific and
        March 2001                               8-45        DRAFT-DO NOT QUOTE OR CITE

-------
  1      nonspecific anti-microbial host defenses against microbes, and the ability of inhaled particles to
  2      modify resistance to bacterial infection could result from a decreased ability to clear or kill
  3      microbes.  Rodent infectivity models frequently have been used to examine the effect of inhaled
  4      particles on host defense and infectivity. Mice or rats are challenged with a bacterial or viral load
  5      either before or after exposure to the particles (or gas) of interest; mortality rate, survival time, or
  6      bacterial clearance are then examined.  A number of studies that have used the infectivity model
  7      to assess the effect of inhaled PM were discussed previously (U.S. Environmental Protection
  8      Agency, 1982, 1989, 1996a). In general, acute exposure to sulfuric acid aerosols at
  9      concentrations up to 5,000 /ug/m3  were not very effective in  enhancing mortality in a bacterially
10      mediated murine model. In rabbits, however, sulfuric acid aerosols altered anti-microbial
11      defenses after exposure for 2 h/day for 4 days to 750 //g/m3 (Zelikoff et al., 1994). Acute or
12      short-term repeated exposures to high concentrations of relatively inert particles have produced
13      conflicting results.  Carbon black  (10,000 /wg/m3) was found to have no effect on susceptibility to
14      bacterial infection (Jakab, 1993), whereas a very high concentration of TiO2 decreased the
15      clearance of microbes and the bacterial response of lymphocytes isolated from mediastinal lymph
16      nodes (Gilmour et al., 1989a,b). In addition, exposure to DPM has been shown to enhance the
17      susceptibility of mice to the lethal effects of some, but not all,  microbial agents (Hatch et al.,
18      1985; Hahon et al.,  1985). Thus, the pulmonary response to microbial agents has been shown to
19      be altered at relatively high particle concentrations in animal models. Moreover, these effects
20      appear to be highly dependent on the microbial challenge and the test animal studied. Pritchard
21      et al. (1996) observed in rats exposed to particles with a high concentration of metals (e.g.,
22      ROFA), that the increased mortality rate after streptococcus  infection was associated with the
23      amount of metal in the PM.
24           Despite the reported association between ambient PM  and deaths caused by pneumonia
25      (Schwartz, 1994), there are few recent studies that have examined the mechanisms that may be
26      responsible for the effect of PM on infectivity.  In one study, Cohen and colleagues (1997)
27      examined the effect of inhaled vanadium (V) on immunocompetence.  Healthy rats were
28      repeatedly exposed to 2 mg/m3 V, as ammonium metavanadate, and then instilled with
29      polyinosinic-polycytidilic acid (poly I:C), a double-stranded polyribonucleotide that acts as a
30      potent immunomodulator. Induction of increases in lavage fluid protein and neutrophils was
31      greater in animals preexposed to V.  Similarly,  IL-6 and interferon-gamma were increased in

        March 2001                               8-46         DRAFT-DO NOT QUOTE  OR CITE

-------
  1     V-exposed animals. Alveolar macrophage function, as determined by zymosan-stimulated
  2     superoxide anion production and by phagocytosis of latex particles, was depressed to a greater
  3     degree after poly I:C instillation in V-exposed rats as compared to filtered air-exposed rats.
  4     These findings provide evidence that inhaled V, a trace metal found in combustion particles and
  5     shown to be toxic  in vivo in studies using instilled or inhaled ROFA (Dreher et al., 1997;
  6     Kodavanti et al., 1997b, 1999), has the potential to inhibit the pulmonary response to microbial
  7     agents. It must be taken into consideration that these effects were found at very high
  8     occupational exposure concentrations of V, and as with many studies, care must be taken in
  9     extrapolating the results to the ambient exposure of healthy individuals or those with preexisting
 10     cardiopulmonary disease to trace concentrations (approximately 3 orders of magnitude lower
 11      concentration) of metals in ambient PM.
 12
 13
 14      8.5  MECHANISMS OF PARTICIPATE  MATTER TOXICITY AND
 15           PATHOPHYSIOLOGY:  IN VITRO EXPOSURES
 16      8.5.1  Introduction
 17           The mechanisms that underlie injury from PM exposure are unclear.  Section 8.5.2
 18      discusses the more recently published in vitro studies  that provide an approach toward
 19      identifying potential mechanisms by which PM mediates health effects. The remaining sections
 20      discuss potential mechanisms in relation to PM characteristics based on these available data.
 21
 22      8.5.2 Experimental Exposure Data
 23           In vitro exposure is a useful technique to provide information on potential hazardous PM
 24      constituents and mechanisms of PM injury, especially when only limited quantities of the test
 25      material are available. Exposing respiratory cells to particles in vitro not only reduces the
 26      amount of material needed for the experiments but also provides an opportunity to investigate the
 27      mechanisms of particle toxicity.  In addition, in vitro exposure allows the examination of the
 28      response to particles in only one or two cell types. Limitations of in vitro studies include
29      difficulty in extrapolating dose-response relationships and from in vitro to in vivo biological
30      response and mechanistic extrapolations. In addition to alterations in physiochemcial
31      characteristics of PM because of the collection and resuspension processes, these exposure
        March 2001                               8-47         DRAFT-DO NOT QUOTE OR CITE

-------
  1      conditions do not simulate the air-cell interface that actually exists within the lungs, and, thus,
  2      the exact dosage delivered to target cells is not known.  Furthermore, unless an in vitro exposure
  3      system that is capable of delivering particles uniformly to monolayers of airway epithelial cells
  4      cultured in an air-liquid interface system is used (Chen et al., 1993), the conventional incubation
  5      system alters the microenvironment surrounding the cells and may alter the mechanisms of
  6      cellular injury induced by these agents.
  7           Even with these limitations, in vitro studies do provide  an approach to identify potential
  8      cellular and molecular mechanisms by which PM mediates health effects.  These mechanisms
  9      can then be evaluated in vivo. In vitro studies are summarized in Table 8-8.
10
11      8.5.2.1 Ambient Particles
12           Several studies have exposed airway epithelial cells, alveolar macrophages, or blood
13      monocytes to ambient PM to investigate cellular processes such as oxidant generation and
14      cytokine production that may contribute to the pathophysiological response seen in vivo.  Among
15      the ambient PM being examined were samples collected from Boston (Goldsmith et al., 1998),
16      North Provo, UT (Ohio et al., 1999a,b), St. Louis, MO (SRM 1648, Dong et al., 1996; Becker
17      and Soukup, 1998), Washington, DC (SRM 1649, Becker and Soukup, 1998), Ottawa, Canada
18      (EHC-93, Becker and Soukup, 1998), Dusseldorf and Duisburg, Germany (Hitzfeld et al., 1997),
19      Mexico City (Bonner et al., 1998), and Terni, Italy (Fabiani et al., 1997).
20           Because soluble metals of ROFA have been shown to be associated with  biological effect
21      and toxicity, several studies have investigated whether the soluble components in ambient PM
22      may have the same biological activities. Ambient PM samples collected from North Provo, UT,
23      (during 1981 and 1982) were used to test whether the soluble components  or ionizable metals,
24      which accounted for approximately 0.1% of the mass, are responsible for the biological activity
25      of ambient PM. The oxidant generation (thiobarbituric acid reactive products), release of IL-8
26      from BEAS-2B cells, and PMN influx in rats exposed to these samples correlated with sulfate
27      content and the ionizable fraction of these PM samples  (Ohio et al., 1999a,b). In addition, these
28      particles stimulated IL-6 and IL-8 production as well as increased IL-8 mRNA and enhanced
29      expression of intercellular adhesion molecule-1  (ICAM-1) in  BEAS-2B cells (Kennedy et al.,
30      1998). Cytokine secretion was preceded by activation of nuclear factor kappa B (NF-KB) and
31      was reduced by treatment with superoxide dismutase (SOD),  Deferoxamine (DBF), or

        March 2001                               8-48        DRAFT-DO NOT QUOTE OR CITE

-------
65
g.
O
O
          TABLE 8-8. IN VITRO EFFECTS OF PARTICULATE MATTER AND PARTICULATE MATTER CONSTITUENTS
oo
Tl
H
O
O

g
H
O
o
H
m
o
&
o
t-H
H
M
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
Constituent
DPM







DPM



Four Urban air
particles.
ROFA
DPM
Volcanic ash
Silica



Urban air
particles,
St. Louis SRM
1648;
Washington, DC,
SRM 1649,
Ottawa, Canada,
EHC-93
PMI(,
Mexico City
1993; volcanic
ash (MSHA)
Exposure
Technique Concentration
In vitro 10-100Mg/mL







In vitro 10-lOO^g/mL



In vitro exposure, Urban and DPM;
2x 105 cells exposed 12,27, 111,333, or
for 2 h 1000 Mg/mL
SiO2 and Ti02:
4, 12, 35, or
167^g/mL
Fe2O3 1:1,3.1,
10:1 particles/cell
ratio
In vitro 33 or 100 Mg/mL







In vitro 1-lOOMg/mL



Particle Size
0 4 //m







0 4//m



Urban particles:
0.3-0.4 A^m
DPM: 0.3 Mm
ROFA: 0 5 Mm
Volcanic ash: 1 8 Mm
Silica: 05- 10 Mm
TiO2: <5 Mm
Latex: 3 8 Mm

0 2 to 0.7 Mm







<10,um



Exposure Duration Effect of Particles Reference
2, 4, 6, 24 h DPM caused no gross cellular Bayram et al.
damage. Ciliary beat frequency was ( 1 998a)
attentuated at all doses. DPM
caused IL-8 release at lower dose m
ASTH than NONA. Higher
concentrations of DPM suppressed
IL-8, GM-CSF, and RANTES m
ASTH cells.
24 h DPM attenuated ciliary beating. Bayram et al
Release of IL-8, protein, GM-CSF, ( 1 998b)
and SICAM-1 increased after DPM
exposure.
2 h for cytotoxicity, 16-18 h UAP-induced cytokme production Becker et al.
for cytokine assay; (TNF, IL-6) m AM of both species (1996)
chemiluminescence at that is not related to respiratory
30 minutes burst or transition metals but may be
related to LPS (blocked by
polymyxin B but not DEF)
ROFA induced strong
chemiluminescence but had weak
effects on TNF production.
3, 6, or 18-20 h Phagocytosis was inhibited by UAP Becker and
at 1 8 h. UAP caused decreased Soukup (1 998)
expression of p2-mtegnns involved
m antigen presentation and
phagocytosis.



24 h PM 10 stimulated alveolar Bonner et al.
macrophages to induce up- (1998)
regulation of PDGF « receptor on
myofiboroblasts. Endotoxin and
NHBE cells
                 ROFA
                              In vitro
                                            0, 50, or 200 Mg/mL
Analysis at 2 and 24 h
postexposure
metal components of PM,0 stimulate
release of IL-p. This is a possible
mechanism for PM10-induced airway
remodeling.

Increase in expression of the       Carter et al
cytokines IL-6, IL-8, and TNF-a;   (1997)
inhibition by DMTU or
deferoxamme.

-------
oo
o
 6
 O
 z
 o
 H
O
 c
 o
 H
 W
 O
 ?o
 O
 H
 W
                          TABLE 8-8 (cont'd).  IN VITRO EFFECTS OF PARTICULATE MATTER AND PARTICULATE
                                                                       MATTER CONSTITUENTS
o
o
k— t
Species, Cell type,
etc
Supercoiled
DNA
Particle or Constituent
PM,,| from Edinburgh,
Scotland
Exposure
Technique
In vitro
Concentration Particle Size
996.2 ±181 8 PM,0
//g/filterin 100//L
Exposure Duration
8h
Effect of Particles
PMK) caused damage to DNA, mediated by
hydroxyl radicals (inhibited by mannitol) and
Reference
Donaldson et
al. (1997)
          Rat AM
                            UAP
                            DPM
                                          In vitro    50 to 200 //g/mL
          Primary cultures of   ROFA
          RTE
                                          In vitro    5, 10, or 20 p;g/cm2
DPM- 1.1 - 1.3 pirn
UAP: St Louis,
between 1974 and
1976 in a baghouse,
sieved through
200-mesh(125/jm)
Same as Dreher et al.
(1997)
2 h exposure;
supernatant
collected 18 h
postexposure
Analysis at 6 and
24 h
Peripheral blood     Organic extract of TSP,     In vitro    42.5 ^g extract/m1    N/A, collected from   2h
monocytes         Italy                               (acetone)           high-volume sampler
                                                                       (60 mVh)
iron (DEF). Clear supernatant has all of the
suspension activity. Free radical activity is
derived either from a fraction that is not
centrifugeable on a bench centrifuge or that the
radical generating system is released into
solution.

Dose dependent increase in TNF-a,  IL-6, CINC,  Dong et al.
MIP-2 gene expression by urban particles but     (1996)
not with DPM; cytokme production  were not
related to ROS, cytokme production can be
inhibited by polymyxin B; LPS was  detected on
UAP but not DPM; endotoxm is responsible for
the cytokme gene expression induced by UAP in
AM..

Particle induced epithelial-cell detachment and   Dye et al.
lytic cell injury; alterations in the permeability of (1997)
the cultured RTE cell layer; increase in LDH, G-
6-PDH, gluathione reductase, glutathione S-
transferase; mechanism of ROFA-mduced RTE
cytotoxicity and pulmonary cellular
inflammation involves the development of an
oxidative burden

Superoxide anion generation was inhibited at a   Fabiani et al
paniculate concentration of 0.17 mg/mL when    (1997)
stimulated with PMA, 50% increase in LDH;
disintegration of plasma membrane.
Rat AM
ROFA, iron sulfate,
nickel sulfate, vanadyl
sulfate
Latex particles with
metal complexed on the
surface
In vitro 001-1 0 mg/mL 3.6^mMMAD
(0.7 x 10'
cells/mL)
Up to 400 mm Increase chemilummescence, inhibited by DEF
and hydroxyl radical scavengers; solutions of
metal sulfates and metal-complexed latex
particles similarly elevated chemilummescence
in a dose-and time-dependent manner.
Ohio et al
(1997a)

-------
                TABLE 8-8 (cont'd).  IN VITRO EFFECTS OF PARTICULATE MATTER AND PARTICULATE
o
0
0



Species, Cell type,
etc.
NHBE
BEAS-2B



Particle or Constituent
ROFA


MA1TJ
Exposure
Technique Concentration
In vitro 5-200 /j.g/mL


tK UUINSlil UEIN IS
Exposure
Particle Size Duration
3 6 Aim 2 and 24 h



Effect of Particles
mRNA for ferritm did not change, ferritm protein
increase; mRNA for transfemn receptor decreased,
mRNA for lactofemn increased; transfemn
decreased whereas lactofemn increased;
deferoxamine alone increased lactofemn mRNA.

Reference
Ghio et al.
(1998c)


BEAS-2B
respiratory
epithelial cells

BEAS-2B
0X174 RF1 DNA
                Oil fly ash
                Provo
                TSP soluble and
                insoluble extract
                 PM |(1 from Edinburgh,
                 Scotland
In vitro     lOOi/g/mL            N/A
In vitro     500 A*g/mL            TSP
                                         In vitro    3.7 or 7.5 Aig/mL       PM,,,
                                              = Ih
                                                                                       24 h
                                                                                       8 h
Lactofemn binding with PM metal occurred within   Ghio et al.
5 mm. V and Fe "">, but not Ni, bound to the        (1999b)
lactofemn receptor.

Water soluble fraction caused greater release of IL-8  Ghio et al.
than insoluble fraction  The effect was blocked by   (1999a)
deferoxamine and presumably because of metals (Fe,
Cu, Zn, Pb)
Significant free radical activity on degrading
supercoiled DNA; mainly because of hydroxyl
radicals (inhibited by manmtol); Fe involvement
(DEF-B conferred protection), more Fe3* was
released compared to Fe2*, especially at pH 4.6 than
at 7.2.
                                                                                                       Gilmour et al.
                                                                                                       (1996)

D
§
K>
H -
1
O
0
2
o
H
O
O
H
M
o
***^
o
H
Hamster AM ROFA or CAP In vitro 0, 25, 50, 100, or
200 fig/mL




Hamster AM CAP, ROFA, and their In vitro 0-200 mg/mL
water-soluble and
particulate fractions



AMs from female Vanadyl chloride sodium In vitro 10-1000 A
-------
p
              TABLE 8-8
IN VITRO EFFECTS OF PARTICULATE MATTER AND PARTICULATE
         MATTER CONSTITUENTS
h-*
o
o
~















OO
t-/i
ho


o

>
H
6
o
z;
0
H
O
c
o
H
W
O
?3
n
HH
H
m

Species, Cell type,
etc.
Human PMN









Human AM





Rat 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


Particle or
Constituent
Aqueous and organic
extracts of TSP in
Dusseldorf and
Duisburg, Germany






UAP
(#1648, 1649)
Volcanic ash
ROFA


ROFA, 10 samples
with differing metal
composition
ROFA

Urban dust SRM
1 649, TiO,, quartz

Ambient air
particles, carbon
black, oil fly ash,
coal fly ash

ROFA





ROFA




Exposure Exposure
Technique Concentration Particle Size Duration
In vitro 042-0. 78 mg Collected by high Upto35min
dust/mL volume sampler, 90%
<5 Aim, 50% < 1 Aim,
maximum at
0.3-0.45 Aim
Extracted using water
and then
dichloromethane to
yield aqueous and
organic extracts
In vitro 0,25, 100, or Volume median 24 h
200 /ig/mL diameter:
ROFA 1 1 Aim
#1648: 1.4 Aim
#1649- 1.1 //m
volcanic ash 2.3 Aim
In vitro Oor50Aig/mL 1.99-255^m l-6h
MMAD

In vitro 0,0 5, or 2.0 mg in N/A Ih
lOmL
In vitro 0-100 A(g in I mL N/A 18 h


In vitro 100 (tgm N/A 40 mm
02mL



In vitro 0, 6, 1 2, 25, or 1 96 Aim 1 and 24 h
50 A^g/mL




In vitro 2, 20, or 60 Aig/cm2 1 .96 Aim 24-h exposure




Effect of Particles
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-mduced LCL is inhibited by both types of
extracts, but aqueous extracts have a stronger inhibitory
effect.



ROFA highly toxic, urban PM toxic at 200Aig/mL;
ROFA produced significant apoptosis as low as
25 Aig/mL; UAP produced apoptosis at 100 Aig/mL;
UAP and ROFA also affect AM phenotype:
increased immune stimulatory, whereas decreased
immune suppressor phenotype
Macrophage activation, as determined by
chemilummescence was maximal with the V-rich
particles as opposed to V plus Ni-nch particles.
ROFA induced production of acetaidehyde in dose-
dependant fashion
Cytotoxicity ranking was quartz > SRM 1 649 > TiO,,
based on cellular ATP decrease and LDH, acid
phosphatase, and p-glucuromdase release.
ROS generation, measured by LCL increased in PMN,
was correlated with Si, Fe, Mn, Ti, and Co content but
not V, Cr, Ni, and Cu Deferoxamme, a metal lon-
chelator, and did not affect LCL in PMN, suggesting that
metal ions are not related to the induction of LCL.
Activation of IL-6 gene by NF-KB activation and
binding to specific sequences in promoter of IL-6 gene;
inhibition ofNF-icB activation by DEF and NAC,
increase in PGE,, 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
prostaglandms E2 and F,n; ROFA-mduced increase in
prostaglandin synthesis was correlated with a marked
increase in PHS activity.
Reference
Hitzfeld et al
(1997)








Hohan et al
(1998)




Kodavanti et al
(1998a)

Madden et al
(1999)
Nadeau et al.
(1996)

Prahalad et al
(1999)



Quay et al.
(1998)




Samet et al.
(1996)




-------
O
o
oo
I!A
U)
D
 a
 o
 2
 o
 H
O
 o
 H
                       TABLE 8-8 (cont'd). IN VITRO EFFECTS OF PARTICULATE MATTER AND PARTICULATE
                                                               MATTER CONSTITUENTS
Species, Cell type,
etc.
Human airway
epithelium-derived
cell line BEAS




Human airway
epithelium-derived
Particle or
Constituent
ROFA
Synthetic ROFA
(soluble Ni, Fe,
and V)



Particle
components As, Cr,
Exposure
Technique
In vitro






In vitro

Concentration
ROFA: 0-200
,ug/mL
Synthetic ROFA
(lOOyug/mL):
Ni, 64 A*M
Fe, 63 ^M
V, 370mM
500 ^M of As, F,
Cr (III), Cu, V, Zn
Particle Size
ROFA- 1.96 Aim
Synthetic ROFA- N/A
(soluble)




N/A (soluble)

Exposure
Duration
Up to 24 h






20 min and
6 and 24 h
Effect of Particles
Tyrosme 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
phosphotyrosmes in BEAS cells.

Noncytotoxic concentrations of As, V, and Zn induced
a rapid phosphorylation of MAPK in BEAS cells;
Reference
Samet et al.
(1997)





Samet et al.
(1998)
         cell lines BEAS-2B   Cu, Fe, Ni, V, and
                         Zn
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.
A549
0X174RF1DNA



Rat (Wistar) AM
RAM cells (a rat AM
cell line)

A549


Urban particles
SRM 1648,
St Louis
SRM 1649,
Washington, DC
TiO2



ROFA, a-quartz,
TiO2

In vitro 1 mg/mL for Fe
mobilization assay



In vitro 20, 50, or 80 ^g/mL



In vitro 1 mg/mL


SRM 1648: 50% <
10 //m
SRM 1649: 30% <
10 pirn

N/A



N/A


Up to 25 h 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; ferritm m A549 cells was
increased with treatment of PM suggesting mobilization
of Fe in the cultured cells.
4 h Opsomzation of TiO2 with surfactant components
resulted in a modest increase in AM uptake compared
with that of unopsomzed TiO2; surfactant components
increase AM phagocytosis of particles.
60 mm Exposure of A549 cells to ROFA, a-quartz, but not TiO2,
caused increased IL-8 production in TNF-ct primed cells
in a concentration-dependent manner.
Smith and
Aust(1997)



Stringer and
Kobzik
(1996)

Stnnger and
Kobzik
(1998)
O
h-H
H
W

-------
2 TABLE 8-8 (cont'd). IN VITRO EFFECTS OF PARTICULATE MATTER AND PARTICULATE
1 MATTER CONSTITUENTS
£^ Species, Cell type,
O £tc-
A549
RLE-6TN cells
(type II like cell line)
rat, Long Evans
epithelial cells
°° BEAS-2B human
V| bronchial epithelial
cells
Particle or
Constituent
TiO2, Fe:O3, CAP,
and the fibrogenic
particle a-quartz
PM, 5, Burlington,
VT;"
Fme/ultrafine TiO,
CFA
PFA
a-quartz
ROFA
Birmingham, AL.
188mg/gofVO
Exposure
Technique Concentration Particle Size
In vitro Ti02 [40 Mg/mL], N/A
Fe20, [lOO^g/mL],
a-quartz
[200 Mg/mL], or
CAP [40 Mg/mL]
In vitro 1,2.5,5, or PM25:39nm
10Mg/mL FmeTiOj 159nm
UFTiO2-37nm
17 7 Mm
2 5 //m
In vitro lOO^g/mL N/A
Exposure
Duration Effect of Particles
24 h TiO2 > Fe2Oj > a-quartz > CAP in particle binding;
binding of particle was found to be calcium-dependent
for TiO2 and Fe,O5, while a-quartz binding was
calcium-independent, scavenger receptor, mediate
paniculate binding; a-quartz, but not TiO2 or CAP,
caused a dose-dependent production of IL-8
24 and 48 h Increases in c-Jun kmase activity, levels of
exposure phosphorylated c-Jun immunoreactive protein, and
transcnptional activation of activator protein- 1-
dependent gene expression; elevation in number of
cells incorporating 5'-bromodeoxyundme.
3 h CFA produced highest level of hydroxyl radicals, iron
content is more important than quartz content.
2-6 h ROFA caused increased intracellular Ca", IL-6, IL-8,
and TNF-a through activation of capsicm- and
pH-sensitive receptors.
Reference
Stringer et al
(1996)
Timblin et al
(1998)
Van Maanen
etal. (1999)
Veronesi et al.
(1999)
 "fl
 H

 6
 o

 z
 o


0
 G
 O
 H
 W

 O
 70

 O
 H-H
 H
 m

-------
  1      N-acetylcysteine.  The addition of similar quantities of Cu2+ as found in the Provo extract
  2      replicated the biological effects observed with particles alone.  When normal constituents of
  3      airway lining fluid (mucin or ceruloplasmin) were added to BEAS cells, particulate-induced
  4      secretion of IL-8 was modified. Mucin reduced IL-8 secretion, whereas ceruloplasmin
  5      significantly increased IL-8 secretion and activation of NF-KB. The authors suggest that copper
  6      ions may cause some of the biologic effects of inhaled PM in the Provo region and may provide
  7      an explanation for the sensitivity of asthmatics to Provo PM seen in epidemiologic studies.
  8           There are regional as well as daily variations in the composition of ambient PM and, hence,
  9      its biological activities. For example, concentrated ambient PM (CAP,  from Boston urban air)
 10      has substantial day-to-day variability in its composition and oxidant effects (Goldsmith et al.,
 11      1998).  Similar to  Utah PM, the water-soluble component of Boston CAPs significantly
 12      increased AM oxidant production and inflammatory cytokine (MIP2 and TNFa) production over
 13      negative control values. These effects can be blocked by metal chelators or antioxidants.  The
 14      regional difference in biological activity of ambient PM has been shown by Becker and Soukup
 15      (1998). The oxidant generation, phagocytosis, as well as the expressions of receptors important
 16      for phagocytosis in human alveolar macrophage and blood monocyte were reduced significantly
 17      by PM exposure.
 18           Becker and Soukup (1998) and others (Dong et al., 1996, Becker et al., 1996) have
 19      suggested that the  biological activity of the ambient PM may result from the presence of
 20      endotoxin on the particles rather than metal-associated oxidant generation. Using the same urban
 21      particles (SRM 1648), cytokine production (TNF-a, IL-1,11-6,  CINC, and  MIP-2) was increased
 22      in macrophages following treatment with 50 to 200 /^g/mL of urban PM (Dong et al., 1996). The
 23      urban particle-induced TNF-a secretion was abrogated completely by treatment with polymyxin
 24      B,  an antibiotic that blocks LPS-associated activities, but not with antioxidants. Although it is
 25      possible that LPS may be responsible for ambient PM induced cytokine  gene expression,
 26      extrapolation of these in vitro results to a potential role for endotoxin in the adverse effects of
27      ambient PM must be done with caution because the investigators could not exclude the
28      possibility that the presence of endotoxin with the PM was caused by inadvertent contamination
29      during the year-long collection process or from the handling of the particles.
30           The involvement of endotoxin, at least partially, in PM induced biological effects was
31      supported more recently by Bonner et al. (1998). Urban PM10 collected  from north, south, and

        March 2001                               8-55         DRAFT-DO NOT QUOTE OR CITE

-------
 1     central regions of Mexico City was used with SD rat AM to examine PM effects on platelet
 2     derived growth factor (PDGF) receptors on lung myofibroblasts (Bonner et al., 1998).
 3     Mexico City PM10 (but not volcanic ash) stimulated secretion of upregulatory factors for the
 4     PDGF a receptor, possibly via IL-lp.  In the presence of an endotoxin-neutralizing protein, the
 5     Mexico City PM10 effect on PDGF was blocked partially, suggesting that LPS was responsible
 6     partially for the effect of the PM10 on macrophages. In addition, both LPS and vanadium (both
 7     present in the PM10) acted directly on lung myofibroblasts. However, the V levels in Mexico
 8     City PM10 were probably not high enough to exert an independent effect. The authors concluded
 9     that PM10 exposure could lead to airway remodeling by enhancing myofibroblast replication and
10     chemotaxis.
11           The effects of water soluble as well as organic components (extracted in dichloromethane)
12     of ambient PM were investigated by exposing human PMN to PM extracts (Hitzfeld et al., 1997).
13     PM was collected with high-volume samplers in two German cities, Dusseldorf and Duisburg;
14     these sites have high traffic and high industrial emissions, respectively.  Organic, but not
15     aqueous, extracts of PM alone significantly stimulated the production and release of ROS in
16     resting human PMN. The effects of the PM extracts were inhibited by SOD, catalase, and
17     sodium azide (NaN3).  Similarly, the organic fraction (extractable by acetone) of ambient PM
18     from Terni, Italy, had been shown to produce cytotoxicity, superoxide release in response to
19     PMA and zymosan in peripheral monocytes (Fabiani et al., 1997).
20
21     8.5.2.2  Residual Oil  Fly Ash
22           In a series of studies using the same ROFA samples, several experiments have investigated
23     the biochemical and molecular mechanisms involved in ROFA induced cellular injury.
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 F2 a.  The ROFA-
27     induced increase in prostaglandin synthesis was correlated with a marked increase in activity of
28     the PHS-2 form of prostaglandin H synthase as well as mRNA coded for this enzyme.
29     In contrast, expression of the PHS1 form of the enzyme was not affected by ROFA treatment of
30     airway epithelial cells. These investigators further demonstrated that noncytotoxic levels of
31     ROFA induced a significant dose- and time-dependent increase in protein tyrosine phosphate, an

       March 2001                               8-56        DRAFT-DO NOT QUOTE OR CITE

-------
  1      important regulator of signal transduction leading to cell growth and proliferation. ROFA-
  2      induced increases in protein phosphotyrosines were associated with its soluble fraction and were
  3      mimicked by V-containing solutions but not iron or nickel solutions (Samet et al., 1997).
  4           ROFA also stimulates respiratory cells to secret inflammatory cytokines such as IL-6, IL-8,
  5      and TNF. Normal human bronchial epithelial (NHBE) cells exposed to ROFA produced
  6      significant amounts of IL-8, IL-6, and TNF, as well as mRNAs coding for these cytokines (Carter
  7      et al., 1997). Increases in cytokine production, but not m-RNA expression, were dose-dependent.
  8      The cytokine production was inhibited by the addition of metal chelator, DBF, or the free radical
  9      scavenger, DMTU. Similar to the data of Samet et al. (1997), V but not Fe or Ni compounds
 10      were responsible for these effects. Cytotoxicity, decreased cellular glutathione levels in primary
 11      cultures of rat tracheal epithelial (RTE) cells exposed to suspensions of ROFA indicated that
 12      respiratory cells exposed to ROFA were under oxidative stress. Treatment with buthionine
 13      sulfoxamine (an inhibitor of y-glutamyl cysteine synthetase) augmented ROFA-induced
 14      cytotoxicity, whereas treatment with DMTU inhibited ROFA-induced cytoxicity further
 15      suggested that ROFA-induced cell injury may be mediated by hydroxyl-radical-like ROS (Dye
 16      et al., 1997). Using BEAS-2B cells, a time- and dose-dependent increase in IL-6 mRNA induced
 17      by ROFA was shown to precede by the activation of nuclear proteins NF-kB (Quay et al., 1998).
 18      Taking together, ROFA exposure increases oxidative stress, perturbs protein tyrosine phosphate
 19      homeostasis, activates NF-kB, and up-regulates inflammatory cytokine and prostaglandin
 20      synthesis and secretion to produce lung injury.
 21           Stringer and Kobzik (1998) observed that "primed" lung epithelial cells exhibited enhanced
 22      cytokine responses to PM. Compared to normal cells, exposure of TNF-cc-primed A549 cells to
 23      ROFA or a -quartz caused increased IL-8 production in a concentration-dependent manner for
 24      particle concentrations ranging from 0-200 /^g/mL.  Addition of the antioxidant NAC (1.0 mM)
 25      decreased ROFA and a -quartz-mediated IL-8 production by approximately 50% in both normal
26      and TNF- a -primed A549 cells. Exposure of A549 cells to ROFA caused an increase in oxidant
27      levels that could be inhibited by NAC. These data suggest that (1) lung epithelial cells primed by
28      inflammatory mediators show increased cytokine production after exposure to PM, and
29      (2) oxidant stress is an important mechanism for this response.
30           In summary, exposure of lung cells to ambient PM or ROFA leads to increased production
31      of cytokines and the effects may be mediated, at least in part, through production of ROS.

        March 2001                               8-57        DRAFT-DO NOT QUOTE OR CITE

-------
 1     Day-to-day variations in the components of PM, such as soluble transition metals, which may be
 2     critical to eliciting the response, are suggested. The involvement of organic components in
 3     ambient PM also was suggested in some studies.
 4
 5     8.5.3  Potential Cellular and Molecular Mechanisms
 6     8.5.3.1  Reactive Oxygen Species
 7          Ambient particulate matter contains transition metals, such as iron (most abundant),
 8     copper, nickel, vanadium, and cobalt.  These metals are capable of catalyzing the one-electron
 9     reductions of molecular oxygen necessary to generate reactive oxygen species (ROS).  These
10     reactions can be demonstrated by the iron-catalyzed Haber-Weiss reactions that follow.
11
12
13                          Reductant" + Fe(III) -> Reductantn+1 + Fe(II)                     (1)
14                                   Fe(II) + 0~2^> Fe(III) + O2                              (2)
15                                  HO2+O~+H+^ O2+H2O2                             (3)
16                      Fe(II) + H2O2 -> Fe(III)+*OH + HO"(Fenton Reaction)                 (4)
17
18     Iron will continue to participate in the redox cycle in the above reactions as long as there is
19     sufficient O2 or H2O2 and reductants.
20          Soluble metals from inhaled  PM dissolved into the fluid lining of the airway lumen can
21     react directly with biological molecules (acting as a reductant in the above reactions) to produce
22     ROS. For example, ascorbic acid in the human lung epithelial lining fluid can react with Fe(III)
23     from inhaled PM to cause single strand breaks in supercoiled plasmid DNA, (j>X174 RFI (Smith
24     and Aust, 1997). The DNA damage caused by a PM10  suspension can be inhibited by mannitol,
25     an hydroxyl radical scavenger, further confirming the involvement of free radicals in these
26     reactions (Gilmour et al., 1996; Donaldson et al., 1997; Li et al., 1997). Because the clear
27     supernatant of the centrifuged PM10 suspension contained all of the suspension activity, the free
28     radical activity is derived either from a fraction that is not centrifugable (10 min at 13,000 rpm
        March 2001                               8-58        DRAFT-DO NOT QUOTE OR CITE

-------
  1      on a bench centrifuge) or the radical generating system is released into solution (Gilmour et al.,
  2      1996; Donaldson et al., 1997; Li et al., 1997).
  3           In addition to measuring the interactions of ROS and biomolecules directly, the role of
  4      ROS in PM-induced lung injury also can be assessed by measuring the electron spin resonance
  5      (ESR) spectrum of radical adducts or fluorescent intensity of dichlorofluorescin (DCFH), an
  6      intracellular dye that fluoresces on oxidation by ROS. Alternatively, ROS can be inhibited using
  7      free radical scavengers, such as dimethylthiourea (DMTU); antioxidants, such as glutathione or
  8      N-acetylcysteine (NAC); or antioxidant enzymes, such as superoxide dismutase (SOD). The
  9      diminished response to PM after treatment with these antioxidants indicates the involvement of
10      ROS.
11           As described earlier, Kadiiska et al. (1997) used the ESR spectra of 4-POBN [a-(4-pyridyl
12      l-oxide)-N-tert-butylnitrone] adducts to measure ROS in rats instilled with ROFA and
13      demonstrated the association between ROS production within the lung and soluble metals in
14      ROFA. Using DMTU to inhibit ROS production, Dye et al. (1997) had shown that systemic
15      administration of DMTU impeded development of the cellular inflammatory response to ROFA,
16      but did not ameliorate biochemical alterations in BAL fluid. Goldsmith et al. (1998), as
17      described earlier, showed that ROFA and CAPs caused increases in ROS production in AMs.
18      The water-soluble component of both CAPs and ROFA significantly increased AM oxidant
19      production over negative control values. In addition, increased PM-induced cytokine production
20      was inhibited by NAC.  Li et al. (1996, 1997) instilled rats with PM10 particles (collected on
21      filters from an Edinburgh, Scotland, monitoring station).  Six hours after intratracheal instillation
22      of PM10, they observed a decrease in glutathione (GSH) levels in the BAL fluid. Although this
23      study does not describe the composition of the PM10, the authors suggest that changes in GSH, an
24      important lung antioxidant, support the contention that the free radical activity of PM10 is
25      responsible for its biological activity in vivo.
26           In addition to ROS generated directly by PM, resident or newly recruited AMs or PMNs
27      also are capable of producing these reactive species on stimulation. The ROS produced during
28      the oxidative burst can be measured using a chemiluminescence (CL) assay. With this  assay,
29      AM CL signals in vitro have been shown to be greatest with ROFA containing primarily soluble
30      V and were less with ROFA containing Ni plus V (Kodavanti et al., 1998a). As described
31      earlier, exposures to Dusseldorf and Duisburg PM increased the resting ROS production in

        March 2001                              8-59         DRAFT-DO NOT QUOTE OR CITE

-------
 1     PMNs, which could be inhibited by SOD, catalase, and sodium azide (Hitzfeld et al., 1997).
 2     Stringer and Kobzik (1998) showed that addition of NAC (1.0 mM) decreased ROFA-mediated
 3     IL-8 production by approximately 50% in normal and TNF-a-primed A549 cells. In addition,
 4     exposures of A549 cells to ROFA caused a substantial (and NAC inhibitable) increase in oxidant
 5     levels as measured by DCFH oxidation. In human AMs, Becker et al. (1996) found a CL
 6     response for ROFA, but not urban air particles (Ottawa and Dusseldorf) or volcanic ash.
 7           Metal compounds of PM are the most probable species capable of catalyzing ROS
 8     generation on exposure to PM. To determine elemental content and solubility in relation to their
 9     ability to generate ROS, PMN or monocytes were exposed to a wide range of ambient air
10     particles from divergent sources (one natural dust, two types of oil fly ash, two types of coal fly
11     ash, five different ambient air samples, and one carbon black sample) (Prahalad et al., 1999), and
12     CL production was measured over a 20-min period postexposure. Percent of sample mass
13     accounted for by XRF detectable elements was 1.2% (carbon black); 22 to 29% (natural dust and
14     ambient air particles); 13 to 22% (oil fly ash particles); and 28 to 49% (coal fly ash particles).
15     The major proportion of elements in most of these particles were aluminosilicates and insoluble
16     iron, except oil derived fly ash particles in which soluble vanadium and nickel were in highest
17     concentration, consistent with particle acidity as measured in the supernatants. All particles
18     induced CL response in cells, except carbon black. The CL response of PMNs in general
19     increased with all washed particles,  with oil fly ash and one urban air particle showing statistical
20     differences between deionized water washed and unwashed particles. These CL activities were
21     significantly correlated with the insoluble Si, Fe, Mn, Ti, and Co content of the particles.
22     No relationship was found between  CL and soluble transition metals such as V, Cr, Ni, and Cu.
23     Pretreatment  of the particles with a metal ion chelator, deferoxamine, did not affect CL activities.
24     Particle sulfate content and acidity of the particle suspension did not correlated with CL activity.
25           Soluble metals can be mobilized into the epithelial cells or AMs to produce ROS
26     intracellularly. Size fractionated coal fly ash particles (2.5, 2.5 to 10, and <10 yum) of bituminous
27     b (Utah coal), c (Illinois coal), and lignite (Dakota coal) were used to compare the amount of iron
28     mobilization  in A549 cells and by citrate (I mM) in cell-free suspensions (Smith et al., 1998).
29     Iron was mobilized by citrate from all three size fractions of all three coal types.  More  iron, in
30     Fe(III) form, was mobilized by citrate from the <2.5-p:m fraction than from the >2.5-p:m
31     fractions. In  addition, the amount of iron mobilized was dependent on the type of coal used to

       March 2001                               8-60        DRAFT-DO NOT QUOTE OR CITE

-------
  1      generate the fly ash (Utah coal > Illinois coal = Dakota coal) but not related to the total amount
  2      of iron present in the particles. Ferritin (an iron storage protein) levels in A549 cells increased by
  3      as much as 11.9-fold in cells treated with coal fly ash (Utah coal > Illinois coal > Dakota coal).
  4      More ferritin was induced in cells treated with the <2.5-/^m fraction than with the >2.5-/um
  5      fractions. Mossbauer spectroscopy of a fly ash sample showed that the bioavailable iron was
  6      assocated with the glassy aluminosilicate fraction of the particles (Ball et al., 2000). As with the
                                                         s
  7      bioavailability of iron, there was an inverse correlation between the production of IL-8 and fly
  8      ash particle size with the Utah coal fly ash being the most potent.
  9          Using ROFA and colloidal iron oxide, Ohio et al. (1997b; 1998a,b,c;  1999c; 2000b) have
 10      shown that exposures to these  particles disrupted iron homeostasis and induced the production of
 11      ROS in vivo and in vitro. Treatment of animals or cells with metal-chelating agents such as DBF
 12      with an associated decrease in response has been used to infer the involvement of metal in PM-
 13      induced lung injury. Metal chelation by DEF (1 mM) caused significant inhibition  of particulate-
 14      induced AM oxidant production, as measured using DCFH (Goldsmith et al., 1998). DEF
 15     treatment also reduced NF-KB activation and cytokine secretion in BEAS-2B cells exposed to
 16     Provo PM (Kennedy et al., 1998). However, treatment of ROFA suspension with DEF was not
 17     effective in blocking leachable metal induced acute lung injury (Dreher et al., 1997). Dreher
 18     et al. (1997) indicated that DEF could chelate Fe(III) and V(II), but not Ni(II), suggesting that metal
 19     interactions played a significant role in ROFA-induced lung injury.
 20          Other than Fe, several V  compounds have been shown to increase mRNA levels for
 21      selected cytokines in BAL cells and also to induce pulmonary inflammation (Pierce et al., 1996).
 22      NaVO3 and VOSO4, highly soluble forms of V, tended to induce pulmonary inflammation and
 23      inflammatory cytokine mRNA expression more rapidly and more intensely than the less soluble
 24      form, V2O5, in rats.  Neutrophil influx was greatest following exposure to VOSO4 and lowest
 25      following exposure to V2O5.  However, metal components of fly ash have not been shown to
 26      consistently increase ROS production from bovine AM treated with combustion particles
                                                        \
 27      (Schluter et al., 1995).  For example, As(III), Ni(II), and Ce(III), which are major components of
28      fly ash, had been shown to inhibit the secretion of superoxide anions (O2~) and hydrogen
29      peroxide. In the same study, O2~ were lowered by Mn(II) and Fe(II), whereas V(IV) increased O2~
30      and H2O2. In contrast, Fe(III) increase O2" productions, demonstrating that the oxidation state of


        March 2001                               8-61        DRAFT-DO NOT QUOTE OR CITE

-------
  1      metal may influence its oxidant generating properties.  Other components of fly ash, such as
  2      Cd(II), Cr(III), and V(V), had no effects on ROS.
  3           It is likely that a combination of several components rather than a single metal in PM is
  4      responsible for the PM induced cellular response.  For example, V and Ni+V but not Fe or Ni
  5      alone (in saline with the final pH at 3.0) resulted in increased epithelial permeability, decreased
  6      cellular glutathione, cell detachment, and lytic cell injury in rat tracheal epithelial cells exposed
  7      to soluble salts of these metals at equivalent concentrations found in ROFA (Dye et al., 1999).
  8      Treatment of V-exposed cells with buthionine sulfoximine further increased cytotoxicity.
  9      Conversely, treatment with radical scavenger dimethyl thiourea inhibited the effects in a dose-
 10      dependent manner. These results showed that soluble metal  or combinations of several metals in
 11      ROFA are responsible for these effects.
 12           Similar to combustion particles such as ROFA, the biological response to exposure to
 13      ambient PM also appear to  depend on the metal content of the particles. Human subject were
 14      instilled with 500 fj.g (in 20 mL sterile saline) of Utah Valley dust (UVD1, 2, 3, collected during
 15      3 successive years) on the left segmental bronchus and on the right side with sterile saline as
 16      control. Twenty-four-hour postinstillation, a second bronchoscopy was performed and
 17      phagocytic cells were obtained on both  side of the segmental bronchus.  AM from subjects
 18      instilled with UVD, obtained by bronchoaveolar lavage 24 h postinstillation, were incubated with
 19      fluoresceinated yeast (Saccharomyces cerevisiae) to assess their phagocytic ability.  Although the
20      same proportion of AMs were exposed to UVD phagocytized yeast, AMs exposed to UVD1,
21      which were collected while a local steel mill was open, took  up significantly less particles than
22      AMs exposed to other extracts (UVD2 when the steel mill was closed and UVD3 when the plant
23      reopened). AMs exposed to UVD1 also exhibited a small decrease in oxidant activity (using
24      dihydrorhodamine-123, DHR).  AMs from healthy volunteers were incubated in vitro with the
25      various UVD extracts to assess whether similar effects on human AMs function could be
26      observed to those seen following in vivo exposure. The percentage of AMs that engulfed yeast
27      particles was significantly decreased by exposure to UVD1 at 100 //g/mL, but not at 25 //g/mL.
28      However, the amount of particles engulfed was the same following exposure to all three UVD
29      extracts. AMs also demonstrated increased oxidant stress (using chemiluminescence)  after in
30      vitro exposure to UVD1 and this effect was not abolished with pretreatment of the extract with
31      the metal chelator deferoxamine. As with the AMs exposed to UVD in vivo, AM exposed to

        March 2001                               8-62        DRAFT-DO NOT QUOTE OR CITE

-------
  1      UVD in vitro had a decreased oxidant activity (DHR assay).  UVD1 contains 61 times and
  2      2 times the amount of Zn compared to UVD 2 and UVD3, respectively, whereas UVD3
  3      contained 5 times more Fe than UVD1. Ni and V were present only in trace amounts. Using the
  4      same particles, Frampton et al. (1999) exposed BEAS-2B cells for 2 and 24 h.  Similar results
  5      were observed for oxidant generation in these cells (i.e., UVD 2, which contains the lowest
  6      concentrations of soluble iron, copper, and zinc, produced the least response).  Only
  7      UVD 3 produced cytotoxicity at a dose  of 500 //g/mL. UVD 1 and 3, but not 2, induced
  8      expression of IL-6 and 8 in a dose-dependent fashion. Taken together, these data showed that
  9      biological response to ambient particles exposure is heavily dependent on the source and, hence,
 10      the chemical composition of PM.
 11
 12      8.5.3.2  Intracellular Signaling Mechanisms
 13           In has been shown that the intracellular redox state of the cell modulates the activity of
 14      several transcription factors, including NF-KB, a critical step in the induction of a variety of
 15      proinflammatory cytokine and adhesion-molecule genes. NF-KB is a heterodimeric protein
 16      complex that in most cells resides in an inactive state in the cell cytoplasm by binding to
 17      inhibitory kappa B alpha (IkBcc). On appropriate stimulation by cytokines or ROS, hcBcc is
 18      phosphorylated and subsequently degraded by proteolysis.  The dissociation of iKBa from NF-KB
 19      allows the latter to translocate into the nucleus and bind to appropriate sites in the DNA to
 20      initiate transcription of various genes. Two studies in vitro have shown the involvement of
 21      NF-KB in particulate-induced cytokine and intercellular adhesion molecule-1 (ICAM-1)
 22      production in human airway epithelial cells (BEAS-2B) (Quay et al., 1998; Kennedy et al.,
 23      1998). Cytokine  secretion was preceded by activation of NF-KB and was reduced by treatment
 24      with  antioxidants or metal chelators. These results suggest that metal-induced oxidative stress
 25      may play a significant role in the initiation phase of the inflammatory cascade following
26      particulate exposure.
27           A second well-characterized human transcription factor, AP-1, also responds to the
28      intracellular ROS concentration. AP-1 exists in two forms, either in a homodimer of c-jun
29      protein or a heterodimer consisting of c-jun and c-fos. Small amounts of AP-1  already exist in
30      the cytoplasm in an inactive form, mainly as phosphorylated c-jun homodimer.  Many different
31      oxidative stress-inducing stimuli, such as UV light and IL-1, can activate AP-1.  Exposure of rat

        March 2001                               8-63        DRAFT-DO NOT QUOTE OR CITE

-------
  1      lung epithelial cells to ambient PM in vitro resulted in increases in c-jun kinase activity, levels of
  2      phosphroylated c-jun immunoreactive protein, and transcriptional activation of AP-1 -dependent
  3      gene expression (Timblin et al., 1998).  This study demonstrated that interaction of ambient
  4      particles with lung epithelial cells initiates a cell signaling cascade related to aberrant cell
  5      proliferation.
  6           Early response gene transactivation has been linked to the development of apoptosis, a
  7      unique type of programmed cell injury and a potential mechanism to account for PM-induced
  8      changes in cellular response. Apoptosis of human AMs exposed to ROFA (25 //g/mL) or urban
  9      PM was observed by Holian et al. (1998).  In addition, both ROFA and urban PM upregulated the
10      expression of the RFD1 + AM phenotype, whereas only ROFA decreased the RFDl+7f phenotype.
11      It has been suggested that an increase in the AM phenotype ratio of RFDl+/RFDl+7f may be
12      related to disease progression in patients with inflammatory diseases. These data showed that
13      ROFA and urban PM can induce apoptosis of human AMs and increase the ratio of AM
14      phenotypes toward a higher immune active state and may contribute to or exacerbate lung
15      inflammation.
16           Another intracellular signaling pathway that can lead to diverse cellular responses such as
17      cell growth, differentiation, proliferation, apoptosis, and stress responses to environmental
18      stimuli, is the phosphorylation-dependent, mitogen-activated protein kinase (MAPK).
19      Noncytotoxic levels of ROFA have been shown to induce  significant dose- and time-dependent
20      increases in protein tyrosine phosphate levels in BEAS cells (Samet et al., 1997). In a
21      subsequent study, the effects of As, Cr, Cu, Fe, Ni, V, and Zn on the MAPK, extracellular
22      receptor kinase (ERK), c-jun N-terminal kinase (JNK), and P38 in BEAS cells were investigated
23      (Samet et al., 1998). Noncytotoxic concentrations of As, V, and Zn induced a rapid
24      phosphorylation of MAPK in BEAS cells.  Activity assays confirmed marked activation of ERK,
25      JNK, and P38 in BEAS cells exposed to As, V, and Zn. Cr and Cu exposure resulted in a
26      relatively small activation of MAPK, whereas Fe and Ni did not activate MAPK.  Similarly, the
27      transcription factors c-Jun and ATF-2, substrates of JNK and P38, respectively, were markedly
28      phosphorylated in BEAS cells treated with As, Cr, Cu, V, and Zn. The same acute exposure to
29      As, V, or Zn that activated MAPK was sufficient to induce a subsequent increase in IL-8 protein
30      expression in BEAS cells.  These data suggest that MAPK may mediate metal-induced
31      expression of inflammatory proteins in human bronchial epithelial cells. The ability of ROFA to

        March 2001                              8-64       DRAFT-DO NOT  QUOTE OR CITE

-------
  1      induce activation of MAPKs in vitro was demonstrated by Silbajoris et al. (2000). In addition,
  2      Gerchen et al. (1996) showed that the ROS production induced by PM was markedly decreased
  3      by the inhibition of protein kinase C as well as phospholipase A2.
  4           The major cellular response downstream of ROS and the cell signaling pathways described
  5      above is the production of inflammatory cytokines or other reactive mediators. In an effort to
  6      determine the contribution of cyclooxygenase to the pulmonary responses to ROFA exposure
  7      in vivo, Samet et  al. (2000) intratracheally instilled Sprague Dawley rats with ROFA (200 or
  8      500 fj.g in 0.5 mL saline). These animals were pretreated, intraperitoneally, with 1 mg/kg ROFA
  9      (in 20% ethanol in saline) 30 min prior to the intratracheal exposure.  At 12 h after intratracheal
 10      instillations, intraperitoneal injections (1 mL) were repeated. ROFA treatment induced a marked
 11      increase in the level of PGE2 recovered in the BALF, which was effectively decreased by
 12      pretreating the animals with specific prostaglandin H synthase 2 (COX2) inhibitor NS398.
 13      Immunohistochemical analyses of rat airway showed concomitant expression of COX2 in the
 14      proximal airway epithelium of rats treated with soluble fraction of ROFA. This study further
 15      showed that, although COX2 products participated in ROFA induced lung inflammation, the
 16      COX metabolites are not involved in IL-6 expression nor the influx of PMN influx into the
 17      airway. However, the rationale for the use of intraperitoneal challenge was not elaborated.
 18           The production of cytokines and mediators also has been shown to depend on the type of
 19      PM used in the experiments.  A549  cells (a human airway epithelial cell line) were exposed to
 20      several PM, carbon black (CB, Elftex-12, Cabot Corp.), diesel soot (ND from NIST, LD
 21      produced from General Motors LH 6.2 V8 engine at light duty cycle), ROFA (from the heat
 22      exchange section of the Boston Edison), OAA (Ottowa ambient air PM, EHC-93), SiO2, and
 23      Ni3S2 at lmg/cm2  (Seagrave and Nikula, 2000). Results indicated that (1) SiO2 and Ni3S2 caused
 24      dose dependent acute toxicity and apototic changes; (2) ROFA and LD, ND were significant only
 25      at the highest concentrations, (3) SiO2 and Ni3S2 increased IL-8 (three and eight times over the
26      control, respectively) at low concentrations but suppressed IL-8 at high concentrations, (4) OAA
27      and ROFA also induced IL-8 but lower than SiO2 and Ni3S2, and (5) both diesel soots suppressed
28      IL-8 production. The order of potency in  alkaline phasphatase production is OAA > LD =
29      ND > ROFA » SiO2 = Ni3S2.  These results  demonstrated that not only the type of particle used
30      but also the exposure-dose influence the biological response.


        March 2001                              8-65        DRAFT-DO NOT QUOTE OR CITE

-------
 1           Expression of MIP-2 and IL-6 genes was significantly upregulated as early as 6 h
 2      post-ROFA-exposure in rat tracheal epithelial cells, whereas gene expression of iNOS was
 3      maximally increased 24 h postexposure.  V but not Ni appeared to be mediating the effects of
 4      ROFA on gene expression.  Treatment with dimethylthiourea inhibited both ROFA and V
 5      induced gene expression in  a dose-dependent manner (Dye et al., 1999).
 6           It appears that many biological responses are produced by PM whether it is composed of a
 7      single component or a complex mixture.  A technical approach is to use the newly developed
 8      gene array to monitor the expressions of many mediator genes, which regulate complex and
 9      coordinated cellular events  involved in tissue injury and repair, in a single assay. Using an array
10      consisting of 27 rat genes representing inflammatory and anti-inflammatory cytokines,  growth
11      factors, adhesion molecules, stress proteins, transcription factors, and antioxidant enzymes,
12      Nadadur et al. (2000) measured the expressions of these genes in rats intratracheally instilled
13      with ROFA (3.3 mg/kg), NiSO4 (1.3 /umol/kg), and VSO4 (2.2 //mol/kg). Their data revealed a
14      twofold induction of IL-6 and TIMP-1 at 24 h post-ROFA or Ni  exposure. The  expression of
15      cellular fibronectin (cFn-EIIIA), ICAM-1, IL-lb, and iNOS gene also were increased 24 h
16      post-ROFA, V, or Ni exposure. This study demonstrated that gene array may provide a tool for
17      screening the expression profile of tissue specific markers following exposure to PM.
18           To investigate the interaction between respiratory cells and PM, Kobzik (1995) showed that
19      scavenger receptors are responsible for AM binding of unopsonized PM and  that different
20      mechanisms mediate binding of carbonaceous dusts such as DPM. In addition, surfactant
21      components can increase AM phagocytosis of environmental particulates in vitro, but only
22      slightly relative to the already avid AM uptake of unopsonized particles (Stringer and Kobzik,
23      1996).  Respiratory tract epithelial cells are also capable of binding with PM  to secrete cytokine
24      IL-8. Using a respiratory epithelial cell line (A549), Stringer et al. (1996) found that binding of
25      particles to epithelial cells was calcium-dependent for TiO2 and Fe2O3, while cc-quartz binding
26      was not calcium dependent. In addition,  as observed in AMs, PM binding by A549 cells also
27      was mediated by scavenger receptors, albeit those distinct from the heparin-insensitive
28      acetylated-LDL receptor. Furthermore, a-quartz, but not TiO2 or CAPs, caused  a dose-dependent
29      production of IL-8 (range 1  to 6 ng/mL),  demonstrating a particle-specific spectrum of epithelial
30      cell cytokine (IL-8) response.
31
        March 2001                               8-66        DRAFT-DO NOT QUOTE OR CITE

-------
  1      8.5.3.3  Other Potential Cellular and Molecular Mechanisms
  2           In addition to inducing cytokine mediated inflammation, PM also may affect the alveolar
  3      surfactant's ability to reduce both the tendency of alveoli to collapse at the end of expiration and
  4      the transudation of fluid from the capillaries to the airspace. Lee et al. (1999) exposed guinea
  5      pigs and rats to high concentrations of sulfuric acid aerosol (43 and 94 nig/m3, 0.9 /urn MMAD)
  6      and investigated the effects of this aerosol on the surface properties of reconstituted phospholipid
  7      using a captive bubble surfactometer. The acid exposure significantly increased the surface
  8      tension of guinea pig but not rat BAL. The most sensitive index of surfactant inhibition was
  9      found to be the maximum film compressibility of the compression isotherm. The index was
 10      119 times greater for the acid exposed guinea pigs compared to control animals.  These results
 11      were associated with an increase in protein and PMN in the BAL.  Although unusually high
 12      concentrations of acid aerosols were used in this study, the results may explain the lack of
 13      response in the rat to acid aerosol exposures.
 14           The potential mechanism involving in the alteration of surface tension may be related to
 15      changes in the expression of matrix metalloproteinases (MMPs), such as pulmonary matrilysin
 16      and gelatinase A  and B, and tissue inhibitor of metalloproteinase (TIMP) (Su et al., 2000a,b).
 17      Sprague Dawley rats exposed to ROFA by intratracheal injection (2.5 mg/rat)  had increased
 18      mRNA levels of matrilysin,  gelatinase A, and TIMP-1,  Gelatinase B, not expressed in control
 19      animals, was increased significantly from 6 to 24 h following ROFA exposure. Alveolar
 20      macrophages, epithelial cells, and inflammatory cells were major cellular sources for the
 21      pulmonary MMP expression. The expression of Gelatinase B in rats exposed  to the same dose of
 22      ambient PM (<1.7 /^m and 1.7 to 3.7 /^m) collected from Washington, DC, was significantly
 23      increased as compared to saline control, whereas the expression of TIMP-2 was suppressed.
 24      Ambient PM between 3.7 and 20 ^m also increased the Gelatinase B expression. Increases in
25      MMPs, which degrade most of the extracellular matrix, suggest that ROFA and ambient PM can
26      similarly increase the total pool of proteolytic activity to the lung and contribute in the
27      pathogenesis of PM-induced lung injury.
28           Sensory nerves originating from trigeminal, nodos, and dorsal root ganglion neurons
29      (DRGs) extend their terminals into the nasal and/or pulmonary epithelium.  These nerve
30      terminals together with sensory irritant receptors (capsaicin and acid sensitive  receptors) found
31      on the cell bodies can be triggered by irritants such as ambient PM or its components. The

        March 2001                               8-67         DRAFT-DO NOT QUOTE OR CITE

-------
 1      activation of these receptors and nerve terminals can result in the release of inflammatory
 2      cytokines leading to airway disorders.  Intracellular calcium levels increased immediately in
 3      BEAS-2B cells exposed to ROFA, which is followed by increased in IL-6, IL-8, and TNFoc gene
 4      expressions (Veronesi et al., 1999). Furthermore, acidic media alone or soluble components of
 5      ROFA produced similar effects. These responses were reduced by pretreating cells with
 6      neuropeptide antagonists.  However, treating cells with capsaicin antagonist, a pH receptor
 7      antagonist, or exposing cells to ROFA in Ca2+ free media inhibited both intracellular Ca2+ as well
 8      as cytokine release. Using synthetic polymer microspheres (SPMs) resembling ROFA particles
 9      in size (2 and 6 ^m in diameter) and surface potential (zeta potential -29 mV) but lacking
10      confounding factors such as metals or biologies, Oortgiesen et al. (2000) demonstrated that
11      BEAS-2B and DRGs responded to both ROFA and charged SPMs with an increase in
12      intracellular Ca2+ ([Ca2+],) concentration and the release of IL-6, whereas neutral SPMs bound
13      with polyethylene glycol (0 mV zeta potential) were relatively ineffective. In DRGs, the SPM-
14      induced increases in [Ca2+], were correlated with the presence of acid- or capsaicin-sensitive
15      pathways. By this pathway, soluble components of ROFA, which  is acidic, and other acidic PM
16      may initiate or exacerbate symptoms of airway inflammation. These data not only demonstrated
17      that the surface chemistry of the particles determines whether cells are activated but also that
18      direct contact of the particle with the target cells and their receptors is necessary for particles to
19      evoke a response.
20
21      8.5.4 Specific Particle Size and Surface Area Effects
22           Most particles used in laboratory animal toxicology and occupational studies are greater
23      than 0.1 /zm in size. However, the enormous number and huge surface area of the ultrafine
24      particles demonstrate the importance of considering the size of the particle in assessing response.
25      Ultrafine particles with a diameter of 20 nm when inhaled at the same mass concentration have a
26      number concentration that is approximately 6 orders of magnitude higher than for a 2.5-/urn
27      diameter particle; particle surface area is also greatly increased (Table 8-9).
28           Many studies summarized in U.S. Environmental Protection  Agency (1996a), as well as in
29      this document, suggest that the surface of particles or substances that are released from the
30      surface (e.g., transition metals) interact with the biological system, and that surface-associated
31      free radicals or free radical-generating systems may be responsible for toxicity. Thus, if ultrafine
        March 2001                               8-68       DRAFT-DO NOT QUOTE OR CITE

-------
                TABLE 8-9. NUMBERS AND SURFACE AREAS OF MONODISPERSE
                PARTICLES OF UNIT DENSITY OF DIFFERENT SIZES AT A MASS
                                   CONCENTRATION OF 10
Particle Diameter
Cum)
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      particles were to cause toxicity by a transition metal-mediated mechanism, for example, then the
  2      relatively large surface area for a given mass of ultrafine particles would mean high
  3      concentrations of transition metals being available to cause oxidative stress to cells.
  4           Two groups have examined the toxic differences between fine and ultrafine particles, with
  5      the general finding that the ultrafine particles show a significantly greater response at similar
  6      mass doses (Oberdorster et al., 1992; Li et al., 1996, 1997, 1999). However, only a few studies
  7      have investigated the ability of ultrafine particles to generate a greater oxidative stress when
  8      compared to fine particles of the same material. Studies by Gilmour et al. (1996) have shown
  9      that at equal mass, ultrafine TiO2 caused more plasmid DNA strand breaks than fine TiO2. This
1 0      effect could be inhibited with mannitol. Osier and Oberdorster (1 997) compared the response of
1 1      rats  (F344) exposed by intratracheal inhalation to "fine" (approximately 250 nm) and "ultrafine"
12      (approximately 2 1 nm) TiO2 particles with rats exposed to similar doses by intratracheal
13      instillation. Animals receiving particles through inhalation showed a smaller pulmonary
14      response, measured by BAL parameters, in both severity and persistence, when compared with
1 5      those animals receiving particles through instillation. These results demonstrate a difference in
16      pulmonary response to an inhaled versus an instilled dose, which may result from differences in
17      dose rate, particle distribution, or altered clearance between the two methods. Consistent with
1 8      these in vivo studies,  Finkelstein et al. (1997) has shown that exposing primary cultures of rat
19      Type II cells to 10 ^g/mL ultrafine TiO2 (20 nm) causes increased TNF and IL-1 release
       March 2001                               8-69        DRAFT-DO NOT QUOTE OR CITE

-------
  1      throughout the entire 48-h incubation period. In contrast, fine TiO2 (200 nm) had no effect.
  2      In addition, ultrafine polystyrene carboxylate-modified microspheres (UFP, fluorospheres,
  3      molecular probes 44 ± 5 nm) have been shown induce a significant enhancement of both
  4      substance P and histamine release after administration of capsaicin (10"4 M), to stimulate C-fiber,
  5      and carbachol (10"4 M), a cholinergic agonist in rabbit intratracheally instilled with UFP
  6      (Nemmar et al,  1999). A significant increase in histamine release also was recorded in the
  7      UFP-instilled group following the administration of both Substance P (10~6 M) plus thiorpan
  8      (10~5 M) and compound 48/80 (C48/80, 10"3 M) to stimulate mast cells.  BAL analysis showed an
  9      influx of PMN, an increase in total protein concentration, and an increase in lung wet weight/dry
10      weight ratio. Electron microscopy showed that both epithelial and endothelial injuries were
11      observed.  The pretreatment of rabbits in vivo with a mixture of either SR 140333 and SR 48368,
12      a tachykinin NK, and  NK2 receptor antagonist, or a mixture of terfenadine and cimetidine,
13      a histamine H, and H2 receptor antagonist, prevented UFP-induced PMN influx and increased
14      protein and lung WW/DW ratio.
15           As discussed earlier, it is believed that ultrafine particles caused greater cellular injury
16      because of the relatively large surface area for a given mass.  However, in a study that compared
17      the response to carbon black particles of two different sizes, Li et al. (1999) demonstrated that in
18      the instillation model, a localized dose of particle over a certain level causes the particle mass to
19      dominate the response, rather than the  surface area. Ultrafine carbon black (ufCB, Printex 90),
20      14 nm in diameter, and fine carbon black (CB, Huber 990), 260 nm in diameter, were instilled
21      intratracheally in rats and BAL profile at 6 h was assessed. At mass of 125 /^g or below, ufCB
22      generated a greater response (increase  LDH, epithelial permeability, decrease in GSH, TNF, and
23      NO productions) than fine CB at various time postexposure.  However, higher dose of CB caused
24      more PMN influx than the ufCB. In contrast to the effect of CB, which showed dose-related
25      increasing inflammatory response, ufCB at the highest dose caused less of a neutrophil influx
26      than at the lower dose. Moreover, when the PMN influx was expressed as a function of surface
27      area, CB produced greater response than ufCB at all doses used in this study. Although particle
28      insterstitialization with a consequent change in the chemotatic gradient for PMN was offered  as
29      an explanation, these results need further scrutinization.
30           Oberdorster et al. (2000) recently completed a series of studies in rats and mice using
31      ultrafine particles of various chemical compositions (PTFE, TiO2, C, Fe, Fe2O3, Pt, V, and V2O5).

        March 2001                               8-70        DRAFT-DO NOT QUOTE OR CITE

-------
  1     In old rats sensitized with endotoxin and exposed to ozone plus ultrafine carbon particles, they
  2     found a ninefold greater release of reactive oxygen species in old rats than in similarly treated
  3     young rats.  Exposure to ultrafine PM alone in sensitized old rats also caused an inflammatory
  4     response.
  5           Although the exact mechanism of ultrafine-induced lung injury remains unclear, it is likely
  6     that ultrafine particles, because of their small size, can easily penetrate the airway epithelium and
  7     cause cellular damage. Using electron microscopy to examine rat tracheal explants treated with
  8     fine (0.12 /urn) and ultrafine (0.021 ^trn) TiO2 partilces for 3 or 7 days, Churg et al. (1998) found
  9     both size particles in the epithelium at both time points, but in the subepithelial tissues, they were
 10     found only at Day 7. The volume proportion (the volume of TiO2 over the entire volume of
 11     epithelium or subepithelium area) of both fine and ultrafine particles in the epithelium increased .
 12     from  3 to 7 days.  It was greater for ultrafine at 3  days but was greater for fine at 7 days. The
 13     volume proportion of particles in the subepithelium at day 7 was equal for both particles, but the
 14     ratio of epithelial to subepithelial volume proportion was 2:1 for fine and 1:1 for ultrafine.
 15     Ultrafine particles persist in the tissue as relatively large aggregates, whereas the size of fine
 16     particle aggregates becomes smaller over time. Ultrafine particles appear to enter the epithelium
 17     faster and, once in the epithelium, a  greater proportion of them is translocated to the subepithelial
 18     space compared to fine particles. However, if it is assumed that the volume proportion is
 19     representative of particle number, the number of particles reaching the interstitial space is
 20     directly proportional to the number applied (i.e., there is no preferential transport from lumen to
 21     interstitium by size). These data are in direct contrast to the results of instillation or inhalation of
 22     fine and ultrafine TIO2 particles reported earlier (Ferin et al., 1990, 1992). Free of inflammatory
 23     cells, possibility of overloading of the explants with dust, and the use of liquid suspension for
 24     exposure were among the possible reasons  cited for the observed effects.
 25          Only two studies examined the influence of specific surface area on biological activity
26     (Lison et al., 1997; Oettinger et al., 1999).  The biological responses to various MnO2 dusts with
27     different  specific surface area  (0.16,  0.5, 17, and 62 m2/g) were compared in vitro and in vivo
28      (Lison et al., 1997). In both systems, the results show that the amplitude of the response is
29      dependent on the total surface area that is in contact with the biological system, indicating that
30      surface chemistry phenomena  are involved  in the biological reactivity. Freshly ground particles
31      with a specific surface area of 5 m2/g also were examined in vitro. These particles exhibited an

        March 2001                                8-71         DRAFT-DO NOT QUOTE OR CITE

-------
 1      enhanced cytotoxic activity, which was almost equivalent to that of particles with a specific
 2      surface area of 62 m2/g, indicating that undefined reactive sites produced at the particle surface
 3      by mechanical cleavage also may contribute to the toxicity of insoluble particles. In another
 4      study, two types of carbon black particles, Printex 90 (P90, Degussa, Germany, formed by
 5      controlled combustion, consists of defined granules with specific surface area of 300 m2/g and
 6      particle size of 14 nm) and FR 101 (Degussa, Germany, with specific surface area of 20 m2/g and
 7      particle size of <95 nm, has a coarse structure, and the ability to adsorb polycyclic and other
 8      carbons) were used in the study (Oettinger et al., 1999). Exposure of AMs to 100 fj.g/106 cells of
 9      FR 101 and P90 resulted in a 1.4- and 2.1-fold increase in ROS release. These exposures also
10      caused a fourfold up-regulation of NF-kB gene expression.  These studies indicated that PM of
11      single component with larger surface properties produce greater biological response than similar
12      particles with smaller surface area.  By exposing bovine AMs to metal oxide coated silica
13      particles, Schluter et al. (1995) showed that most of the metal coatings (Li, Cr, Fe, Mn, Ni, Ph,
14      and V) had no effect on ROS production by these cells.  However, coating with CuO markedly
15      lowered the O2" and H2O2, whereas V(IV) increases both ROI. This study demonstrated that,  in
16      addition to specific area, chemical composition of the particle surface also influence its cellular
17      response.
18
19      8.5.5 Pathophysiological Mechanisms for the Effects of Low Concentrations
20            of Particulate Air Pollution
21           The pathophysiological mechanisms involved in PM-associated cardiovascular and
22      respiratory health effects still are not elucidated fully, but progress has been made since the 1996
23      PM AQCD (U. S. Environmental Protection Agency, 1996a) was prepared. This section
24      summarizes current hypotheses and reviews the toxicological evidence for these potential
25      pathophysiological mechanisms.
26
27      8.5.5.1  Direct Pulmonary Effects
28           When the 1996 PM AQCD (U. S. Environmental Protection Agency, 1996a) was written,
29      the lung was thought  to be the primary organ to affected by particulate air pollution.  There is
30      growing toxicological and epidemiological evidence that the cardiovascular system is affected as
31      well.  Nonetheless, understanding how particulate  air pollution causes or exacerbates respiratory

        March 2001                               8-72        DRAFT-DO NOT QUOTE OR CITE

-------
  1     disease remains an important goal. There is some toxicological evidence for the following three
  2     mechanisms for direct pulmonary effects.
  3
  4     Paniculate Air Pollution Causes Lung Injury and Inflammation
  5          In the last few years, numerous studies have shown that instilled and inhaled ROFA, a
  6     product of fossil fuel combustion, can cause substantial lung injury and inflammation. The toxic
  7     effects of ROFA are largely caused by its high content of soluble metals, and the pulmonary
  8     effects of ROFA can be reproduced by equivalent exposures to soluble metal salts.  In contrast,
  9     controlled exposures of animals to sulfuric acid aerosols, acid coated carbon, and sulfate salts
 10     cause little lung injury or inflammation, even at high concentrations. Inhalation of concentrated
 11     ambient PM (which contains only small amounts of metals) by laboratory animals at
 12     concentrations in the range of 100 to 1000 /^g/m3 have been shown in some (but not all) studies
 13     to cause mild pulmonary injury and inflammation.  Rats with SO2-induced bronchitis and
 14     monocrotaline-treated rats have been reported to have a greater inflammatory response to
 15     concentrated ambient PM than normal rats. These studies suggest that exacerbation of
 16     respiratory disease by ambient PM may be caused in part by lung injury and inflammation.
 17
 18     Particulate Air Pollution Causes Increased Susceptibility to Respiratory Infections
 19          At this time there are no newly published studies on the effects of inhaled concentrated
 20     ambient PM on host susceptibility to infectious agents. Ohtsuka et al. (2000a,b) have shown that
 21      in vivo exposure of mice to acid-coated carbon particles at a mass concentration of 10,000 /ug/m3
 22     causes decreased phagocytic activity of alveolar macrophages, even in the absence of lung injury.
 23
 24     Particulate Air Pollution Increases Airway Reactivity and Exacerbates Asthma
 25           The strongest evidence supporting this hypothesis is from studies on diesel particulate
26      matter (DPM).  DPM has been shown to increase production of antigen-specific IgE in mice and
27      humans (summarized in Section 8.2.4.2). In vitro studies have suggested that the organic
28      fraction of DPM is involved in the increased IgE production. ROFA leachate also has been
29      shown to enhance antigen-specific airway reactivity in mice (Goldsmith et al., 1999) indicating
30      that soluble metals can also enhance an allergic response. However, in this same study, exposure
31      of mice to concentrated ambient PM did not affect antigen-specific airway reactivity. It is

        March 2001                               8-73        DRAFT-DO NOT QUOTE OR CITE

-------
  1      premature to conclude from this one experiment that concentrated ambient PM does not
  2      exacerbate allergic airways disease because the chemical composition of the PM (as indicated by
  3      studies with DPM and ROFA) may be more important than the mass concentration.
  4
  5      8.5.5.2 Systemic Effects Secondary to Lung Injury
  6           When the 1996 PM AQCD was written, it was thought that cardiovascular-related
  7      morbidity and mortality most likely would be secondary to impairment of oxygenation or some
  8      other consequence of lung injury and inflammation. Newly available toxicologic studies provide
  9      some additional evidence regarding such possibilities.
10
11      Lung Injury from Inhaled Paniculate Matter Causes Impairment of Oxygenation and
12      Increased Work of Breathing That Adversely Affects the Heart
14           Instillation of ROFA has been shown to cause a 50% mortality rate in monocrotaline-
15      treated rats (Watkinson et al., 2000).  Although blood oxygen levels were not measured in this
16      study, there were ECG abnormalities consistent with severe hypoxemia in about half of the rats
17      that subsequently died. Given the severe inflammatory effects of instilled ROFA and the fact
18      that monocrotaline-treated rats have increased lung permeability as well as pulmonary
19      hypertension, it is plausible that instilled ROFA can cause severe hypoxemia leading to death in
20      this rat model. Results from studies in which animals (normal and compromised) were exposed
21      to concentrated ambient PM (at concentrations many times higher than would be encountered in
22      the United States) indicate that ambient PM is unlikely to cause severe disturbances in
23      oxygenation or pulmonary function. However, even a modest decrease in oxygenation can have
24      serious consequences in individuals with ischemic heart disease.  Kleinman et al.  (1998) has
25      shown that a reduction in arterial blood saturation from 98 to 94% by either mild hypoxia or by
26      exposure to 100 ppm CO significantly reduced the time to onset of angina in exercising
27      volunteers. Thus, information is needed on the effects of PM on arterial blood gases and
28      pulmonary function to fully address the above hypothesis.
29
30      Lung Inflammation and Cytokine Production Cause Adverse Systemic Hemodynamic Effects
31           It has been suggested that systemic effects of particulate air pollution may result from
32      activation of cytokine production in the lung (Li et al., 1997). In support of this idea,

        March 2001                               8-74       DRAFT-DO NOT QUOTE OR CITE

-------
  1      monocrotaline-treated rats exposed to inhaled ROFA (15,000 /ug/m3, 6 h/day for 3 days) showed
  2      increased pulmonary cytokine gene expression, bradycardia, hypothermia, and increased
  3      arrhythmias (Watkinson et al., 2000). However, spontaneously hypertensive rats had a similar
  4      cardiovascular response to inhaled ROFA (except that they also developed ST segment
  5      depression) with no increase in pulmonary cytokine gene expression. Studies in dogs exposed to
  6      concentrated ambient PM showed minimal pulmonary inflammation and no positive staining for
  7      IL-8, IL-1, or TNF in airway biopsies.  However, there was a significant decrease in the time of
  8      onset of ischemic ECG changes following coronary artery occlusion in PM-exposed dogs
  9      compared to controls (Godleski et al., 2000). Thus, there is not a clear-cut link between changes
 10      in cardiovascular function and production of cytokines in the lung. Because human and animal
 11      exposure studies of ambient PM are using increasingly sophisticated and sensitive measures of
 12      cardiac function, basic information  on the effects of mild pulmonary injury on these cardiac
 13      endpoints is needed to understand the mechanisms of how inhaled PM affects the heart.
 14
 15      Lung Inflammation from Inhaled Particulate Matter Causes Increased Blood Coagulability
 16      That Increases the Risk of Heart Attacks and Strokes
 18          There is abundant evidence linking risk of heart attacks and strokes to small prothrombotic
 19      changes in the blood coagulation system. However, the published toxicological evidence that
 20      moderate lung inflammation causes increased blood coagulability is inconsistent. Ohio et al.
 21      (2000) have shown that inhalation of concentrated ambient PM in healthy nonsmokers causes
 22      increased levels of blood fibrinogen. Gardner et al. (2000) have shown that a high dose
 23      (8,300 /wg/kg) of instilled ROFA in rats causes increased levels of fibrinogen, but no effect was
 24      seen at lower doses. Exposure of dogs to concentrated ambient PM had no effect on fibrinogen
 25      levels (Godleski et al., 2000). The coagulation system is as multifaceted and complex as the
 26      immune system, and there are many other sensitive and clinically significant parameters that
27      should be examined in addition to fibrinogen. Thus, it is premature to draw any conclusions on
28      the relationship between PM and blood coagulation.
29
30      Interaction of Particulate Matter with the Lung Affects Hematopoiesis
31          Terashima et al. (1997) found  that instillation of fine carbon particles (20,000 //g/rabbit)
32      stimulated release of PMNs from the bone marrow. In further support of this hypothesis, Gordon

        March 2001                              8-75        DRAFT-DO NOT QUOTE OR CITE

-------
 1     and colleagues reported that the percentage of PMNs in the peripheral blood increased in rats
 2     exposed to ambient PM in some but not all exposures. On the other hand, Godleski et al. (2000)
 3     found no changes in peripheral blood counts of dogs exposed to concentrated ambient PM.
 4     Thus, direct evidence that PM ambient concentrations can affect hematopoiesis remains to be
 5     demonstrated.
 6
 7     8.5.5.3  Direct Effects on the Heart
 8          Changes in heart rate, heart rate variability, and conductance associated with ambient PM
 9     exposure have been reported in animal studies (Godleski et al., 2000; Gordon et al., 2000;
10     Watkinson et al., 2000), in several human panel studies (described in Chapter 6), and in a
11     reanalysis of data from the MONICA study (Peters et al., 1997). Some of these  studies included
12     endpoints related to respiratory effects but few significant adverse respiratory changes were
13     detected.  This raises the possibility that ambient PM may have effects on the heart that are
14     independent of adverse changes in the lung.  There is certainly precedent for this idea.
15     For example, tobacco smoke (which is a mixture of combustion-generated gases and PM) causes
16     cardiovascular disease by mechanisms  that are independent of its effect on the lung.  Two types
17     of hypothesized direct effects of PM on the heart are noted below.
18
19     Inhaled Paniculate Matter Affects the Heart by Uptake of Particles into the Circulation or
20     Release of a Soluble Substances into the Circulation.
21
22          Drugs can be rapidly and efficiently delivered to the systemic circulation by inhalation.
23     This implies that the pulmonary vasculature  absorbs inhaled materials, including charged
24     substances such as small proteins and peptides. Cigarettes are a widely used method for
25     delivering nicotine to the blood stream. It is likely that soluble materials absorbed onto airborne
26     particles find their way into the blood stream, but it is not clear whether the particles themselves
27     enter the blood.  It is anticipated that more information will be available on this  important
28     question in the next few years.
29
30     Inhaled Particulate Matter Affects Autonomic Control of the Heart and Cardiovascular
31     System
33          There is growing evidence for this idea as described above. This raises the question of how
34     inhaled particles could affect the autonomic  nervous system.  Activation of neural receptors in
       March 2001                               8-76       DRAFT-DO NOT QUOTE OR CITE

-------
  1      the lung is a logical area to investigate. Studies in conscious rats have shown that inhalation of
  2      wood smoke causes marked changes in sympathetic and parasympathetic input to the
  3      cardiovascular system that are mediated by neural reflexes (Nakamura and Hayashida, 1992).
  4      Although research on airway neural receptors and neural-mediated reflexes is a well established
  5      discipline, the cardiovascular effects of stimulating airway receptors continue to receive less
  6      attention than the pulmonary effects. Previous studies of airway reflex-mediated cardiac effects
  7      usually employed very high doses of chemical irritants, and the results may not be applicable to
  8      air pollutants. There is a need for basic physiological studies to examine effects on
  9      cardiovascular system when airway and alveolar neural receptors are stimulated in a manner
 10      relevant to air pollutants.
 11
 12
 13      8.6  RESPONSES TO PARTICULATE MATTER AND GASEOUS
 14           POLLUTANT MIXTURES
 15           Ambient PM itself is a mixture of particles of varying size and composition. The following
 16      discussion examines effects of mixtures of ambient PM, or PM surrogates, with gaseous
 17      pollutants. Ambient PM co-exists in indoor and outdoor air with a number of co-pollutant gases,
 18      including ozone, sulfur dioxide, oxides of nitrogen, and  carbon monoxide.  Toxicological
 19      interactions between PM and gaseous co-pollutants may be antagonistic, additive, or synergistic
 20      (Mauderly, 1993). The presence and nature of any interaction appears to depend on 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 8-10).
 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      underlie 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

        March 2001                               8-77        DRAFT-DO NOT QUOTE OR CITE

-------
                        TABLE 8-10. RESPIRATORY AND CARDIOVASCULAR EFFECTS OF MIXTURES
C- Species,
^ Gender, Strain
O Age, or Body
2 Weight
Rats, Fischer
NNia, male,
22 to 24 mo
old





Rats






Gases and Exposure
Particles Technique
Carbon, Inhalation
ammonium
bisulfate,
and O3





O3 and Ottawa Inhalation
urban dust





Mass Particle
Concentration Size
50 A
-------
TABLE 8-10 (cont'd). RESPIRATORY AND CARDIOVASCULAR EFFECTS OF MIXTURES
o
N)
0
o
H— *













oo
~Lj
^o



O
i
H
6
o
z
o
H
O
a,
o
H
M
o
o
H
W
Species,
Gender, Strain
Age, or Body Gases and
Weight Particles
Rats H,SO4 and O3




Healthy and H2SO4,
asthmatic SO,, and 03
children

Pigeons Ambient gases
(Columba and particles
livia)

Canine Ambient gases
and particles





Rats 03 and
resuspended
urban PM















Exposure Particle
Technique Mass Concentration Size
Inhalation, 20 to 1 50 Aig/W 0.4 to 0 8 ^m
whole body H,SO4 and 0. 1 2 or
0.2 ppm O3


Inhalation 60 to 1 40 pig/m3 06^mIl,SO4
H2SO4, 0.1 ppm SO,,
and 0. 1 ppm O3

Natural 24-h
exposure in
urban and
rural areas
Natural 24-h
exposure in
four urban
areas of
Mexico City
and one
rural area
Inhalation, 0 8 ppm O3 and
whole-body 5,000 or
50,000 ^g/m3 PM















Exposure
Duration
Intermittent
(12h/day)or
continuous
exposure for up
to 90 days
Single 4-h
exposure with
intermittent
exercise
Continuous
ambient exposure


Continuous
ambient exposure





Single 4-h
exposure
















Respiratory Effects of Inhaled Particles on Markers
m Lavage Fluid
No interactive effect of H,SO4 and O3 on
biochemical and morphometnc endpoints.



A positive association between acid concentration
and symptoms, but not spirometry, in asthmatic
children. No changes in healthy children.

Increased number of AMs and decreased number of
lamellar bodies in type II epithelial cells in urban
pigeons.

No significant differences in AMs or total cell
counts in lavage from dogs studied among the
five regions. A significant increase in lavage fluid
neutrophils and lymphocytes in the southwest
region, where the highest O, levels were recorded,
compared to the two industnal 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 prohferative changes
induced by O3 alone. These changes were greatest
in the epithelium of the terminal bronchioles and
alveolar ducts













Reference
Last and Pmkerton
(1997)



Linn et al.
(1997)


Lorz and Lopez
(1997)


Vanda et al.
(1998)





Vincent et al.
(1997)















-------
  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 that may be more active lexicologically than the primary materials
  5      and that can then be carried to a sensitive site.  The hypothesis of such chemical interactions has
  6      been examined in the gas and particle exposure studies by Amdur and colleagues (Amdur and
  7      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.
1 1           Another potential mechanism of gas-particle interaction may involve a pollutant-induced
12      change in the local microenvironment of the lung, enhancing the effects of the co-pollutant.
13      For example, Last et al. (1984) suggested that the observed synergism between ozone and acid
14      sulfates in rats was due to a decrease in the local microenvironmental pH of the lung following
1 5      deposition of acid, enhancing the effects of ozone by producing a change in the reactivity or
1 6      residence time of reactants, such as radicals, involved in ozone-induced tissue injury.
1 7           As noted in U.S. Environmental Protection Agency  (1996a), the toxicology database for
1 8      mixtures containing PM other than acid sulfates is still quite sparse. Vincent et al. (1997)
19      exposed rats to 0.8 ppm ozone in combination with 5 or 50 mg/m3 of resuspended urban particles
20      for 4 h. Although PM alone caused no change in cell proliferation (3H-thymidine labeling),
2 1      co-exposure to either concentration of resuspended PM with ozone greatly potentiated the
22      proliferative effects of exposure to ozone alone.  These interactive changes occurred in epithelial
23      cells of the terminal bronchioles and the alveolar ducts. These findings using resuspended dusts,
24      although at high concentrations, are consistent with studies demonstrating interaction between
25      sulfuric acid (H2SO4) aerosols and ozone.  Kimmel and colleagues (1997) examined the effect of
26      acute co-exposure to ozone and fine or ultrafine H2SO4 aerosols on rat lung morphology. They
27      determined morphometrically that alveolar septal volume  was increased in animals  co-exposed to
28      ozone and ultrafine, but not fine, H2SO4. Interestingly, cell labeling, an index  of proliferative cell
29      changes, was increased only in animals co-exposed to fine H2SO4 and ozone, as compared to
30      animals exposed to ozone alone.  Importantly, Last and Pinkerton (1997)  extended their previous
3 1      work and found that subchronic exposure to acid aerosols (20 to 1 50 A^g/m3 H2SO4) had no
       March 2001                               8-80        DRAFT-DO NOT QUOTE OR CITE

-------
  1      interactive effect on the biochemical and morphometric changes produced by either intermittent
  2      or continuous ozone exposure (0.12 to 0.2 ppm).  Thus, the interactive effects of ozone and acid
  3      aerosol co-exposure in the lung disappeared during the long-term exposure.
  4           Kleinman et al. (1999) examined the effects of ozone plus fine, H2SO4-coated, carbon
  5      particles (MMAD = 0.26 /^m) for 1 or 5 days. They found the inflammatory response with the
  6      ozone-particle mixture was greater after 5 days (4 h/day) than after Day 1. This contrasted with
  7      ozone exposure alone (0.4 ppm), which caused marked inflammation on acute exposure, but no
  8      inflammation after 5 consecutive days of exposure.
  9           Studies have examined interaction between carbon particles and gaseous co-pollutants.
 10      Jakab et al. (1996) challenged mice with a single 4-h exposure to a high concentration of carbon
 11      (10 mg/m3) in the presence of SO2 at low and high relative humidities.  Macrophage phagocytosis
 12      was depressed significantly only in mice exposed to the combined pollutants under high relative
 13      humidity conditions. This study demonstrates that fine carbon particles can serve as an effective
 14      carrier for acidic sulfates where chemical conversion of adsorbed SO2 to acid sulfate species
 15      occurred. Interestingly, the depression in macrophage function was present as late as 7  days
 16      postexposure. Bolarin et al. (1997) exposed rats to only 50 or 100 /ug/m3 carbon particles in
 17      combination with ammonium bisulfate and ozone. Despite 4 weeks  of exposure, they observed
 18      no changes in protein concentration in lavage fluid or blood prolyl 4-hydroxylase, an enzyme
 19      involved in collagen metabolism.  Slight decreases in plasma fibronectin were present in animals
 20      exposed to the combined pollutants versus ozone  alone. Thus as, previously noted, the  potential
 21      for adverse effects in the lungs of animals challenged with a combined exposure to particles and
 22      gaseous pollutants is dependent on numerous factors, including the gaseous co-pollutant,
 23      concentration, and time.
 24           In a complex series of exposures, Oberdorster and colleagues examined the interaction of
 25      ultrafine carbon particles (100 /wg/m3) and ozone (1 ppm) in young and old Fischer 344  rats that
 26      were pretreated with aerosolized endotoxin (Elder et al., 2000). In old rats, exposure to carbon
27      and ozone produced an interaction that resulted in a greater influx in neutrophils than that
28      produced by either agent alone. This interaction was not seen in young rats.  Oxidant release
29      from lavage fluid cells was also assessed and the combination of endotoxin, carbon particles, and
30      ozone produced an increase in oxidant release in old rats.  This combination produced the
31      opposite response in the cells recovered from the lungs of the young  rats, indicating that the

        March 2001                                8-81         DRAFT-DO NOT QUOTE OR CITE

-------
 1      lungs of the aged animals underwent greater oxidative stress in response to this complex
 2      pollutant mix of particles, ozone, and a biogenic agent.
 3           Linn and colleagues (1997) examined the effect of a single exposure to 60 to 140 /wg/m3
 4      H2SO4, 0.1 ppm SO2, and 0.1 ppm ozone in healthy and asthmatic children. The children
 5      performed intermittent exercise during the 4-h exposure to increase the inhaled dose of the
 6      pollutants. An overall effect on the combined group  of healthy and asthmatic children was not
 7      observed. A positive association between acid concentration and symptoms was seen, however,
 8      in the subgroup of asthmatic children. The combined pollutant exposure had no effect on
 9      spirometry in asthmatic children, and no changes in symptoms or spirometry were observed in
10      healthy children. Thus, the effect of combined exposure to PM and gaseous co-pollutants
11      appeared to have less effect on asthmatic children exposed under controlled laboratory conditions
12      in comparison with field studies of children attending summer camp (Thurston et al., 1997).
13      However, prior exposure to H2SO4 aerosol may enhance the subsequent response to ozone
14      exposure (Linn et al., 1994; Frampton et al., 1995); the timing and sequence of the exposures
15      may be important.
16           Three unique animal field studies have examined the adverse respiratory effects of complex
17      mixtures in urban and rural environments.  These studies have taken advantage of the differences
18      in pollutant makeup of urban and rural environments and studied animals under natural,
19      continuous exposure conditions. Gulisano et al. (1997) examined the morphologic changes
20      produced by continuous ambient exposure to air pollutants in lambs raised for 3 mo in rural
21      (n = 2) or urban (n = 10) environments. Compared to the lungs of the rural lambs, irritation, as
22      characterized by mucus hypersecretion and morphological changes in the epithelial cells lining
23      the nasopharyngeal region, was present in the lambs exposed to urban air pollution.  Lorz and
24      Lopez (1997) performed a similar study using pigeons as the test animal.  They observed an
25      increase in the number of AMs and a decrease in the  number of lamellar bodies in Type II
26      epithelial cells in the lungs of urban pigeons. Extrapolation of these studies is hampered by an
27      incomplete characterization of the exposure atmospheres. A more thorough examination of the
28      ambient level of pollutants was performed in the study  by Vanda et al. (1998), who studied the
29      effect of pollutant exposure in dogs raised in four urban regions of Mexico City and one nearby
30      rural area.  They found no significant differences in AM number or total cell counts in lavage
31      fluid from the dogs among the five regions. A significant increase in lavage fluid neutrophils and

        March 2001                               8-82       DRAFT-DO NOT QUOTE OR CITE

-------
  1      lymphocytes was found in dogs from the urban region with the highest ozone levels in
  2      comparison to the regions with the highest PM levels. Thus, the effect of ozone on cellular
  3      parameters in lavage fluid appeared to be greater than that for PM. In summary, each of these
  4      three animal field studies provides evidence that urban air pollutants can produce greater lung
  5      changes than would occur from exposure to rural pollution. However, extrapolation of these
  6      results is severely hampered by the uncontrolled exposure conditions, small sample size,
  7      behavior patterns, and nutritional factors. Thus, in these field studies, it is difficult to assign a
  8      role to PM in the observed adverse pulmonary effects.
  9
 10
 11      8.7  SUMMARY
 12      8.7.1 Biological Plausibility
 13           Toxicological studies can play an integral role in answering the following two key
 14      questions regarding biological plausibility of PM health effects.
 15         (1) What component (or components) of ambient PM cause health effects?
 16         (2) Are the statistical associations between PM and health effects biologically plausible?
 17      This summary focuses on the progress that toxicological studies have made towards answering
 18      these questions.
 19
 20      8.7.1.1  Link Between Specific Participate Matter Components and Health Effects
 21           Key to the validity of the biological plausibility is the need to identify the components of
 22      airborne PM responsible for the adverse effects and the individuals at risk. The plausibility of
 23      the association between PM and increases in morbidity and mortality has been questioned
 24      because the adverse cardiopulmonary effects have been observed at very low PM concentrations,
 25      often below the current NAAQS for PM10.  To date, toxicology studies on PM have provided
 26      only very limited evidence for specific PM components being responsible for observed
 27      cardiopulmonary effects of ambient PM.  Studies have shown that some components of particles
28      are more toxic than others.  For example, high concentrations of ROFA and associated soluble
29      metals have produced clinically significant effects (including death) in compromised animals.
30      The relevance of these findings to understanding the adverse effects of PM components is

        March 2001                               8-83         DRAFT-DO NOT QUOTE OR CITE

-------
 1     tempered, however, by the large difference between metal concentrations delivered to the test
 2     animals and metal concentrations present in the ambient urban environment.  Such comparisons
 3     must be applied to the interpretation of all studies that examine the individual components of
 4     ambient urban PM.  A summary of potential contributions of individual physical/chemical factors
 5     of particles to cardiopulmonary effects is given below.
 6
 7     Acid Aerosols
 8          There is relatively little new information on the effects of acid aerosols, and the conclusions
 9     of the  1996 PM AQCD are unchanged.  It was previously concluded that acid aerosols cause
10     little or no change in pulmonary function in healthy subjects, but asthmatics may develop small
11     changes in pulmonary function. This conclusion is supported by the recent study of Linn and
12     colleagues (1997) in which children (26 children with allergy or asthma and 15 healthy children)
13     were exposed to sulfuric acid aerosol (100 //g/m3) for 4 h. There were no significant effects on
14     symptoms or pulmonary function when data from the entire group was analyzed, but the allergy
15     group  had a significant increase in symptoms after the acid aerosol exposure.
16          Although pulmonary effects of acid aerosols have been the subject of extensive research in
17     past decades, the cardiovascular effects of acid aerosols have received little attention. Zhang
18     et al. (1997) reported that inhalation of acetic acid fumes caused reflex mediated increases in
19     blood  pressure  in normal and spontaneously hypertensive rats. Thus, acid components should
20     not be ruled out totally as possible mediators of PM health effects. In particular, the
21     cardiovascular  effects of acid aerosols at realistic concentrations need further investigation.
22
23     Metals
24          The previous PM AQCD (U.S. Environmental Protection Agency, 1996a) mainly relied on
25     data related to occupational exposures to evaluate the potential toxicity of metals in particulate
26     air pollution. Since that time, in vivo and in vitro studies using ROFA or soluble transition
27     metals have contributed substantial new information on the health effects of particle-associated
28     soluble metals. Although there are some uncertainties about differential effects of one transition
29     metal versus another, water soluble metals leached from ROFA have been shown consistently
30     (albeit at high concentrations) to cause cell injury and inflammatory changes in vitro and in vivo.


       March 2001                                8-84        DRAFT-DO NOT QUOTE OR CITE

-------
  1           Even though it is clear that combustion particles that have a high content of soluble metals
  2      can cause lung injury and even death in compromised animals, it has not been established that the
  3      small quantities of metals associated with ambient PM are sufficient to cause health effects.
  4      Moreover, it cannot be assumed that metals are the primary toxic component of ambient PM.
  5      In studies in which various ambient and emission source particulates were instilled into rats, the
  6      soluble metal content did appear to be the primary determinant of lung injury (Costa and Dreher,
  7      1997).  However, one published study has compared the effects of inhaled ROFA (at 1 mg/m3) to
  8      concentrated ambient PM (four experiments, at mean concentrations of 475 to 900 yUg/m3) in
  9      normal and SO2- induced bronchitic  rats. A  statistically significant increase in at least one lung
 10      injury marker was seen in bronchitic rats with only one out of four of the concentrated ambient
 11      exposures, whereas inhaled ROFA had no effect even though the content of soluble iron,
 12      vanadium, and nickel was much higher in the ROFA sample than in the concentrated ambient
 13      PM. Although the role of metals in contributing to health effects of ambient PM is not
 14      established, the recent studies based  on ROFA have important implications.
 15
 16      Ultraflne Particles
 17           When this subject was reviewed in the  1996 PM AQCD (U. S. Environmental Protection
 18      Agency, 1996a), it was not known whether the pulmonary toxicity of freshly generated ultrafine
 19      teflon particles was due to particle size or a result of absorbed fumes. Subsequent studies with
 20      other types of ultrafine particles have shown  that the chemical constituents of ultrafines
 21      substantially modulate their toxicity.  For example, Kuschner et al. (1997) have established that
 22      inhalation of MgO particles produces far fewer respiratory effects than does ZnO. Also,
 23      inhalation exposure of normal rats to ultrafine carbon  particles generated by electric arc discharge
 24      (100 //g/m3 for 6 h) caused minimal lung inflammation (Elder et al., 2000), compared to ultrafine
 25      Teflon or metal particles. On the other hand, instillation of 125 ^g of ultrafine carbon black
26      (20 nm) caused substantially more inflammation than  did the same dose of fine particles of
27      carbon black (200 to 250 nm), suggesting that ultrafine particles may cause more inflammation
28      than larger particles (Li et al., 1997).  However, the chemical constituents of the two sizes of
29      carbon black used in this study were not analyzed, and it cannot be assumed that the chemical
30      composition was the same for the two sizes.  Thus, there is still insufficient toxicological
31      evidence to conclude that ambient concentrations of ultrafine particles contribute to the health

        March 2001                              8-85       DRAFT-DO NOT QUOTE OR CITE

-------
  1      effects of particulate air pollution.  However, with acid aerosols, studies of ultrafine particles
  2      have focused largely on effects in the lung, and it is possible that inhaled ultrafine particles may
  3      have systemic effects that are independent of effects on the lung.
  4
  5      Bioaerosols
  6           Recent studies support the conclusion of the 1996 PM AQCD (U. S. Environmental
  7      Protection Agency, 1996a), which stated that bioaerosols, at concentrations present in the
  8      ambient environment, would not account for the reported health effects of ambient PM.
  9      Dose-response studies in healthy volunteers exposed to 0.55 and 50 //g endotoxin, by the
10      inhalation route, showed a threshold for pulmonary and systemic effects for endotoxin between
11      0.5 and 5.0 //g (Michel et al., 1997). Monn and Becker (1999) examined effects of size
12      fractionated outdoor PM on human monocytes and found cytokine induction characteristic of
13      endotoxin activity in the coarse-size fraction but not in the fine fraction. Available information
14      suggests that ambient concentrations of endotoxin are very low and do not exceed 0.5 ng/m3.
15
16      Diesel Exhaust Particles
17           As described in Section 8.2.4.2, there is growing toxicological evidence that diesel PM
18      exacerbates the allergic response to inhaled antigens.  The organic fraction of diesel exhaust has
19      been linked to eosinophil degranulation and induction of cytokine production, suggesting that the
20      organic constituents of diesel PM is responsible part for the immune effects. It is not known
21      whether the adjuvant-like activity of diesel PM is unique  or whether other combustion particles
22      have similar effects. It is important to compare the immune effects of other source-specific
23      emissions, as well as concentrated ambient PM, to diesel  PM to determine the extent to which
24      exposure to diesel exhaust may contribute to the incidence and severity of allergic rhinitis and
25      asthma.
26
27      Organic Compounds
28           Published research on the acute effects of particle-associated organic carbon constituents is
29      conspicuous by its relative absence, except for diesel exhaust particles.  Like metals, organics are
30      common constituents of combustion-generated particles and have been  found in ambient PM
31      samples over a wide geographical range. Organic carbon constituents comprise a substantial

        March 2001                                8-86        DRAFT-DO NOT QUOTE OR CITE

-------
  1     portion of the mass of ambient PM (10 to 60% of the total dry mass [Turpin, 1999]). The
  2     organic fraction of ambient PM has been evaluated for its mutagenic effects. Although the
  3     organic fraction of ambient PM is a poorly characterized heterogeneous mixture of an unknown
  4     number of different compounds, strategies have been proposed for examining the health effects
  5     of this potentially important constituent (Turpin, 1999).
  6
  7     Ambient Particle Studies
  8          Ambient particle studies should be the most relevant in understanding the susceptibility of
  9     individuals to PM and the underlying mechanisms.  Studies have used collected urban PM for
 10     intratracheal administration to healthy and compromised animals. Despite the difficulties in
 11     extrapolating from the bolus delivery used in such studies, they have provided strong evidence
 12     that the chemical composition of ambient particles has a major influence on toxicity. More
 13     recent work with inhaled concentrated ambient PM has observed cardiopulmonary changes in
 14     rodents and dogs at high concentrations of fine PM.  No comparative studies to examine the
 15     effects of ultrafine and coarse ambient PM have been done, although a new ambient particle
 16     concentrator developed by Sioutas and colleagues should permit the direct toxicological
 17     comparison of various ambient particle sizes.  Importantly, it has become evident that, although
 18     the concentrated ambient PM studies can provide important dose-response information, identify
 19     susceptibility factors in  animal models, and permit examination of mechanisms related to PM
 20     toxicity, they are not particularly well suited, however, for the identification of toxic components
 21     in urban PM.  Because only a limited number of exposures using concentrated ambient PM can
 22     be reasonably conducted by a given laboratory in a particular urban environment, there may be
 23     insufficient information to conduct a  factor analysis on an exposure/response matrix. This may
 24     also hinder principal component analysis techniques that are useful in identifying particle
25     components responsible for adverse outcomes.
26
27     8.7.1.2 Susceptibility
28          Progress has been  made in understanding the role of individual susceptibility to ambient
29     PM effects.  Studies have consistently shown that animals with compromised health, either
30     genetic or induced, are more susceptible to instilled or inhaled particles, although the increased
31      animal-to-animal variability in these models has created problems.  Moreover, because PM

        March 2001                               8-87        DRAFT-DO NOT QUOTE OR CITE

-------
 1     seems to affect broad categories of disease states, ranging from cardiac arrhythmias to pulmonary
 2     infection, it can be difficult to know what disease models to use in understanding the biological
 3     plausibility of the adverse health effects of PM.  Thus, the identification of susceptible animal
 4     models has been somewhat slow, but overall it represents solid progress when one considers that
 5     data from millions of people are necessary in epidemiology studies to develop the statistical
 6     power to detect small increases in PM-related morbidity and mortality.
 7
 8     8.7.2 Mechanisms of Action
 9           The mechanisms that underlie the biological responses to ambient PM are not clear.
10     Various toxicologic studies using particulate matter having diverse physicochemical
11     characteristics have shown that these characteristics have a great impact on the specific response
12     that is observed. Thus, there may, in fact, be multiple biological mechanisms that may be
13     responsible for observed morbidity/mortality because of exposure to ambient PM, and these
14     mechanisms may be highly dependent on the type  of particle in the exposure atmosphere.
15     However, it should be noted that many controlled  exposure studies used particle concentrations
16     much higher than those typically occurring in ambient air. Thus, some of the mechanisms
17     elicited may not occur with exposure to lower levels.  Clearly, controlled exposure studies have
18     not as yet been able to unequivocally determine the particle characteristics and the toxicological
19     mechanisms by which ambient PM may affect biological systems.
20
       March 2001                               8-88        DRAFT-DO NOT QUOTE OR CITE

-------
   1       REFERENCES

  2       Amdur, M. O.; Chen, L. C. (1989) Furnace-generated acid aerosols: speciation and pulmonary effects.
  3             In: Symposium on the health effects of acid aerosols; October 1987; Research Triangle Park, NC. Environ.
  4             Health Perspect. 79: 147-150.
  5       Amdur, M. O.; Dubriel, M.; Creasia, D. A. (1978) Respiratory response of guinea pigs to low levels of sulfuric acid.
  6             Environ. Res. 15:418-423.
  7       Ball, B. R.; Smith, K. R.; Veranth, J. M.; Aust, A. E. (2000) Bioavailability of iron from coal fly ash: mechanisms
  8             of mobilization and of biological effects. In: Grant, L. D., ed. PM2000: particulate matter and health.
  9             Inhalation Toxicol. 12(suppl. 4): 209-225.
 10       Bayram, H.; Devalia, J. L.; Khair, O. A.; Abdelaziz, M. M.;  Sapsford, R. J.; Sagai, M.; Davies, R. J. (1998a)
 11             Comparison of ciliary activity and inflammatory mediator release from bronchial epithelial cells of nonatopic
 12             nonasthmatic subjects and atopic asthmatic patients and the effect of diesel exhaust particles in vitro.
 13             J. Allergy Clin. Immunol. 102: 771-782.
 14       Bayram, H.; Devalia, J. L.; Sapsford, R. J.; Ohtoshi, T.; Miyabara, Y.; Sagai, M.; Davies, R. J. (1998b) The effect
 15             of diesel exhaust particles on cell function and release of inflammatory mediators from human bronchial
 16             epithelial cells in vitro. Am. J. Respir. Cell Mol. Biol. 18: 441-448.
 17       Becker, S.; Soukup, J. M. (1998) Decreased CD1 IB expression, phagocytosis, and oxidative burst in urban
 18             particulate pollution-exposed human monocytes and alveolar macrophages. J. Toxicol. Environ. Health Part
 19             A 55: 455-477.
 20       Becker, S.; Soukup, J. M.; Gilmour, M. I.; Devlin, R. B. (1996) Stimulation of human and rat alveolar macrophages
 21             by urban air particulates: effects on oxidant radical generation and cytokine production. Toxicol. Appl.
 22             Pharmacol.  141:637-648.
 23       Blomberg, A.; Sainsbury, C.; Rudell, B.; Frew, A. J.; Holgate, S. T.; Sandstrom, T.; Kelly, F. J. (1998) Nasal cavity
 24             lining fluid ascorbic acid concentration increases in healthy human volunteers following short term exposure
 25             to diesel exhaust. Free Radical Res. 28: 59-67.
 26       Bolarin, D. M.; Bhalla, D. K.; Kleinman, M. T. (1997)  Effects of repeated exposures of geriatric rats to ozone and
 27             particle-containing atmospheres: an analysis of bronchoalveolar lavage and plasma proteins. Inhalation
 28             Toxicol. 9: 423-434.
 29       Bonner, J. C.; Rice, A. B.; Lindroos, P. M.; O'Brien, P. O.; Dreher, K. L.; Rosas, I.; Alfaro-Moreno, E.;
 30             Osornio-Vargas, A. R. (1998) Induction of the lung myofibroblast PDGF receptor system by urban ambient
 31             particles from Mexico City. Am. J. Respir. Cell Mol. Biol. 19: 672-680.
 32       Bouthillier, L.; Vincent, R.; Goegan, P.; Adamson, I. Y. R.; Bjarnason, S.; Stewart, M.; Guenette, J.; Potvin, M.;
 33             Kumarathasan, P. (1998) Acute effects of inhaled urban particles and ozone: lung morphology, macrophage
 34             activity, and plasma endothelin-1. Am. J. Pathol. 153: 1873-1884.
 35       Brain, J. D.; Long, N. C.; Wolfthal, S. F.; Dumyahn, T.; Dockery, D. W. (1998) Pulmonary toxicity in hamsters of
 36             smoke particles from Kuwaiti oil fires. Environ.  Health Perspect. 106: 141-146.
 37       Broeckaert, F.; Buchet, J.  P.; Huaux, F.; Lardot, C.; Lison, D.; Yager, J. W. (1997) Reduction of the ex vivo
 3 8            production of tumor necrosis factor alpha by alveolar phagocytes after administration of coal fly ash and
 39            copper smelter dust. J. Toxicol. Environ. Health  51:189-202.
 40       Campen, M. J.; Costa, D. L.; Watkinson, W. P. (2000) Cardiac and thermoregulatory toxicity of residual oil fly ash
 41            in cardiopulmonary-compromised rats. Inhalation Toxicol. 12(suppl. 2): 7-22.
 42       Carter, J. D.; Ohio, A. J.; Samet, J. M.; Devlin, R. B. (1997)  Cytokine production by human airway epithelial cells
 43            after exposure to an air pollution particle is metal-dependent. Toxicol. Appl. Pharmacol. 146:  180-188.
 44       Chen, L. C.; Fine, J. M.; Qu, Q.-S.; Amdur, M. O.; Gordon, T. (1992) Effects of fine and ultrafine sulfuric acid
 45            aerosols in guinea pigs: alterations in alveolar macrophage function and intracellular pH. Toxicol. Appl.
 46            Pharmacol. 113: 109-117.
47       Chen, L. C.; Fang, C. P.; Qu, Q. S.; Fine, J. M.; Schlesinger,  R. B. (1993) A novel system for the in vitro exposure
48            of pulmonary cells to acid sulfate aerosols. Fundam. Appl. Toxicol. 20: 170-176.
49      Churg, A.; Stevens, B.; Wright, J. L. (1998) Comparison of the uptake of fine and ultrafine TiO2 in a tracheal
 50            explant system. Am. J. Physiol. 274: L81-L86.
 51       Clarke, R. W.; Catalano, P. J.; Koutrakis, P.; Krishna Murthy, G. G.;  Sioutas, C.; Paulauskis, J.; Coull, B.;
 52            Ferguson, S.; Godleski, J. J. (1999) Urban air particulate inhalation alters pulmonary function and induces
53             pulmonary inflammation in a rodent model of chronic  bronchitis. Inhalation Toxicol. 11: 637-656.
         March 2001                                     8-89         DRAFT-DO NOT QUOTE OR CITE

-------
  1       Clarke, R. W.; Catalano, P.; Coull, B.; Koutrakis, P.; Krishna Murthy, G. G.; Rice, T.; Godleski, J. J. (2000)
 2             Age-related responses in rats to concentrated urban air particles (CAPs). In: Inhalation Toxicology:
 3             proceedings of the third colloquium on particulate air pollution and human health; June, 1999; Durham, NC.
 4             Inhalation Toxicology 129(suppl. 1): 73-84.
 5       Cohen, M. D.; Becker, S.; Devlin, R.; Schlesinger, R. B.; Zelikoff, J. T. (1997) Effects of vanadium upon
 6             polyl:C-induced responses in rat lung and alveolar macrophages. J. Toxicol. Environ. Health 51: 591-608.
 7       Cormier, Y.; Laviolette, M.; Bedard, G.; Dosman, J.; Israel-Assayag, E. (1998) Effect of route of breathing on
 8             response to exposure in a swine confinement building. Am. J. Respir. Crit. Care Med. 157: 1512-1521.
 9       Costa, D. L.; Dreher, K. L. (1997) Bioavailable transition metals in particulate matter mediate cardiopulmonary
10             injury in healthy and compromised animal models. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of
11             the sixth international meeting on the toxicology of natural and man-made fibrous and non-fibrous particles;
12             September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1053-1060.
13       Creutzenberg, O.; Bellmann, B.; Muhle, H.; Dasenbrock, C.; Morrow, P.; Mermelstein, R. (1998) Lung clearance
14             and retention of toner, TiO2, and crystalline silica, utilizing a tracer technique during chronic inhalation
15             exposure in Syrian Golden hamsters. Inhalation Toxicol. 10: 731-751.
16       Diaz-Sanchez, D.; Tsien, A.; Casillas, A.; Dotson, A. R.; Saxon, A. (1996) Enhanced nasal cytokine production in
17             human beings after in vivo challenge with diesel exhaust particles. J. Allergy Clin.  Immunol. 98: 114-123.
18       Diaz-Sanchez, D.; Tsien, A.; Fleming, J.; Saxon, A. (1997)  Combined diesel exhaust particulate and ragweed
19             allergen challenge markedly enhances human in vivo nasal ragweed-specific IgE and skews cytokine
20             production to a T helper cell 2-type pattern. J. Immunol. 158: 2406-2413.
21       Donaldson, K.; Brown, D. M.; Mitchell, C.; Dineva,  M.; Beswick, P. H.; Gilmour, P.; MacNee, W. (1997) Free
22             radical activity of PM10: iron-mediated generation of hydroxyl radicals. In: Driscoll, K. E.; Oberdorster, G.,
23             eds. Proceedings of the sixth international meeting on the toxicology of natural and man-made fibrous and
24             non-fibrous particles; September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5):
25             1285-1289.
26       Dong, W.; Lewtas, J.; Luster, M. I. (1996) Role of endotoxin in tumor necrosis factor ocexpression from alveolar
27             macrophages treated with urban  air particles. Exp. Lung Res. 22: 577-592.
28       Dormans, J. A. M. A.; Steerenberg, P. A.; Arts, J. H. E.; Van Bree, L.; De Klerk, A.; Verlaan, A. P. J.; Bruijntjes,
29             J. P.; Beekhof, P.; Van Soolingen, D.; Van Loveren,  H. (1999) Pathological and immunological effects of
30             respirable  coal fly ash in male wistar rats. Inhalation  Toxicol.  11: 51-69.
31       Dreher, K. L.; Jaskot, R. H.; Lehmann, J. R.; Richards, J. H.; McGee, J. K.; Ohio, A. J.; Costa, D. L. (1997)
32             Soluable transition metals mediate residual oil fly ash induced acute lung injury. J.  Toxicol. Environ. Health
33             50:285-305.
34       Driscoll, K. E.; Costa, D. L.; Hatch, G.; Henderson, R.; Oberdorster, G.; Salem, H.; Schlesinger, R. B. (2000)
35             Intratracheal instillation as an exposure technique for the evaluation of respiratory tract toxicity: uses and
36             limitations. Toxicol. Sci. 55: 24-35.
37       Dye, J. A.; Adler, K. B.; Richards, J. H.; Dreher, K. L. (1997) Epithelial injury induced by exposure to residual oil
38             fly-ash particles: role of reactive oxygen species? Am. J. Respir. Cell Mol. Biol. 17: 625-633.
39       Dye, J. A.; Adler, K. B.; Richards, J. H.; Dreher, K. L. (1999) Role of soluble metals in oil fly ash-induced airway
40             epithelial injury and cytokine gene expression. Am. J. Physio). 277: L498-L510.
41       Elder, A. C. P.; Gelein, R.; Finkelstein  J. N.; Cox, C.; Oberdorster, G. (2000) Endotoxin priming affects the lung
42             response to ultrafine particles and ozone in young and old rats. In: Inhalation Toxicology: proceedings of the
43             third colloquium on particulate air pollution and human health; June, 1999; Durham, NC. Inhalation
44             Toxicology 12(suppl. 1): 85-98.
45       Fabiani, R.; Pampanella, L.; Minelli, A.; Mezzasoma, I.; Vecchiarelli, A.; Morozzi, G. (1997) Effect of airborne
46             particulate extracts on monocyte oxidative metabolism. J. Environ. Pathol. Toxicol. Oncol. 16:  195-199.
47       Ferin, J.; Oberdorster, G.; Penney, D. P.; Soderholm, S. C.;  Gelein, R.; Piper, H. C. (1990) Increased pulmonary
48             toxicity of ultrafine particles? I.  Particle  clearance, translocation, morphology.  J. Aerosol Sci. 21: 381-384.
49       Ferin, J.; Oberdorster, G.; Penney, D. P. (1992) Pulmonary retention of ultrafine and fine particles in rats. Am. J.
50             Respir. Cell Mol. Biol. 6: 535-542.
51       Fine, J. M.; Gordon, T.; Chen, L. C.; Kinney, P.; Falcone, G.; Beckett, W. S. (1997) Metal fume fever:
52             characterization of clinical and plasma IL-6 responses in controlled human exposures to zinc oxide fume at
53             and below the threshold limit value. J. Occup. Environ. Med. 39: 722-726.
54
         March 2001                                     8-90          DRAFT-DO NOT QUOTE OR CITE

-------
  1       Finkelstein, J. N.; Johnston, C.; Barrett, T.; Oberdorster, G. (1997) Particulate-cell interactions and pulmonary
  2             cytokine expression. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth international meeting
  3             on the toxicology of natural and man-made fibrous and non-fibrous particles; September 1996; Lake Placid,
  4             NY. Environ. Health Perspect. Suppl. 105(5): 1179-1182.
  5       Folinsbee, L. J.; Kim, C. S.; Kehrl, H. R.; Prah, J. D.; Devlin, R. B. (1997) Methods in human inhalation toxicology.
  6             In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press LLC; pp. 607-670.
  7       Fortoul, T. I.; Osorio, L. S.; Tovar, A. T.; Salazar, D.; Castilla, M. E.; Olaiz-Fernandez, G. (1996) Metals in lung
  8             tissue from autopsy cases in Mexico City residents: comparison of cases from the 1950s and the 1980s.
  9             Environ. Health Perspect. 104: 630-632.
 10       Frampton, M. W.; Voter, K. Z.; Morrow, P. E.; Roberts, N. J., Jr.; Culp, D. J.; Cox, C.; Utell, M. J. (1992) Sulfuric
 11             acid aerosol exposure in humans assessed by bronchoalveolar lavage. Am. Rev. Respir. Dis. 146: 626-632.
 12       Frampton, M. W.; Morrow, P. E.; Cox, C.; Levy, P. C.; Condemi, J. J.; Speers, D.; Gibb, F. R.; Utell, M. J. (1995)
 13             Sulfuric acid aerosol followed by ozone exposure in healthy and asthmatic subjects. Environ. Res. 69: 1-14.
 14       Frampton, M. W.; Ohio, A. J.; Samet, J. M.; Carson, J. L.; Carter, J. D.; Devlin, R. B. (1999) Effects of aqueous
 15             extracts of PM10 filters from the Utah valley on human airway epithelial cells. Am. J. Physiol.
 16             277: L960-L967.
 17       Fujimaki, H.; Katayama, N.; Wakamori, K. (1992) Enhanced histamine release from lung mast cells of guinea pigs
 18             exposed to sulfuric acid aerosols. Environ. Res. 58: 117-123.
 19       Gardner, S. Y.;  Lehmann, J. R.; Costa,  D. L. (2000) Oil fly ash-induced elevation of plasma fibrinogen levels in
 20             rats. Toxicol. Sci. 56: 175-180.
 21       Gavett, S. H.; Madison, S.  L.; Dreher, K. L.; Winsett, D. W.; McGee, J. K.; Costa,  D. L. (1997) Metal and sulfate
 22             composition of residual oil fly ash determines airway hyperreactivity and lung injury in rats. Environ. Res.
 23             72: 162-172.
 24       Gavett, S. H.; Madison, S.  L.; Stevens,  M.  A.; Costa, D. L. (1999) Residual oil fly ash amplifies allergic cytokines,
 25             airway responsiveness, and inflammation in mice. Am. J. Respir. Crit. Care Med. 160: 1897-1904.
 26       Gearhart, J. M.; Schlesinger, R. B. (1986) Sulfuric acid-induced airway hyperresponsiveness. Fundam. Appl.
 27             Toxicol. 7: 681-689.
 28       Gerchenetal. (1996)
 29       Ohio, A. J.; Meng, Z. H.; Hatch, G. E.;  Costa, D. L. (1997a) Luminol-enhanced chemiluminescence after in vitro
 30             exposures of rat alveolar macrophages to oil fly ash is metal dependent. Inhalation Toxicol. 9: 255-271.
 31       Ohio, A. J.; Piantadosi, C.  A.;  Crumbliss, A. L. (1997b) Hypothesis: iron chelation plays a vital role in neutrophilic
 32             inflammation. BioMetals 10:  135-142.
 33       Ghio, A. J.; Carter, J. D.; Richards, J. H.; Brighton, L. E.; Lay, J. C.; Devlin, R. B.  (1998a) Disruption of normal
 34             iron homeostasis after bronchial  instillation of an iron-containing particle. Am. J. Physiol. 274: L396-L403.
 35       Ghio, A. J.; Richards, J. H.; Dittrich, K. L.; Samet, J. M. (1998b) Metal storage and transport proteins increase after
 36             exposure of the rat lung to an air pollution particle. Toxicol. Pathol. 26: 388-394.
 37       Ghio, A. J.; Carter, J. D.; Samet, J. M.;  Reed, W.; Quay, J.; Dailey, L. A.; Richards, J. H.; Devlin, R. B. (1998c)
 38             Metal-dependent expression of ferritin and lactoferrin by respiratory epithelial cells. Am. J. Physiol.
 39             274: L728-L736.
40       Ghio, A. J.; Stonehuerner, J.; Dailey, L. A.; Carter, J. D. (1999a) Metals associated with both the water-soluble and
41             insoluble fractions of an ambient air pollution particle catalyze an oxidative stress. Inhalation Toxicol.
42             11:37-49.
43       Ghio, A. J.; Carter, J. D.; Dailey, L. A.; Devlin, R. B.; Samet, J. M. (1999b) Respiratory epithelial cells demonstrate
44             lactoferrin receptors that increase after metal exposure. Am. J. Physiol. 276:  L933-L940.
45       Ghio, A. J.; Stoneheurner, J.; McGee, J. K.; Kinsey, J. S. (1999c) Sulfate content correlates with  iron concentrations
46             in ambient air pollution particles. Inhalation Toxicol. 11: 293-307.
47       Ghioetal. (1999d)
48       Ghio, A. J.; Kim, C.; Devlin, R. B. (2000a) Concentrated ambient air particles induce mild pulmonary inflammation
49             in healthy human volunteers. Am. J. Respir. Crit. Care Med. 162: 981-988.
50      Ghio, A.  J.; Carter, J. D.; Richards, J. H.; Crissman, K. M.; Bobb, H. H.; Yang, F. (2000b)  Diminished injury in
51             hypotransferrinemic mice after exposure to a metal-rich particle. Am. J. Physiol. 278: L1051-LI061.
52      Gift, J. S.; Faust, R. A. (1997) Noncancer inhalation toxicology of crystalline silica: exposure-response assessment.
53            J. Exposure Anal. Environ. Epidemiol. 7: 345-358.
54      Gilmour, M. I.; Taylor, F. G. R.; Baskerville, A.; Wathes, C. M. (1989a) The effect of titanium dioxide inhalation
55            on the pulmonary clearance of Pasteurella haemolytica in the mouse. Environ. Res. 50: 157-172.


         March 2001                                     8-91           DRAFT-DO NOT QUOTE OR CITE

-------
  1       Gilmour, M. I.; Taylor, F. G. R.; Wathes, C. M. (1989b) Pulmonary clearance of Pasteurella haemolytica and
  2             immune responses in mice following exposure to titanium dioxide. Environ. Res. 50: 184-194.
  3       Gilmour, P. S.; Brown, D. M.; Lindsay, T. G.; Beswick, P. H.; MacNee, W.; Donaldson, K. (1996) Adverse health
  4             effects of PM10 particles: involvement of iron in generation of hydroxyl radical. Occup. Environ. Med.
  5             53:817-822.
  6       Godleski, J. J.; Verrier, R. L.; Koutrakis, P.; Catalano, P. (2000) Mechanisms of morbidity and mortality from
  7             exposure to ambient air particles. Cambridge, MA: Health Effects Institute; research report no. 91.
  8       Goldsmith, C.-A.; Frevert, C.; Imrich, A.; Sioutas, C.; Kobzik, L. (1997) Alveolar macrophage interaction with air
  9             pollution particulates. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth international meeting
10             on the toxicology of natural and man-made fibrous and non-fibrous particles; September 1996; Lake Placid,
11             NY. Environ. Health Perspect. Suppl.  105(5): 1191-1195.
12       Goldsmith, C.-A. W.; Imrich, A.; Danaee, H.; Ning, Y. Y.; Kobzik, L. (1998) Analysis of air pollution
13             particulate-mediated oxidant stress in alveolar macrophages. J. Toxicol. Environ. Health Part A 54: 529-545.
14       Goldsmith, C.-A. W.; Hamada, K.; Ning, Y. Y.; Qin, G.; Catalano, P.; Murthy, G. G. K.; Lawrence, J.; Kobzik, L.
15             (1999) The effects of environmental aerosols on airway hyperresponsiveness in a murine model of asthma.
16             Inhalation Toxicol. 11: 981-998.
17       Gordon, T.; Nadziejko, C.; Schlesinger, R.; Chen, L. C. (1998) Pulmonary and cardiovascular effects of acute
18             exposure to concentrated ambient particulate matter in rats. Toxicol. Lett. 96-97: 285-288.
19       Gordon, T.; Nadziejko, C.; Chen, L. C.; Schlesinger, R. (2000) Effects of concentrated ambient particles in rats and
20             hamsters: an exploratory study. Cambridge, MA: Health Effects Institute; research report no. 93.
21       Grabowski, G. M.; Paulauskis, J. D.; Godleski, J. J. (1999) Mediating phosphorylation events in the
22             vanadium-induced respiratory burst of alveolar macrophages. Toxicol. Appl. Pharmacol. 156: 170-178.
23       Gulisano, M.; Marceddu, S.; Barbara, A.; Pacini, A.; Buiatti, E.; Martini, A.; Pacini, P. (1997) Damage to the
24             nasopharyngeal mucosa induced by current levels of urban air pollution: a field study in lambs. Eur. Respir.
25             J.  10:567-572.
26       Hahon, N.; Booth, J. A.; Green, F.; Lewis, T. R.  (1985) Influenza virus infection in mice after exposure to coal dust
27             and diesel engine emissions. Environ.  Res. 37: 44-60.
28       Hamada, K.; Goldsmith, C.-A.; Kobzik, L. (1999) Increased airway hyperresponsiveness and inflammation in a
29             juvenile mouse model of asthma exposed to air-pollutant aerosol. J. Toxicol. Environ. Health 58: 129-143.
30       Hamada, K.; Goldsmith, C.-A.; Goldman, A.; Kobzik, L. (2000) Resistance of very young mice to inhaled allergen
31             sensitization Is overcome by coexposure to an air-pollutant aerosol. Am. J. Respir. Crit. Care Med.
32             161:1285-1293.
33       Hatch, G. E.; Boykin, E.; Graham, J. A.; Lewtas, J.; Pott, F.; Loud, K.; Mumford, J. L. (1985) Inhalable particles
34             and pulmonary host defense: in vivo and in vitro effects of ambient air and combustion particles. Environ.
35             Res. 36: 67-80.
36       Health Effects Institute. (1995) Diesel exhaust: a critical analysis of emissions, exposure, and health effects:
37             a special report of the Institute's Diesel Working Group. Cambridge, MA: Health Effects Institute.
38       Heyder, J.; Beck-Speier, I.; Busch, B.; Dirscherl, P.; Heilmann, P.; Ferron, G. A.; Josten, M.; Karg, E.; Kreyling,
39             W. G.; Lenz, A.-G.; Maier, K. L.; Miaskowski, U.; Platz, S.; Reitmeir, P.; Schulz, H.; Takenaka, S.;
40             Ziesenis, A. (1999) Health effects of sulfur-related environmental air pollution. I. Executive summary.
41             Inhalation Toxicol. 11: 343-359.
42       Hitzfeld, B.; Friedrichs, K. H.; Ring, J.; Behrendt, H. (1997) Airborne particulate matter modulates the production
43             of reactive oxygen species in  human polymorphonuclear granulocytes. Toxicology  120:  185-195.
44       Holian, A.; Hamilton, R. F., Jr.; Morandi, M. T.; Brown, S. D.; Li, L. (1998) Urban particle-induced apoptosis and
45             phenotype shifts in human alveolar macrophages. Environ. Health Perspect. 106:  127-132.
46       Irsigler, G. B.; Visser, P. J.; Spangenberg, P. A. L. (1999) Asthma and chemical bronchitis in vanadium plant
47             workers. Am. J. Ind. Med. 35: 366-374.
48       Jacobs, J.; Kreutzer, R.; Smith, D. (1997) Rice burning and asthma hospitalizations, Butte County, California,
49             1983-1992. Environ. Health Perspect.  105: 980-985.
50       Jakab, G. J. (1992) Relationship between carbon black particulate-bound formaldehyde, pulmonary antibacterial
51             defenses, and alveolar macrophage phagocytosis. Inhalation Toxicol. 4: 325-342.
52       Jakab, G. J. (1993) The toxicologic interactions resulting from inhalation of carbon black and acrolein on
53             pulmonary antibacterial and antiviral defenses. Toxicol. Appl. Pharmacol. 121: 167-175.
         March 2001                                    8-92          DRAFT-DO NOT QUOTE OR CITE

-------
  1       Jakab, G. J.; Hemenway, D. R. (1993) Inhalation coexposure to carbon black and acrolein suppresses alveolar
  2             macrophage phagocytosis and TNF-arelease and modulates peritoneal macrophage phagocytosis. Inhalation
  3             Toxicol. 5: 275-289.
  4       Jakab, G. J.; Clarke, R. W.; Hemenway, D. R.; Longphre, M. V.; Kleeberger, S. R.; Frank, R. (1996) Inhalation of
  5             acid coated carbon black particles impairs alveolar macrophage phagocytosis. Toxicol. Lett. 88: 243-248.
  6       Johnston, C. J.; Finkelstein, J. N.; Gelein, R.; Baggs, R.; Oberdorster, G. (1996) Characterization of the early
  7             pulmonary inflammatory response associated with PTFE fume exposure. Toxicol. Appl. Pharmacol.
  8             140:  154-163.
  9       Johnston, C. J.; Finkelstein, J. N.; Gelein, R.; Oberdorster, G. (1998) Pulmonary inflammatory responses and
 10             cytokine and antioxidant mRNA levels in the lungs of young and old C57BL/6 mice after exposure to Teflon
 11             fumes. Inhalation Toxicol. 10: 931-953.
 12       Kadiiska, M. B.; Mason, R. P.; Dreher, K. L.; Costa, D. L.; Ohio, A. J. (1997) In vivo evidence of free radical
 13             formation in the rat lung after exposure to an  emission source air pollution particle. Chem. Res. Toxicol.
 14             10:1104-1108.
 15       Kennedy, T.; Ghio, A. J.; Reed, W.; Samet, J.; Zagorski, J.; Quay, J.; Carter, J.; Dailey, L.; Hoidal, J. R.; Devlin,
 16             R. B. (1998) Copper-dependent inflammation and nuclear factor-KB activation by particulate air pollution.
 17             Am. J. Respir. Cell  Mol. Biol. 19: 366-378.
 18       Killingsworth, C. R.; Alessandrini, F.; Krishna Murthy, G. G.; Catalano, P. J.; Paulauskis, J. D.; Godleski, J. J.
 19             (1997) Inflammation, chemokine expression,  and death in monocrotaline-treated rats following fuel oil fly
 20             ash inhalation. Inhalation Toxicol. 9: 541-565.
 21       Kimmel, T.  A.; Chen, L. C.; Bosland, M. C.; Nadziejko,  C. (1997)  Influence of acid aerosol droplet size on
 22             structural changes in the rat lung caused by acute exposure to sulfuric acid and ozone. Toxicol. Appl.
 23             Pharmacol. 144: 348-355.
 24       Kitabatake,  M.; Imai, M.; Kasama, K.; Kobayashi, I.; Tomita, Y.; Yoshida, K. (1979) Effects of air pollutants on the
 25             experimental induction of asthma attacks in guinea pigs: sulfuric acid mist and mixture of the  mist and sulfur
 26             dioxide. Mie  Med. J. 29: 29-36.
 27       Kleeberger,  S. R.; Levitt, R. C.; Zhang, L.-Y.; Longphre, M.; Harkema, J.; Jedlicka, A.; Eleff, S. M.;
 28             DiSilvestre, D.; Holroyd, K. J. (1997) Linkage analysis of susceptibility to ozone-induced lung inflammation
 29             in inbred mice. Nat. Genet. 17: 475-478.
 30       Kleinman, M. T.; Leaf, D.  A.; Kelly, E.; Caiozzo, V.; Osann, K.; O'Niell, T. (1998) Urban angina in  the mountains:
 31             effects of carbon monoxide and mild hypoxemia on subjects with chronic stable angina. Arch. Environ.
 32             Health 53: 388-397.
 33       Kleinman, M. T.; Mautz, W. J.; Bjarnason, S. (1999) Adaptive and non-adaptive responses in rats exposed to ozone,
 34             alone and in mixtures, with acidic aerosols. Inhalation Toxicol. 11: 249-264.
 35       Knox, R. B.; Suphioglu, C.; Taylor, P.; Desai, R.; Watson, H. C.; Peng, J. L.; Bursill, L. A. (1997) Major grass
 36             pollen allergen Lol p 1 binds to diesel exhaust particles: implications for asthma and air pollution. Clin.  Exp.
 37             Allergy 27: 246-251.
 38       Kobzik, L. (1995) Lung macrophage uptake of unopsonized environmental particulates: role of scavenger-type
 39             receptors. J. Immunol. 155: 367-376.
 40       Kodavanti, U.; Costa, D. L. (1999) Animal models to study for pollutant effects. In: Holgate, S. T.; Koren, H. S.;
 41             Samet, J. M.;  Maynard, R. L., eds. Air pollution and health. London, United Kingdom: Academic Press;
 42            pp. 165-197.
43       Kodavanti, U. P.; Jaskot, R. H.; Bonner, J.; Badgett,  A.; Dreher, K. L. (1996) Eosmophilic lung inflammation in
44            particulate-induced lung injury: technical consideration in isolating RNA for gene expression studies.
45            Exp. Lung Res. 22: 541-554.
46       Kodavanti, U. P.; Jaskot, R. H.; Su, W. Y.; Costa, D. L.; Ghio, A. J.; Dreher, K. L. (1997a) Genetic variability  in
47            combustion particle-induced chronic lung injury. Am. J. Physiol. 272: L521-L532.
48       Kodavanti, U. P.; Jaskot, R. H.; Costa, D. L.; Dreher, K. L. (1997b) Pulmonary proinflammatory gene induction
49            following acute exposure to residual oil fly ash: roles of particle-associated metals. Inhalation  Toxicol.
50            9:679-701.
51        Kodavanti, U. P.; Hauser, R.; Christiani, D. C.; Meng, Z. H.; McGee, J.; Ledbetter, A.; Richards, J.; Costa, D. L.
52            (1998a) Pulmonary responses to oil fly ash particles in the rat differ by virtue of their specific soluble metals.
53             Toxicol. Sci. 43:  204-212.
54      Kodavanti, U. P.; Costa, D. L.; Bromberg, P. A. (1998b) Rodent models of cardiopulmonary disease: their potential
55             applicability in studies of air pollutant susceptibility. Environ. Health  Perspect. 106(suppl. 1):  111-130.


         March 2001                                      8-93          DRAFT-DO NOT QUOTE OR CITE

-------
  1      Kodavanti, U. P.; Jackson, M. C.; Ledbetter, A. D.; Richards, J. R.; Gardner, S. Y.; Watkinson, W. P.; Campen,
  2            M. J.; Costa, D. L. (1999) Lung injury from intratracheal and inhalation exposures to residual oil fly ash in a
  3            rat model of monocrotaline-induced pulmonary hypertension. J. Toxicol. Environ. Health Part A
  4            57:101-121.
  5      Kodavanti, U. P.; Mebane, R.; Ledbetter, A.; Krantz, T.; McGee, J.; Jackson, M. C.; Walsh, L.; Hilliard, H.; Chen,
  6            B. Y.; Richards, J.; Costa, D. L. (2000a) Variable pulmonary responses from exposure to concentrated
  7            ambient air particles in a rat model of bronchitis. Toxicol. Sci. 54: 441-451.
  8      Kodavanti, U. P.; Schladweiler, M. C.; Ledbetter, A. D.; Watkinson, W. P.; Campen, M. J.; Winsett, D. W.;
  9            Richards, J. R.; Crissman, K. M.; Hatch, G. E.; Costa, D. L. (2000b) The spontaneously hypertensive rat as a
 10            model of human cardiovascular disease: evidence of exacerbated cardiopulmonary injury and oxidative stress
 11            from inhaled emission particulate matter. Toxicol. Appl. Pharmacol. 164: 250-263.
 12      Kodavanti, U. P.; Schladweiler, M. C. J.; Richards, J. R.; Costa, D. L. (2001) Acute lung injury from intratracheal
 13            exposure to  fugitive residual oil fly ash and its constituent metals in normo- and spontaneously hypertensive
 14            rats. Inhalation Toxicol. 13: 37-54.
 15      Kuschner, W. G.; Wong, H.; D'Alessandro, A.; Quinlan, P.; Blanc, P. D. (1997) Human pulmonary responses to
 16            experimental inhalation of high concentration fine and ultrafine magnesium oxide particles. Environ. Health
 17            Perspect.  105:  1234-1237.
 18      Lambert, A. L.; Dong, W.; Winsett, D. W.; Selgrade, M. K.; Gilmour, M. I. (1999) Residual oil fly ash exposure
 19            enhances allergic sensitization to house dust mite. Toxicol. Appl. Pharmacol. 158: 269-277.
 20      Lambert, A. L.; Dong, W.; Selgrade, M. K.; Gilmour, M. I. (2000) Enhanced allergic sensitization by residual oil fly
 21            ash particles is mediated by soluble metal constituents.  Toxicol. Appl. Pharmacol. 165: 84-93.
 22      Last, J. A.; Pinkerton, K. E. (1997) Chronic exposure of rats to ozone and sulfuric acid aerosol: biochemical and
 23            structural responses. Toxicology 116: 133-146.
 24      Last, J. A.; Hyde, D. M.; Chang, D. P. Y. (1984) A mechanism of synergistic lung damage by ozone and a
 25            respirable aerosol. Exp. Lung Res. 7: 223-235.
 26      Lay, J. C.; Bennett, W. D.; Kim, C. S.; Devlin, R. B.; Bromberg, P. A.  (1998) Retention and intracellular
 27            distribution of instilled iron oxide particles in human alveolar macrophages. Am. J. Respir. Cell Mol. Biol.
 28            18:687-695.
 29      Leduc, D.; Fally, S.; De Vuyst, P.; Wollast, R.; Yernault, J.-C. (1995) Acute exposure to realistic acid fog: effects
 30            on respiratory function and airway responsiveness in asthmatics. Environ. Res. 71: 89-98.
 31      Lee, J.-T.; Shin, D.; Chung, Y. (1999) Air pollution and daily mortality in Seoul and Ulsan, Korea. Environ. Health
 32            Perspect. 107:  149-154.
 33      Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1996) Free radical activity and pro-inflammatory effects of
 34            particulate air pollution (PM10)  in vivo and in vitro.  Thorax 51: 1216-1222.
 35      Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1997) In vivo and in vitro proinflammatory effects of
 36            particulate air pollution (PM10). In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the  sixth
 37            international meeting on the toxicology of natural and man-made fibrous and non-fibrous particles;
 38            September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1279-1283.
 39      Li, X. Y.; Brown, D.; Smith, S.; MacNee, W.; Donaldson,  K. (1999) Short-term inflammatory responses following
40            intratracheal instillation of fine  and ultrafine carbon black in rats. Inhalation Toxicol.  11: 709-731.
41      Linn, W. S.; Shamoo, D. A.; Anderson, K.  R.; Peng, R.-C.; Avol, E. L.; Hackney, J. D. (1994) Effects of prolonged,
42            repeated exposure to ozone, sulfuric acid, and their combination in healthy and asthmatic volunteers. Am. J.
43            Respir. Crit. Care Med. 150: 431 -440.
44      Linn, W. S.; Gong, H., Jr.; Shamoo, D. A.; Anderson, K. R.; Avol, E. L. (1997) Chamber exposures of children to
45            mixed ozone, sulfur dioxide, and sulfuric acid. Arch. Environ. Health 52: 179-187.
46      Lipsett, M.; Hurley, S.; Ostro, B. (1997) Air pollution and  emergency room visits for asthma in Santa Clara County,
47            California. Environ. Health Perspect. 105: 216-222.
48      Lison, D.; Lardot, C.; Huaux, F.; Zanetti, G.; Fubini, B. (1997) Influence of particle surface area on the toxicity of
49            insoluble manganese dioxide dusts. Arch. Toxicol. 71: 725-729.
50      Lorz, C.; Lopez, J.  (1997) Incidence of air pollution in the  pulmonary surfactant system of the pigeon (Columbia
51             livid). Anat.  Rec. 249: 206-212.
52      Madden, M. C.; Thomas, M. J.; Ghio,  A. J. (1999) Acetaldehyde (CH3CHO) production in rodent lung after
53            exposure to metal-rich particles. Free Radical Biol. Med. 26: 1569-1577.
         March 2001                                     8-94          DRAFT-DO NOT QUOTE OR CITE

-------
  1       Madl, A. K.; Wilson, D. W.; Segall, H. J.; Pinkerton, K. E. (1998) Alteration in lung particle translocation,
  2            macrophage function, and microfilament arrangement in monocrotaline-treated rats. Toxicol. Appl.
  3            Pharmacol. 153:28-38.
  4       Mauderly, J. L. (1993) Toxicological approaches to complex mixtures. Environ. Health Perspect. Suppl.
  5             101(4): 155-165.
  6       McKenna, I. M; Gordon, T.; Chen, L. C.; Anver, M. R.; Waalkes, M. P. (1998) Expression of metallothionein
  7            protein in the lungs of Wistar rats and C57 and DBA mice exposed to cadmium oxide fumes. Toxicol. Appl.
  8            Pharmacol. 153: 169-178.
  9       Michel, O.;  Nagy, A.-M.; Schroeven, M.; Duchateau, J.; Neve, J.; Fondu, P.; Sergysels, R. (1997) Dose-response
 10            relationship to inhaled endotoxin in normal subjects. Am. J. Respir. Crit. Care Med. 156:  1157-1164.
 11       Minami, M.; Endo, T.; Hamaue, N.; Hirafuji, M.; Mori, Y.; Hayashi, H.; Sagai, M.; Suzuki, A. K. (1999)
 12            Electrocardiographic changes induced by diesel  exhaust particles (DEP) in guinea pigs. Res. Comm. Mol.
 13            Pathol. Pharmacol. 105: 67-76.
 14       Monn, C.; Becker, S. (1999) Cytotoxicity and induction of proinflammatory cytokines from human monocytes
 15            exposed to fine (PM2 5) and coarse particles (PM,0.2 5) in outdoor and indoor  air. Toxicol. Appl. Pharmacol.
 16            155:245-252.
 17       Muggenberg, B. A.; Barr, E. B.; Cheng, Y. S.; Seagrave, J. C.; Tilley, L. P.; Mauderly, J. L. (2000) Effect of
 18            inhaled residual oil fly ash on the electrocardiogram of dogs. In: Grant, L. D., ed. PM2000: particulate matter
 19            and health. Inhalation Toxicol. 12(suppl. 4): 189-208.
 20       Murphy, S. A.; BeruBe, K. A.; Pooley,  F. D.; Richards, R. J. (1998) The response of lung epithelium to well
 21            characterised fine particles. Life Sci. 62: 1789-1799.
 22       Nadadur, S.  S.; Schladweiler, M. C. J.;  Kodavanti, U. P. (2000) A pulmonary rat gene array for screening altered
 23            expression profiles in air pollutant-induced lung injury. Inhalation Toxicol. 12: 1239-1254.
 24       Nadeau, D.;  Vincent, R.; Kumarathasan, P.; Brook, J.; Dufresne, A. (1996) Cytotoxicity of ambient air particles to
 25            rat lung macrophages: comparison  of cellular and functional assays. Toxicol. In Vitro 10:  161-172.
 26       Nakamura and Hayashida (1992)
 27       National Institutes of Health. (1997) Guidelines for the diagnosis and management  of asthma: expert panel report 2.
 28            Bethesda, MD: U.S. Department of Health and Human Services, National Heart, Lung, and Blood Institute;
 29            publication no. 97-4051.
 30       Nel, A. E.; Diaz-Sanchez, D.; Ng, D.; Hiura, T.; Saxon, A. (1998) Enhancement of allergic inflammation by the
 31            interaction between diesel exhaust particles and the immune system. J. Allergy Clin. Immunol. 102: 539-554.
 32       Nemmar, A.; Delaunois, A.; Nemery, B.; Dessy-Doize, C.; Beckers, J.-F.; Sulon, J.; Gustin, P. (1999) Inflammatory
 33            effect of intratracheal instillation of ultrafine particles in the rabbit: role of C-fiber and mast cells.
 34            Toxicol. Appl. Pharmacol. 160: 250-261.
 35       Oberdorster, G.; Ferin, J.; Gelein, R.; Soderholm, S. C.; Finkelstein, J. (1992) Role  of the alveolar macrophage in
 36            lung injury: studies with ultrafine particles. Environ. Health Perspect. 97: 193-199.
 37       Oberdorster  etal. (1995)
 38       Oberdorster, G.; Cox, C.; Gelein, R. (1997) Intratracheal instillation versus intratracheal inhalation of tracer
 39            particles for measuring lung clearance function. Exp. Lung Res. 23: 17-34.
 40       Oberdorster, G.; Finkelstein, J. N.; Johnston, C.; Gelein, R.; Cox, C.; Baggs, R.; Elder, A. C. P. (2000) Acute
 41             pulmonary effects of ultrafine particles in rats  and mice. Cambridge, MA: Health Effects Institute; research
 42            report no. 96.
 43       Oettinger, R.; Drumm, K.; Knorst, M.; Krinyak, P.; Smolarski, R.; Kienast, K. (1999) Production of reactive oxygen
 44            intermediates by human macrophages exposed to soot particles and asbestos  fibers and increase in NF-KB
45             P50/pl05mRNA. Lung 177: 343-354.
46       Ohtoshi, T.;  Takizawa, H.; Okazaki, H.; Kawasaki, S.; Takeuchi, N.; Ohta, K.;  Ito, K. (1998) Diesel exhaust
47            particles stimulate human  airway epithelial cells to produce cytokines relevant to airway inflammation in
48             vitro.  J. Allergy Clin. Immunol. 101: 778-785.
49       Ohtsuka, Y.; Clarke, R. W.; Mitzner, W.; Brunson, K.; Jakab, G. J.; Kleeberger, S. R. (2000a) Interstrain variation
 50             in murine susceptibility to inhaled acid-coated particles. Am. J. Physiol.  L469-L476.
51       Ohtsuka, Y.; Brunson, K. J.; Jedlicka, A. E.; Mitzner, W.; Clarke, R. W.; Zhang, L.-Y.; Eleff, S. M.; Kleeberger,
52             S. R. (2000b) Genetic linkage analysis of susceptibility to particle exposure in mice. Am. J. Respir. Cell Mol.
53             Biol. 22:574-581.
54       Oortgiesen, M.; Veronesi, B.; Eichenbaum, G.; Kiser, P. F.; Simon, S. A. (2000) Residual oil fly ash and charged
55             polymers activate epithelial cells  and nociceptive sensory neurons. Am. J. Physiol. 278: L683-L695.


         March 2001                                     8-95          DRAFT-DO NOT QUOTE OR CITE

-------
  1       Osier, M.; Oberdorster, G. (1997) Intratracheal inhalation vs intratracheal instillation: differences in particle effects.
  2             Fundam. Appl. Toxicol. 40: 220-227.
  3       Osier, M.; Baggs, R. B.; Oberdorster, G. (1997) Intratracheal instillation versus intratracheal inhalation: influence of
  4             cytokines on inflammatory response. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth
  5             international meeting on the toxicology of natural and man-made fibrous and non-fibrous particles;
  6             September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1265-1271.
  7       Osman, K.; Zejda, J. E.; Schiitz, A.; Mielzynska, D.; Elinder, C. G.; Vahter, M. (1998) Exposure to lead and other
  8             metals in children from Katowice district, Poland. Int. Arch. Occup.  Environ. Health 71: 180-186.
  9       Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997) Respiratory effects are associated with the
10             number of ultrafme particles. Am. J. Respir. Crit. Care Med. 155:  1376-1383.
11       Pierce, L. M.; Alessandrini, F.; Godleski, J. J.; Paulauskis, J. D. (1996) Vanadium-induced chemokine mRNA
12             expression and pulmonary inflammation. Toxicol. Appl. Pharmacol.  138: 1-11.
13       Pinto, M.; Birnbaum, S. C.; Kadar, T.; Goldberg, G. M. (1979) Lung injury in mice induced by factors acting
14             synergistically with inhaled particulate antigen. Clin. Immunol. Immunopathol. 13: 361-368.
15       Prahalad, A. K.; Soukup, J. M.; Inmon, J.;  Willis, R.; Ohio, A.  J.; Becker, S.; Gallagher, J. E. (1999) Ambient air
16             particles: effects on cellular oxidant radical generation in relation to  particulate elemental chemistry. Toxicol.
17             Appl. Pharmacol. 158:81-91.
18       Pritchard, R. J.; Ohio, A. J.; Lehmann, J. R.; Winsett, D. W.; Tepper, J. S.;  Park, P.; Gilmour, M. I.; Dreher, K. L.;
19             Costa, D. L. (1996) Oxidant generation and lung injury after particulate air pollutant exposure increase with
20             the concentrations of associated metals. Inhalation Toxicol. 8: 457-477.
21       Prows, D. R.; Shertzer, H. G.; Daly, M. J.; Sidman, C. L.; Leikauf, G. D. (1997) Genetic analysis of ozone-induced
22             acute lung injury in sensitive and resistant strains of mice. Nat. Genet. 17: 471-474.
23       Quay, J. L.; Reed, W.; Samet, J.; Devlin, R. B. (1998) Air pollution particles induce IL-6 gene expression in human
24             airway epithelial cells via NF-KB activation. Am. J.  Respir. Cell Mol. Biol. 19: 98-106.
25       Rose, C. S.; Martyny, J. W.; Newman, L. S.; Milton, D. K.; King, T. E., Jr.; Beebe, J. L.; McCammon, J. B.;
26             Hoffman, R. E.; Kreiss, K. (1998) "Lifeguard lung": endemic granulomatous pneumonitis in an indoor
27             swimming pool. Am. J. Public Health 88: 1795-1800.
28       Rudell, B.; Sandstrom, T.; Stjernberg, N.; Kolmodin-Hedman, B. (1990) Controlled diesel exhaust exposure in an
29             exposure chamber: pulmonary effects investigated with bronchoalveolar lavage. J. Aerosol Sci.
30             21(suppl. 1):S411-S414.
31       Sagai, M.; Furuyama, A.; Ichinose, T. (1996) Biological effects of diesel exhaust particles (DEP). III. Pathogenesis
32             of asthma like symptoms in mice. Free Radical Biol. Med. 21:199-209.
33       Samet, J. M.; Reed, W.; Ohio, A.  J.; Devlin, R. B.; Carter, J. D.; Dailey, L.  A.; Bromberg, P. A.; Madden, M. C.
34             (1996) Induction of prostaglandin H synthase 2 in human airway epithelial cells exposed to residual oil fly
35             ash. Toxicol. Appl. Pharmacol. 141: 159-168.
36       Samet, J. M.; Stonehuerner, J.; Reed, W.; Devlin, R. B.; Dailey, L. A.; Kennedy, T. P.; Bromberg, P. A.; Ohio, A. J.
37             (1997) Disruption of protein tyrosine phosphate homeostasis in bronchial epithelial cells exposed to oil fly
38             ash. Am. J. Physiol. 272: L426-L432.
39       Samet, J. M.; Graves, L. M.; Quay, J.; Dailey, L. A.; Devlin, R. B.; Ohio, A. J.; Wu, W.; Bromberg, P. A.; Reed, W.
40             (1998) Activation of MAPKs in human bronchial epithelial cells exposed to metals. Am. J. Physiol.
41             275: L551-L558.
42       Samet, J. M.; Ghio, A. J.; Costa, D. L.; Madden, M. C. (2000) Increased expression of cyclooxygenase 2 mediates
43             oil fly ash-induced lung injury. Exp. Lung Res. 26: 57-69.
44       Seagrave, J. C.; Nikula, K. J. (2000) Multiple modes of responses to air pollution particulate materials in
45             A549 alveolar type II cells. In: Grant, L. D., ed.  PM2000: particulate matter and health. Inhalation Toxicol.
46             12(suppl. 4): 247-260.
47       Seaton, A.; MacNee, W.; Donaldson, K.; Godden, D. (1995) Particulate air pollution and acute health effects.
48             Lancet (8943): 176-178.
49       Schluter, T.; Berg, I.; Dorger, M.; Gercken, G. (1995) Effect of heavy metal ions on the release of reactive oxygen
50             intermediates by bovine alveolar macrophages. Toxicology 98: 47-55.
51       Shukla et al. (2000)
52       Silbajoris, R.; Ghio, A. J.; Samet, J. M.; Jaskot, R.; Dreher, K.  L.; Brighton, L. E. (2000) In vivo and in vitro
53             correlation of pulmonary map kinase activation following metallic exposure. Inhalation Toxicol. 12: 453-468.
54       Sioutas, C.; Koutrakis, P.; Burton, R. M. (1995) A technique to expose animals to concentrated fine ambient
55             aerosols. Environ. Health Perspect.  103: 172-177.


         March 2001                                     8-96         DRAFT-DO NOT QUOTE OR CITE

-------
  1      Sjogren, B. (1997) Occupational exposure to dust: inflammation and ischaemic heart disease. Occup. Environ. Med.
  2            54:466-469.
  3      Smith, K.  R.; Aust, A. E. (1997) Mobilization of iron from urban particulates leads to generation of reactive oxygen
  4            species in vitro and induction of ferritin synthesis in human lung epithelial cells. Chem. Res. Toxicol.
  5            10: 828-834.
  6      Smith, K.  R.; Veranth, J. M.; Lighty, J. S.; Aust, A. E. (1998) Mobilization of iron from coal fly ash was dependent
  7            upon the particle size and the source of coal. Chem. Res. Toxicol. 11: 1494-1500.
  8      Spengler,  J. D.; Thurston, G. D. (1983) Mass and elemental composition of fine and coarse particles in six U.S.
  9            cities. J. AirPollut. Control Assoc. 33: 1162-1171.
 10      Steerenberg, P. A.; Dormans, J. A. M. A.; Van Doom, C. C. M.; Middendorp, S.; Vos, J. G.; Van Loveren, H.
 11            (1999) A pollen model in the rat for testing adjuvant activity of air pollution components. Inhalation Toxicol.
 12            11:1109-1122.
 13      Stringer, B.; Kobzik, L. (1996) Alveolar macrophage uptake of the environmental particulate titanium dioxide:
 14            role of surfactant components. Am. J. Respir. Cell Mol. Biol. 14: 155-160.
 15      Stringer, B.; Kobzik, L. (1998) Environmental particulate-mediated cytokine production in lung epithelial cells
 16            (A549): role of preexisting inflammation and oxidant stress. J. Toxicol. Environ. Health Part A 55: 31-44.
 17      Stringer, B.; Imrich, A.; Kobzik, L. (1996) Lung epithelial cell (A549) interaction with unopsonized environmental
 18            particulates: quantitation of particle-specific binding and IL-8 production. Exp. Lung Res. 22: 495-508.
 19      Su, W.-Y.; Jaskot, R. H.; Richards, J.; Abramson,  S. R.; Woessner, F. J.; Wu, W.-H.; Dreher, K. L. (2000a)
 20            Induction of pulmonary matrilysin expression by combustion  and ambient air particles. Am. J. Physiol.
 21            279:L152-L160.
 22      Su, W.-Y.; Jaskot, R. H.; Dreher, K. L. (2000b) Particulate matter induction of pulmonary gelatinase A, gelatinase
 23            B, tissue inhibitor of metaloproteinase expression. Inhalation  Toxicol. 12(suppl. 2): 105-119.
 24      Takano, H.; Yoshikawa, T.; Ichinose, T.; Miyabara, Y.; Imaoka, K.;  Sagai, M. (1997) Diesel exhaust particles
 25            enhance antigen-induced airway inflammation and local cytokine expression in mice. Am. J. Respir. Crit.
 26            Care Med. 156:36-42.
 27      Takano, H.; Ichinose, T.; Miyabara,  Y.; Shibuya, T.; Lim, H.-B.; Yoshikawa, T. Sagai, M. (1998)  Inhalation of
 28            diesel exhaust enhances allergen-related eosinophil recruitment and airway hyperresponsiveness  in mice.
 29            Toxicol. Appl. Pharmacol. 150: 328-337.
 30      Terashima, T.; Wiggs, B.; English, D.; Hogg, J. C.; Van Eeden, S. F. (1997) Phagocytosis of small carbon particles
 31            (PMio) by alveolar macrophages stimulates the release of polymorphonuclear leukocytes from bone marrow.
 32            Am. J. Respir. Crit. Care Med. 155: 1441-1447.
 33      Thurston,  G. D.; Lippmann, M.; Scott, M. B.; Fine, J. M. (1997) Summertime haze air pollution and children with
 34            asthma. Am. J. Respir. Crit. Care Med. 155: 654-660.
 35      Timblin, C.; BeruBe, K.; Chrug, A.; Driscoll, K.; Gordon, T.; Hemenway, D.; Walsh, E.; Cummins, A. B.;
 36            Vacek, P.; Mossman, B. (1998) Ambient particulate matter causes activation of the c-jun
 37            kinase/stress-activated protein kinase cascade and DNA synthesis in lung epithelial cells. Cancer Res.
 38            58:4543-4547.
 39      Tsien, A.;  Diaz-Sanchez, D.; Ma, J.; Saxon, A. (1997) The organic component of diesel exhaust particles and
40            phenanthrene, a major polyaromatic hydrocarbon constituent,  enhances IgE production by IgE-secreting
41            EBV-transformed human B cells in vitro. Toxicol. Appl. Pharmacol. 142: 256-263.
42      Tsuchiyama, F.; Hisanaga, N.; Shibata, E.; Aoki, T.;  Takagi, H.; Ando, T.; Takeuchi, Y. (1997) Pulmonary metal
43            distribution in urban dwellers. Int.  Arch. Occup. Environ. Health 70: 77-84.
44      Turpin, B. J. (1999) Options for characterizing organic particulate matter. Environ. Sci. Technol. 33: 76A-79A.
45      U.S. Environmental Protection Agency. (1982) Air quality criteria for particulate matter and sulfur oxides. Research
46            Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment
47            Office; EPA report no. EPA-600/8-82-029aF-cF. 3v. Available from: NTIS, Springfield, VA; PB84-156777.
48      U.S. Environmental Protection Agency. (1989) An acid aerosols issue paper: health effects and aerometrics.
49            Research Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and
50            Assessment Office; EPA report no. EPA-600/8-88-005F. Available from: NTIS, Springfield, VA;
51             PB91-125864.
52      U.S. Environmental Protection Agency. (1996a) Air quality criteria for particulate matter. Research Triangle Park,
53            NC: National Center for Environmental Assessment-RTP Office; report nos. EPA/600/P-95/001aF-cF. 3v.
         March 2001                                     8-97          DRAFT-DO NOT QUOTE OR CITE

-------
  1       U.S. Environmental Protection Agency. (1996b) Ambient levels and noncancer health effects of inhaled crystalline
  2             and amorphous silica: health issue assessment [draft]. Research Triangle Park, NC: National Center for
  3             Environmental Assessment; EPA report no. EPA/600/R-95/115.
  4       U.S. Environmental Protection Agency. (1999) Health assessment document for diesel emissions, SAB review draft.
  5             Washington, DC: Office of Research and Development; report no. EPA/600/8-90/057C.
  6       Utell, M. J.; Morrow, P. E.; Hyde, R. W. (1983) Latent development of airway hyperreactivity in human subjects
  7             after sulfuric acid aerosol exposure. J. Aerosol Sci. 14: 202-205.
  8       Van Maanen, J. M.; Borm, P. J.; Knaapen, A.; Van Herwijnen, M.; Schilderman, P. A.; Smith, K. R.; Aust, A. E.;
  9             Tomatis, M.; Fubini, B. (1999) In vitro effects of coal fly ashes: hydroxyl radical generation, iron release,
10             and DNA damage and toxicity in rat lung epithelial cells. Inhalation Toxicol. 11:1123-1142.
11       Vanda, B.; de Buen, N.; Jasso, R.; Valero, G.; Vargas, M. H.; Olmos, R.; Arreola, J. L.; Santillan, P.; Alonso, P.
12             (1998) Inflammatory cells and ferruginous bodies in bronchoalveolar lavage in urban dogs. Acta Cytol.
13             42:939-944.
14       Veronesi, B.; Oortgiesen, M.; Carter, J. D.; Devlin, R. B. (1999) Particulate matter initiates inflammatory cytokine
15             release by activation of capsaicin and acid receptors in a human bronchial epithelial cell line. Toxicol. Appl.
16             Pharmacol. 154: 106-115.
17       Vincent, R.; Bjarnason, S. G.; Adamson, I. Y. R.; Hedgecock, C.; Kumarathasan, P.; Guenette, J.; Potvin, M.;
18             Goegan, P.; Bouthillier, L. (1997) Acute pulmonary toxicity of urban particulate matter and ozone. Am. J.
19             Pathol. 151: 1563-1570.
20       Vogelzang, P. F. J.; Van Der Gulden, J. W. J.; Folgering, H.; Kolk, J. J.;  Heederik, D.; Preller, L.; Tielen. M. J. M.;
21             Van Schayck, C. P. (1998) Endotoxin exposure as a major determinant of lung function decline in pig
22             farmers. Am. J. Respir. Crit. Care Med. 157: 15-18.
23       Watkinson, W. P.; Campen, M. J.; Costa, D. L. (1998) Cardiac arrhythmia induction after exposure to residual oil
24             fly ash particles in a rodent model of pulmonary hypertension. Toxicol. Sci. 41: 209-216.
25       Watkinson, W. P.; Campen, M. J.; Dreher, K. L.; Su, W.-Y.; Kodavanti,  U.P.; Highfill, J. W.; Costa, D. L. (2000)
26             Thermoregulatory effects following exposure to particulate matter in healthy and
27             cardiopulmonary-compromised rats.  J. Therm. Biol.  25: 131-137.
28       Wesselkamper, S. C.; Prows, D. R.; Biswas, P.; Willeke, K.; Bingham, E.; Leikauf, G. D. (2000) Genetic
29             susceptibility to irritant-induced acute lung injury in mice. Am.  J.  Physiol. 279: L575-L582.
30       Wheatley, L. M.; Platts-Mills, T. A. E. (1996) Perennial allergens and  the asthma epidemic. Sci. Med. 3: 6-13.
31       Woodin, M. A.; Hauser, R.; Liu, Y.; Smith, T. J.; Siegel, P. D.; Lewis, D. M.; Tollerud, D. J.; Christiani, D. C.
32             (1998) Molecular markers of acute upper airway inflammation in workers exposed to fuel-oil ash. Am. J.
33             Respir. Crit. Care Med. 158: 182-187.
34       Yoshinari et  al., 2000
35       Zelikoff, J. T.;  Sisco, M.; Cohen, M. D.; Frampton, M. W.; Utell, M. J; Schlesinger, R. B. (1994) Interspecies
36             comparison of immunotoxicity of inhaled sulfuric acid. II. New Zealand white rabbits. In: 1994 international
37             conference sponsored by the American Lung Association and the American Thoracic Society; May; Boston,
38             MA. Am. J. Respir. Crit. Care Med.  149: A621.
39       Zelikoff, J. T.;  Frampton, M. W.; Cohen, M. D.; Morrow, P. E.; Sisco, M.; Tsai, Y.; Utell, M. J.; Schlesinger, R. B.
40             (1997) Effects of inhaled sulfuric acid aerosols on pulmonary immunocompetence: a comparative study in
41             humans and animals. Inhalation Toxicol. 9: 731-752.
42       Zhang, T.; Huang, C.; Johns, E. J. (1997) Neural regulation of kidney function by the somatosensory system in
43             normotensive and hypertensive rats.  Am. J. Physiol. 273: R1749-R1757.
44       Zock, J.-P.; Hollander, A.; Heederik, D.; Douwes, J. (1998) Acute lung function changes and low endotoxin
45             exposures in the potato processing industry. Am. J. Ind. Med. 33:  384-391.
46
         March 2001                                     8-98         DRAFT-DO NOT QUOTE OR CITE

-------
 i               CHAPTER 9.  INTEGRATIVE  SYNTHESIS:
 2       PARTICULATE  MATTER ATMOSPHERIC SCIENCE,
 3         AIR QUALITY, HUMAN EXPOSURE, DOSIMETRY,
 4                              AND HEALTH RISKS
 5
 6
 7     9.1 INTRODUCTION
 8          This chapter focuses on integration of key information on exposure-dose-response risk
 9     assessment components drawn from the preceding detailed chapters, to provide a coherent
10     framework for assessment of human health risks posed by ambient particulate matter (PM) in the
11     United States.  As such, the chapter updates the integrated assessment provided in the 1996
12     Particulate Matter Air Quality Criteria Document (1996 PM AQCD; U.S. Environmental
13     Protection Agency, 1996) of available scientific information regarding ambient PM sources,
14     exposures, and health risks as they pertain to the United States. This assessment must be
15     considered provisional at this time, pending public comment and Clean Air Scientific Advisory
16     Committee (CASAC) review of other earlier, more detailed chapters from which key findings
17     were extracted for discussion and preliminary integration here.  More complete integration and
18     conclusions will be incorporated in chapter revisions to be made subsequent to the CASAC
19     review.
20          This chapter first provides background information  on key features of atmospheric
21     particles, highlighting important distinctions between fine- and coarse-mode particles with regard
22     to their size, chemical composition, sources, atmospheric behavior, and potential human
23     exposure relationships—distinctions that collectively continue to suggest that fine- and coarse-
24     mode particles should be treated as two distinct subclasses of air pollutants. Information on
25     recent trends in U.S. concentrations of different ambient PM size and composition fractions (e.g.,
26     PM,0, PM2 5, and PM10_2 5) and ranges of variability seen in U.S. regions and urban air sheds also
27     is summarized to place the ensuing health effects discussions in perspective.
28         The chapter next summarizes key points regarding respiratory tract dosimetry, followed by
29     discussion of the extensive PM epidemiologic database that has expanded greatly during recent
30     years.  The latter includes numerous new studies of populations throughout the world published

       March 2001                              9-1         DRAFT-DO NOT QUOTE OR CITE

-------
  1      since the 1996 PM AQCD that contain further evidence that serious health effects (mortality,
  2      exacerbation of chronic disease, increased hospital admissions, etc.) are associated with
  3      exposures to ambient levels of PM found in contemporary U.S. urban air sheds.  Evaluations of
  4      other possible explanations for the reported PM epidemiology results (e.g., effects of weather,
  5      other co-pollutants, choice of models, etc.) also are discussed, ultimately leading to the
  6      conclusion that the reported associations of PM exposure and effects are valid.  The newer
  7      evidence is then discussed that (a) further substantiates associations of such serious health effects
  8      with U.S. ambient PM10 levels, (b) also more strongly establishes fine particles (as indexed by
  9      various indicators, e.g., PM2 5) as likely being important contributors to the observed human
10      health effects, and (c) now provides additional information on associations between coarse-
11      fraction (PM10_2 5) particles and adverse health impacts. The overall coherence of the newer
12      epidemiologic database also is discussed, which strengthens the 1996 PM AQCD evaluation
13      suggesting a likely causal role of ambient PM in contributing to the reported effects.
14           The nature of the observed effects and the biological mechanisms that might underlie such
15      effects then are discussed. The discussion of potential mechanisms of injury examines ways in
16      which PM could induce health effects. The increased, but still limited, availability of new
17      experimental evidence necessary to evaluate or directly substantiate the viability of hypothesized
18      mechanisms is noted. Information concerning possible contributions of particular classes of
19      specific ambient PM constit aents also is summarized.
20           The chapter also provides information on the identification of population groups at special
21      risk for ambient PM effects and factors placing them at increased risk, which need to be
22      considered in generating risk estimates for the possible occurrence  of PM-related health events in
23      the United States.
24
25
26      9.2 ATMOSPHERIC SCIENCE CONSIDERATIONS
27           As discussed in Chapter 2 of this document, airborne PM is not a single pollutant but many
28      classes of pollutants; each class consists of several to many individual chemical species. One
29      classification is based on the natural division of the atmospheric aerosol into fine- and coarse-
30      mode particles. Fine-mode particles, in general, are smaller than coarse-mode particles; they also
31      differ in many other aspects such as formation mechanisms,  chemical composition, sources,
        March 2001                                9-2         DRAFT-DO NOT QUOTE OR CITE

-------
  1     physical behavior, human exposure relationships, and control approaches required for risk
  2     reduction.  Such differences alone are sufficient to justify consideration of fine- and coarse-mode
  3     particles as separate pollutants, regardless of the extent or lack of evidence regarding differences
  4     in respiratory tract dosimetry or associated health effects in laboratory animals or humans. The
  5     various physical and chemical differences between fine- and coarse-mode particles, their sources,
  6     ambient concentrations, factors affecting human exposure, and their respiratory tract deposition
  7     are summarized concisely below as a prelude to discussion of key health effects associated with
  8     ambient PM exposures and other information useful in assessing PM-related public health risks
  9     in the United States.
 10          Atmospheric particles originate from a variety of sources and possess a range of
 11     morphological, chemical, physical, and thermodynamic properties.  The composition and
 12     behavior of airborne particles are linked with those of surrounding gases. Aerosol may be
 13     defined as a suspension of solid or liquid particles in air and includes both the particles and all
 14     vapor or gas phase components of air. However, the term aerosol often is used, as is PM, to refer
 15     to the suspended particles only.  A complete description of the atmospheric aerosol would
 16     include an accounting of the size, morphology, and chemical composition of each particle and the
 17     relative abundance of each particle type as a function of particle size.
 18
 19     9.2.1  Ambient Particulate Matter Size Distinctions
 20          Atmospheric particles differ in density and are not always spherical. Therefore, their
 21      diameters often are described by an "equivalent" diameter. The aerodynamic equivalent diameter
 22     (AED), defined as the diameter of a spherical particle with a density of 1  g/cm3 that would have a
 23      settling velocity equal to the particle in question, is important for particle transport, collection,
 24      and respiratory tract deposition.
 25           The distribution of particles with respect to size is an important physical parameter
 26      governing their behavior. Because atmospheric particles cover several orders of magnitude in
 27      particle size, size distributions often are expressed in terms of the logarithm of the particle
 28      diameter (D), on the X-axis, and the differential concentration on the Y-axis. If the differential
29      concentration is plotted on a linear scale, the surface, volume, mass, or number of particles
30      between D and D + AD is proportional to the area under the curve.  Atmospheric aerosol size
31      distributions frequently are approximated by a sum of log-normal distributions.
        March 2001                                9-3         DRAFT-DO NOT QUOTE OR CITE

-------
  1           The aerosol community uses various approaches or conventions in the classification of
  2      particles by size, including modes, based on the observed size distributions in the atmosphere and
  3      formation mechanisms and cut point, usually based on the 50% cut point of the specific sampling
  4      device i.e., the particle size at which 50% of the particles enter and 50% of the particles are
  5      excluded, as summarized below.
  6           Atmospheric size distributions show that most atmospheric particles are quite small, below
  7      0.1 jum, whereas most of the particle volume (and therefore most of the mass) and much of the
  8      surface area is found in particles greater than 0.1 //m. The surface area peaks around ca. 0.2 ,um.
  9      An important  feature of the mass or volume size distributions of atmospheric aerosols is their
10      multimodal nature.  Volume-size distributions, measured in ambient air in the United States,
11      almost always are found to be bimodal, with a minimum between 1.0 and 3.0 //m (see
12      Figure 9-1). The distribution of particles that are mostly larger than the minimum is termed the
13      coarse mode, whereas the distribution of particles that are mostly smaller than the minimum is
14      termed the fine mode.  In the ambient atmosphere, fine-mode particles include both the nuclei
15      mode and the  accumulation mode.  The nuclei mode, that portion of the fine-particle fraction
16      with diameters below about 0.1  /^m, can be observed as a separate  mode in mass or volume
17      distributions only in clean or remote areas or near sources of new particle formation by
18      nucleation.  Accumulation-mode particles are that portion of the fine-particle fraction with
19      diameters above about 0.1 /zm.  Toxicologists use the term "ultrafme" to refer to particles in the
20      nuclei-mode size range. Aerosol physicists and material scientists  tend to use "nanoparticles" to
21      refer to particles in this size range generated in the laboratory.
22           Another set of definitions  of particle size fractions arises from considerations related to
23      size-selective  sampling (see Figure  9-2).  Size-selective sampling refers to the collection of
24      particles below or within a specified aerodynamic size range, usually defined by the upper 50%
25      cut point size, and has arisen in  an effort to measure particle size fractions  with some special
26      significance (e.g., health, visibility,  source apportionment). Dichotomous samplers split the
27      particles into smaller and larger fractions, which may be collected on separate filters. However, a
28      fraction (-10%) of the fine particles are collected with the coarse particle fraction. Cascade
29      impactors use  multiple size cuts to obtain a distribution of size cuts for mass or chemical
30      composition measurements.


        March 2001                               9-4        DRAFT-DO NOT  QUOTE  OR CITE

-------
         o
            6  -
            5
            4  -
         O)
         I 3

         I
            2  -
            1  -
              0.002
                                                                 Mechanically
                                                                  Generated
                      Nuclei Mode
 0.1              1
Particle Diameter, Dp, |jm
Accumulation Mode
         Coarse Mode
                             Fine-Mode Particles
                       Coarse-Mode Particles
      Figure 9-1.  Volume-size distribution, measured in traffic, showing fine- 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 ag (geometric standard deviation) are shown for each mode.
                  Also shown are transformation and growth mechanisms (e.g., nucleation,
                  condensation, coagulation).

      Source: Adapted from Wilson and Suh (1997).
1          Prior to 1987, the indicator for the National Ambient Air Quality Standards (NAAQS) for

2     PM was total suspended particulate matter (TSP). TSP is defined by the design of the High
3     Volume Sampler (hivol), which collects all of the fine particles but only part of the coarse

4     particles. The upper cut off size of the hivol depends on the wind speed and direction and may

5     vary from 25 to 40 ^m.  In 1987, the NAAQS for PM were revised to use PM10, rather than TSP,

6     as the indicator for the PM NAAQS (Federal Register, 1987). The use of PM10 as an indicator is

7     an example of size-selective sampling.  The selection of PM10 as an indicator was based on
      March 2001
        9-5
DRAFT-DO NOT QUOTE OR CITE

-------
        n
70

60

50
         J§  40
        o>
        .o
            30   -
        CO
        CO
        <   20
            10   -
                       Fine-Mode Particles
                                    Coarse-Mode Particles
                                                         TSP
                                                         HiVol
                                                         WRAC
               0.1
 i
0.2
                                 I
                                    I
I
                     0.5    1.0     2         5      10
                       Aerodynamic Particle Diameter Da,
I
20
 I	
50     100
                                 Total Suspended Particles (TSP)
                                       PM
                                          10
                                PM
                                  '25
                               PM
                                                         (10-2.5)
      Figure 9-2.  An idealized distribution of ambient particulate matter showing fine- and
                  coarse-mode particles and the fractions collected by size-selective samplers.
                  WRAC is the Wide Range Aerosol Classifier, which collects the entire coarse
                  mode (Lundgren and Burton, 1995).
      Source: Adapted from Wilson and Suh (1997).
1     dosimetric considerations and was intended to focus regulatory concern on those particles small
2     enough to enter the thoracic region of the human respiratory tract. The PM2 5 indicator
3     promulgated by U. S. Environmental Protection Agency (EPA) in 1997 is also an example of
4     size-selective sampling.
5          An idealized distribution showing the typically observed division of ambient aerosols into
6     fine- and coarse-mode particles and size fractions collected by the WRAC, TSP, PM10, PM2 5, and
      March 2001
                        9-6
                                              DRAFT-DO NOT QUOTE OR CITE

-------
  1     PM(10.2 5) samplers, is shown in Figure 9-2. PM10 samplers, as defined in Appendix J to 40 Code
  2     of Federal Regulations (CFR) Part 50 (Code of Federal Regulations, 199 la; Federal Register,
  3     1987), collect all of the fine particles and part of the coarse particles. The upper cut point is
  4     defined as having a 50% collection efficiency at 10 ± 0.5 jum AED. The slope of the collection
  5     efficiency curve is defined in amendments to 40 CFR, Part 53, (Code of Federal Regulations,
  6     1991b).
  7          Over the years, the terms "fine" and "coarse", as applied to particle sizes, have lost the
  8     original precise meaning of fine and coarse mode.  In any given article, therefore, the meaning of
  9     fine and coarse, unless defined, must be inferred from the author's usage. In particular, PM2 5
 10     and fine-mode particles are not equivalent. In this chapter and document, the term "mode" is
 11     used with fine and coarse when it is desired to specify the distribution of fine- or coarse-mode
 12     particles as shown in Figures 9-1 and 9-2.
 13
 14     9.2.2 Fine- and Coarse-Mode Particle Distinctions vis-a-vis Sources,
 15           Formation Mechanisms, and Atmospheric Behavior
 16          Table 9-1 summarizes important physical and chemical properties, sources, and
 17     atmospheric behavior that distinguish between nuclei-mode (ultrafine) and accumulation-mode
 18     components of fine particles, as well as coarse-mode particles.
 19          Several processes influence the formation and growth of particles. New particles may be
 20     formed by nucleation from gas-phase material.  Particles may grow by condensation as gas-phase
 21      material  condenses onto existing particles. Particles also may grow by coagulation as two
 22     particles combine to form one.  Gas-phase material condenses preferentially on smaller particles
 23      and the rate constant for coagulation of two particles decreases as the particle size increases.
 24      Therefore, nuclei-mode particles grow into the accumulation mode, but accumulation mode
 25      particles  do not grow into the coarse mode under normal atmospheric conditions.
 26           As  discussed in Chapter 2 of this document, the major constituents of atmospheric PM are
 27      sulfate, nitrate, ammonium, and hydrogen ions; particle-bound water; elemental carbon; a great
 28      variety of organic compounds; and crustal material.  Atmospheric PM contains a large  number of
29      elements in various compounds and concentrations and hundreds to thousands  of specific organic
30      compounds. Particulate matter can be primary or secondary.  Particulate matter is called primary
31      if it is in the same chemical form in which it was emitted into the atmosphere.  Particulate matter

        March 2001                               9-7        DRAFT-DO NOT QUOTE OR CITE

-------
                 TABLE 9-1.  COMPARISON OF AMBIENT PARTICLES,
            FINE (Nuclei Mode Plus Accumulation Mode) AND COARSE MODE
                                      Fine
                       Nuclei
                          Accumulation
                                         Coarse
  Formed from:


  Formed by:
 Composed of:
 Solubility:
 Sources:
 Atmospheric
 half-life:

 Removal
 processes:
         Combustion, high-temperature
      processes, and atmospheric reactions
Nucleation
Condensation
Coagulation
Sulfates
Elemental carbon
Metal compounds
Organic compounds
with very low,
saturation vapor
pressure at ambient
temperature
Probably less
soluble than
accumulation mode

Combustion
Atmospheric
transformation of
SO2 and some
organic compounds
High temperature
processes
Condensation
Coagulation
Evaporation of fog and
cloud droplets in which
gases have dissolved and
reacted

Sulfate, SC-4
Nitrate, NO,
Ammonium, NHJ
Hydrogen ion, H+
Elemental carbon
Large variety of organic
compounds
Metals: compounds of Pb,
Cd, V, Ni, Cu, Zn, Mn, Fe,
etc.
Particle-bound water

Largely soluble,
hygroscopic, and
deliquescent

Combustion of coal, oil,
gasoline, diesel fuel, and
wood
Atmospheric transformation
products of NOX, SO2, and
organic compounds,
including biogenic organic
species (e.g., terpenes)
High-temperature
processes, smelters, steel
mills, etc.
Minutes to hours      Days to weeks
Grows into
accumulation mode
 Travel distance:   <1 to 10s of km
Forms cloud droplets and
rains out
Dry deposition

100s to 1000s of km
                           Break-up of large solids/droplets
Mechanical disruption (crushing,
grinding, and abrasion of surfaces)
Evaporation of sprays
Suspension of dusts
Reactions of gases in or on particles


Suspended soil or street dust
Fly ash from uncontrolled combustion
of coal, oil, and wood
Nitrates and chlorides from HNO3 and
HC1
Oxides of crustal elements
(Si, Al, Ti, and Fe)
CaCO3, NaCl, and sea salt
Pollen, mold, and fungal spores
Plant and animal fragments
Tire, brake pad, and road wear debris

Largely insoluble and nonhygroscopic
Resuspension of industrial dust and
soil tracked onto roads and streets
Suspension from disturbed soil (e.g.,
farming, mining, unpaved roads)
Construction and demolition
Uncontrolled coal and oil combustion
Ocean spray
Biological sources
Minutes to hours


Dry deposition by fallout
Scavenging by falling rain drops


<1 to 10s of km
(100s to 1000s in dust storms)
 Source: Adapted from Wilson and Suh (1997).
March 2001
                              9-8
                       DRAFT-DO NOT QUOTE OR CITE

-------
  1      is called secondary if it is formed by chemical reactions in the atmosphere.  Primary coarse
  2      particles usually are formed by mechanical processes. Primary fine particles are emitted from
  3      sources either directly as particles or as vapors that rapidly condense to form particles.
  4           Most of the sulfate and nitrate and a portion of the organic compounds in atmospheric
  5      particles are secondary (i.e., they are formed by chemical reactions in the atmosphere).
  6      Secondary aerosol formation depends on numerous factors, including the concentrations of
  7      precursors; the concentrations of other gaseous reactive species such as ozone (O3), hydroxyl
  8      radical, peroxy radicals, or hydrogen peroxide; atmospheric conditions, including solar radiation
  9      and relative humidity; and the interactions of precursors and preexisting particles within cloud or
 10      fog droplets or on or in the liquid film on solid particles. As a result of such transformations, it is
 11      considerably more difficult to relate ambient concentrations of secondary species to individual
 12      sources of precursor emissions than it is to identify the sources of primary particles.
 13           The atmospheric lifetimes of particles vary with the aerodynamic diameter of the particle.
 14      Coarse particles can settle rapidly from the atmosphere within minutes or hours and normally
 15      travel only short distances.  However, when mixed high into the atmosphere, as in dust storms,
 16      the smaller-sized, coarse-mode particles may have longer atmospheric residence times and travel
 17      greater distances. Nuclei-mode particles rapidly grow into accumulation-mode fine particles,
 18      which are kept suspended by normal air motions and have very low deposition rates to surfaces.
 19      They can be transported thousands of kilometers and remain in the atmosphere for a number of
 20      days. Particulate matter can be removed  from the atmosphere by wet and dry deposition. Dry
 21      deposition rates are expressed in terms of a deposition velocity, which varies with particle size,
 22      reaching a minimum between 0.1 and 1.0 yum AED.  For small particles, dry deposition is
 23      accomplished by impaction on surfaces by turbulent motion.  For larger particles (i.e., coarse
 24      mode), buoyancy forces are not large enough to overcome the force of gravity, and gravitational
 25      settling becomes important. Soluble particles are removed from the atmosphere primarily by
26      incorporation into cloud droplets, which then rain out. Coarse-mode and ultrafine, but not
27      accumulation-mode, particles are removed by impaction with falling rain drops.
28
       March 2001                               9-9         DRAFT-DO NOT QUOTE OR CITE

-------
 1      9.2.3 Particle Size-Related Distinctions vis-a-vis Number, Surface Area,
 2            and Mass
 3           The distribution of particles in terms of numbers, surface area, and mass in relation to size
 4      is gaming more attention as efforts focus on trying to identify specific toxic components of the
 5      ambient PM mix that may contribute to observed human health and environmental effects.
 6      Examples of averaged atmospheric size distributions are shown in Figures 9-3 and 9-4.
 7      Figure 9-3 describes the number of particles as a function of particle diameter for rural,
 8      urban-influenced rural, urban, and freeway-influenced urban aerosols. For the urban data, the
 9      particle volume distribution is shown in Figure 9-4. The particle diameter is always shown on a
10      logarithmic scale. The particle number frequently is shown on a logarithmic scale to display the
11      wide range in number concentration for different particle sizes  and different sites. When shown
12      on an arithmetic scale the volume, surface area, or number of particles in any specified size range
13      is proportional to the corresponding area under the curve (see Figure 9-5). These distributions
14      show that most of the particles are very small (<0.1 /um), whereas most of the particle volume
15      (and therefore most of the mass) and  the surface area is found in particles >0.1 /um. Also,
16      particle surface area peaks around 0.2 /um.
17           The number concentrations of coarse particles are usually too small to be seen in arithmetic
18      plots (Figure 9-3b) but can be seen in a logarithmic plot (Figure 9-3a). Whitby and Sverdrup
19      (1980) observed that rural aerosols, not much influenced by nearby sources, have a small
20      accumulation mode and no observable nuclei mode. For urban aerosols, the accumulation- and
21      coarse-particle modes are comparable in volume.  However, in urban aerosols, the nuclei-mode
22      can be observed only in volume distributions that are influenced by nearby traffic or other
23      sources of nuclei-mode particles. Still, the nuclei-mode dominates the number distributions of
24      urban aerosols.  Whitby's conclusions were based on extensive studies of size distributions in a
25      number of western and midwestern locations during the 1970s (Whitby, 1978; Whitby and
26      Sverdrup, 1980). No size-distribution studies of similar scope have been published since then.
27      Newer data from particle counting techniques and size-segregation impactor studies, including
28      data from Europe (U.S. Environmental Protection Agency, 1996) and Australia (Keywood et al.,
29      1999), show similar results.
30
       March 2001                               9-10        DRAFT-DO NOT QUOTE OR CITE

-------
             1,000,000-

               10,000 -

          
-------
   70
   65 -
   60 -
   55
   50 -
»  45 -
^40 -
 "S
 930 -
 O)
 % 25 H
 i 20
   15 -
   10 -
    5 -
    0
                                       Clean Rural
                                       Urban Influenced
                                       Rural
                                       South-Central
                                       New Mexico
   70
   65 -
   60 -
   55 -
   50 -
„  45 -
J<0 -
 ^35 H
 "S
 9 30 -
 2 25 -
 £ 20 -
   15 -
   10 -
    5 -
    0
                                                                      Average Urban
                                                                      Urban + Freeway
              0.01      0.1        1        10
                       Particle Diameter, Dp (u.m)
                                      100     0.01      0.1       1       10
                                                       Particle Diameter, Dp (urn)
                                                                                  100
       Figure 9-4.  Particle volume distribution as a function of particle diameter: (a) for the
                   averaged rural and urban-influenced rural number distributions shown in
                   Figure 9-3 and a distribution from south central New Mexico, and (b) for the
                   averaged urban and freeway-influenced urban number distributions shown in
                   Figure 9-3.
       Source: Whitby and Sverdrup (1980) and Kim et al. (1993).
1
2
3
4
5
6
7
(1) Particles containing heavy metals. Nuclei-mode particles of metal oxides or other
metal compounds are generated during metal smelting processes or, more widely, when
metallic impurities in coal or oil are vaporized during combustion and the vapor undergoes
nucleation. Metallic ultrafine particles also may be formed from metals in lubricating oil or
fuel additives that are vaporized during combustion of gasoline or diesel fuels.

(2) Elemental carbon (EC) or soot. Elemental carbon particles are formed primarily by
condensation of C2 molecules generated during combustion processes. Because EC has a
very low equilibrium vapor pressure, ultrafine EC particles can nucleate even at high
      March 2001
                                     9-12
                                                    DRAFT-DO NOT QUOTE OR CITE

-------
    CD
    .Q
    E
     CO
     0)
0)
o
1
co
         x
        IE
         o
         D)
         O
         o
 0.
Q
CD
O
        CO
              15-
              10 -
           5-
             600-
             400-
             200-
              30 -1
       E   ^ 20 H
       5   Q
           D)
           JO
              10H
                                    Nn = 7.7x 10
                                 DGNn = 0.013
                                                                     (a)
                                Na = 1.3x10
                             DGNa = 0.069
                                  = 2.03
   Nc = 4.2
DGNC = 0.97
  agc=2.15
                      Vn = 0.33
                   DGVn = 0.031
0.001       0.01
                                        0.1
                                                1.0
                                                        10
                   100
Figure 9-5. Distribution of coarse (c), accumulation (a), and nuclei- or ultrafine (n) -mode
           particles by three characteristics, (1) number (N), (2) surface area (S), and
           (3) volume (V) for the grand average continental size distribution. DGV =
           geometric mean diameter by volume; DCS = geometric mean diameter by
           surface area; DGN = geometric mean diameter by number; Dp = geometric
           diameter.
Source: Whitby (1978).

March 2001
                                   9-13
                                          DRAFT-DO NOT QUOTE OR CITE

-------
  1           temperatures (Kittelson, 1998; Morawska et al., 1998).  Thus, substantial amounts of EC
  2           can be released into the air as the result of biomass burning (because of agricultural
  3           clearing, forest fires, etc.) or combustion of fossil fuels (e.g., gasoline or diesel fuel derived
  4           from oil or coal used for power generation, industrial boilers, etc.).
  5
  6           (3) Sulfates and nitrates. Sulfuric acid (H2SO4), or its neutralization products with
  7           ammonia (NH3) (i.e., ammonium sulfate [(NH4)2SO4] or ammonium acid sulfate
  8           [NH4HSO4]), are generated in the atmosphere by conversion of sulfur dioxide (SO2) to
  9           H2SO4. As H2SO4 is formed, it can either nucleate to form new ultrafine particles, or it can
10           condense onto existing nuclei-mode or accumulation-mode particles (Clark and Whitby,
11           1975; Whitby, 1978). However, the possible formation of ultrafine ammonium nitrate
12           (NH4NO3) by reaction of NH3 and nitric acid (HNO3) apparently has not been investigated.
13
14           (4) Organic carbon (OC). Recent smog chamber studies and indoor experiments show
15           that atmospheric oxidation of certain organic compounds found in the atmosphere can
16           produce highly oxidized organic compounds with an equilibrium vapor pressure low
17           enough to result in nucleation (Kamens et al., 1999; Weschler and Shields, 1999). Organic
18           carbon compounds originate from a wide variety of processes, including biomass  burning,
19           fossil fuel combustion, use of various dry cleaning or industrial solvents, and release of
20           naturally occurring substances (e.g., terpenes) from certain terrestrial plant species.
21
22           Ambient concentrations of nuclei-mode particles importantly reflect a balance between
23      formation and removal processes.  Nuclei-mode particles are removed mainly by growth into the
24      accumulation mode but also may be removed by dry deposition. Such growth takes place as
25      other low-vapor-pressure material  condenses onto the particles or as nuclei-mode particles
26      coagulate with themselves or with accumulation-mode particles.  Because the rate of coagulation
27      will vary with the concentration of accumulation-mode particles, it might be expected that
28      atmospheric concentrations of nuclei-mode particles would increase with decreases in
29      accumulation-mode mass.  On the  other hand, the concentration of particles would be expected to
30      decrease with a decrease in the rate of generation of particles by reduction in emissions  of metal
31      and carbon particles or a decrease in the rate of generation of H2SO4 or condensable organic

        March 2001                               9-14        DRAFT-DO NOT QUOTE OR CITE

-------
 1     vapor. The rate of generation of H2SO4 depends on the concentration of SO2 and OH radicals.
 2     OH is generated primarily by the photolysis of ozone at wavelengths <320 nm, followed by the
 3     reaction of electronically excited oxygen atoms with water vapor.
 4           Exposure to ultrafine particles may occur near sources of primary ultrafme particles (e.g., in
 5     traffic). Secondary ultrafme particles are generated by photochemistry throughout the boundary
 6     layer; so exposure to ultrafine particles is not limited to locations near primary sources. Models
 7     exist to predict formation and coagulation rates, but no careful analyses of how rapidly various
 8     ultrafine (nuclei-mode) particles may agglomerate or adhere to larger particles as they "age" in
 9     the ambient air or how this may impact lung deposition of such particles has been published.
10     Thus, it may be important to monitor particle number and surface area, as well as mass, to further
11     delineate PM exposure-response relationships and to determine the relative effectiveness of
12     strategies for reducing particle mass, surface area, and number to ameliorate PM-related health
13     risks.
14
15
16     9.3  CHARACTERIZATION OF U.S. AMBIENT PARTICULATE MATTER
17          CONCENTRATIONS AND CONTRIBUTING SOURCES AND
18          EMISSIONS
19     9.3.1  Ambient Particulate Matter Measurement Methods
20           The EPA decision to revise the PM standards by adding daily and annual PM2 5 NAAQS
21     has led to renewed interest in the measurement of atmospheric particles and better understanding
22     of problems in obtaining precise and accurate airborne particle measurements.
23           The U.S. Federal Reference Methods (FRJM) for PM2 5 and PM10 provide relatively precise
24     (±10%) methods for determining the mass of material remaining on a Teflon filter after
25     equilibration at 25 °C and 40% relative humidity. However, many uncertainties exist as to
26     relationships between the mass and composition of material remaining on the filter, as measured
27     by the FRM, and the mass and composition of material that exists in the atmosphere as
28     suspended PM. It is currently not possible to characterize accurately the material that exists as a
29     particle in the atmosphere, in part because of difficulties in creating a reference standard for
30     particles suspended in the atmosphere. As a result, EPA defines accuracy for PM measurements
31     in terms of agreement of a candidate sampler with a reference sampler under standardized

       March 2001                              9-15        DRAFT-DO NOT QUOTE OR CITE

-------
 1     conditions for sample collection, storage, and analysis.  Therefore, intercomparisons of samplers
 2     become very important in determining how well various samplers agree and how various design
 3     choices influence what is actually measured.  Data from ambient PM monitoring is needed to
 4     guide implementation of a standard; to determine whether or not a standard has been attained;
 5     and to determine effects on health, ecosystems, visibility, and the transfer of solar ultraviolet and
 6     visible radiation.
 7          Current filtration-based mass measurements lead to significant evaporative losses, during
 8     and possibly after collection, of a variety of semivolatile components (i.e., species that exist in
 9     the atmosphere in dynamic equilibrium between the condensed phase and gas phase). Important
10     examples include ammonium nitrate,  semivolatile organic compounds, and particle-bound water.
11     In designing an aerosol indicator, choices must be made regarding the treatment of the
12     semivolatile components.  Other areas where choices must be made include selection of an upper
13     cut point; separation of fine- and coarse-mode PM; and treatment of pressure, temperature, and
14     relative humidity.
15          It is becoming increasingly apparent that the semivolatile component of PM impacts
16     significantly the quality of the measurement, and leads to both positive and negative sampling
17     artifacts.  Negative artifacts, because of the loss of ammonium nitrate and semivolatile organic
18     compounds, occur during sampling, because of temperature, relative humidity, composition of
19     the aerosol,  or because of pressure drop across the filter. Negative artifacts also occur during
20     handling and storage because of evaporation. Positive artifacts occur when volatile species
21     adsorb onto, or react with, filter media or collected PM.
22          The loss of particulate nitrate may be determined by comparing nitrate collected on a
23     Teflon filter to that collected on a nylon filter (which absorbs nitrate), preceded by a denuder to
24     remove nitric acid.  In two studies, the PM25 mass lost because of volatilization of ammonium
25     nitrate was found to represent 10 to 20% of the total PM2 5 mass and almost a third of the nitrate.
26     Semivolatile organic compounds (SVOC) can similarly be lost from Teflon filters because of
27     volatilization during or after collection. Such losses can cause the PM2 5 mass  to be
28     underestimated significantly. The FRM for PM2 5 will suffer loss of particulate nitrates and
29     SVOC, similar to the losses experienced with other single-filter collection systems.
30          It is generally desirable to collect and measure ammonium nitrate and SVOC. However, it
31     is also desirable to remove the particle-bound water before determining the mass. Calculations

       March 2001                                9-16        DRAFT-DO NOT QUOTE OR CITE

-------
  1      and measurements indicate that aerosol water content is strongly dependent on composition, but
  2      that liquid water could represent a significant mass fraction of aerosol concentration at relative
  3      humidities above 60%.
  4           Federal Reference Methods for equilibrated mass have been specified for PM10 and PM2 5.
  5      In addition to FRM sampling to determine compliance with PM standards, EPA requires states to
  6      conduct speciation sampling to determine contributions from different source categories and to
  7      evaluate exposure to trace elements. The current speciation samplers include three filters:
  8      (1) a  Teflon filter for equilibrated mass and elemental analysis; (2) a nylon filter, preceded by a
  9      nitric acid denuder, to collect nitrate; and (3) a quartz fiber filter for elemental and organic
10      carbon (but without any correction for positive or negative artifacts caused by adsorption of
11      volatile organic compounds on the quartz filters or evaporation of semivolatile organic
12      compounds from the collected particles).
13           The EPA expects that more than 200 local agency monitoring sites throughout the United
14      States will operate continuous PM monitors. However, EPA has not yet provided any guidance
15      regarding appropriate  continuous monitoring techniques.  All currently available continuous
16      measurements of suspended particle mass share the problem of dealing with semivolatile PM
17      components (i.e., so as not to include particle-bound water as part of the mass, the particle-bound
18      water must be removed by heating or dehumidification). However, heating also causes loss of
19      ammonium nitrate and semivolatile organic components.  Several candidates for continuous PM
20      mass measurements, which use dehumidification instead of heating to remove particle-bound
21      water, currently are being field tested.  In addition to continuous mass measurement, a number of
22      techniques for continuous  measurement of sulfate, nitrate, or elements are being tested.  Aerosol
23      time-of-flight mass spectroscopy provides a new technique for real-time measurement of
24      correlated size and composition profiles of individual atmospheric aerosol particles.
25           For measurement of the chemical composition of PM collected on a filter, adequate
26      techniques exist for measurement of the heavier elements (Na and higher); sulfate, nitrate,
27      ammonium, and hydrogen ions; and total carbon. The split between elemental carbon and
28      organic carbon is defined operationally and depends on the measurement technique used. The
29      definition of elemental carbon (measured by oxidation to CO2 and quantification of the CO2
30      formed) and black carbon (measured by optical absorption or transmission) is also operational
31      and determined by the methods used. Determination of the mass of organic material (carbon plus

        March 2001                               9-17        DRAFT-DO NOT QUOTE OR CITE

-------
 1      molecularly bound hydrogen and oxygen) remains a problem, as does the identification of the
 2      many individual organic compounds. However, measurement techniques for polynuclear
 3      aromatic hydrocarbons (PAH) and some other toxic compounds in PM are well developed.
 4
 5      9.3.2  Patterns and Trends in U.S. Particulate Matter Concentrations
 6          Since the  1987 setting of the PM,0 NAAQS, extensive PM10 monitoring has been carried
 7      out throughout the United States, allowing for confident characterization of PMIO patterns and
 8      trends during the past decade or so. However, only very recently, with the deployment of a
 9      nationwide PM2 5 monitoring network during 1998, has it became possible to begin, in a
10      systematic fashion, to characterize PM25 patterns and trends, starting with data for 1999.
11
12      9.3.2.1  PM10 Trends and Concentrations
13          Annual average PM10 mass concentrations throughout the United States, for different
14      regions within the United States, and for most subregions or cities, have generally decreased over
15      the past decade. Nationwide average PM10 concentrations decreased from 31.7 /ug/m3 in 1989 to
16      23.7 Aig/m3 in 1998.  Decreases were  largest (38%) in the Pacific Northwest and smallest in the
17      Southeast (18%).
18          Annual mean PMIO concentrations in urban areas, found in EPA's Air Information
19      Retrieval System  database (Fitz-Simons et al., 2000), generally were greater than about 20 /ug/m3
20      for 1999.  Annual average concentrations  above 50 //g/m3 are found in several locations in
21      southern and  central California, Nevada, Arizona, Texas, South Carolina, and Puerto Rico.
22      At rural sites  in national parks, wilderness areas, and national monuments in the western United
23      States, the annual average PM10 concentrations were in the  range of 5 to 10 /ug/m3. Higher PM10
24      concentrations have been reported at  some rural sites in the eastern United States. The
25      corresponding PM2 5 concentrations in western rural or remote sites were approximately 3 Aig/m3
26      and, in eastern rural or remote sites, were  in the range of 5 to 10 yUg/m3.
27          A few attempts to infer various  types of "background" levels of PM25 and PM10 have been
28      made.  The background levels most relevant to the present criteria document include: (l)an
29      uncontrollable"background" (which includes the "natural background" defined below and
30      anthropogenic sources outside of North America), and (2) a "natural background" (which
31      includes all natural sources but excludes all anthropogenic  sources anywhere in the world).
        March 2001                              9-18        DRAFT-DO NOT QUOTE OR CITE

-------
  1     Annual average background levels of PM10 (according to the first definition) have been estimated
  2     to range from 4 to 8 /ug/m3 in the western United States and 5 to 11 Aig/m3 in the eastern United
  3     States. Corresponding PM2 5 background levels have been estimated to range from 1 to 4 ,ug/m3
  4     in the western United States and from 2 to 5 /wg/m3 in the eastern United States.
  5
  6     9.3.2.2 PM2 5 Trends and Concentrations
  7          The recently deployed PM25 FRM network has returned data for a large number of sites
  8     across the United States beginning in January of 1999. As of the end of 1999, the network
  9     consisted of 1025 monitors. Annual mean PM2 5 concentrations for 1999 ranged from about
 10     5 yUg/m3 to more than 20 //g/m3.  As might be expected, annual average PM2 5 concentrations
 11      towards the low end of the range were found in relatively small, nonindustrialized cities such as
 12     Bangor, ME; Fargo, ND; Cheyenne, WY; and Albuquerque, NM. Higher annual averages were
 13     found in larger urban areas such as Atlanta, GA, and Los Angeles, CA, as well as in a number of
 14     urban areas in the eastern United States. Because FRM measurements of PM25 only began in
 15     January 1999, data tend to be limited in many areas, especially for the first quarter. However, a
 16     number of observations can be made regarding  PM2 s concentrations and the patterns of seasonal
 17     variability in urban areas across the United States.  Generally, similar patterns of seasonal
 18     variability were found at all sites within Metropolitan Statistical Areas (MSAs) sampled
 19     nationwide, although there were exceptions at individual sites, which may have been related to
 20     contributions from local sources as opposed to contributions from regional background sources.
 21      At sites in the eastern United States, highest quarterly mean values and maximum values
 22      occurred during the third quarter (summer) of 1999, with exceptions occurring at several
 23      locations. For example, at monitoring sites in Miami and Puerto Rico, maximum concentrations
 24      occurred during the second quarter and may have been related to the transport of dust from the
 25      Sahara Desert.  At sites west of the Mississippi River, highest mean values occurred during the
26      first or fourth quarter (winter or autumn) of 1999, and, again, there were exceptions. Because of
27      the limited nature of these  data, definitive conclusions regarding long-term patterns of seasonal
28      variability cannot be drawn from these data alone.  These findings are generally consistent with
29      those based on longer term data sets such as the Metropolitan Acid Aerosol Characterization
30      Study (MAACS) in the eastern United States and the California Air Resources Board (CARS)
31      network of dichotomous samplers in California. Very limited data sets are available for obtaining

        March 2001                                9-19        DRAFT-DO NOT QUOTE OR CITE

-------
  1      trends in PM2 5 concentrations in urban areas. Data obtained by the CARB indicate that annual
 2      average PM2 5 concentrations decreased from 35 to 50% in large urban areas in California from
 3      1990 to 1995. Smaller decreases ranging from 2 to 34% were observed as part of the children's
 4      health study in southern California. In contrast, the urban IMPROVE site in Washington, DC
 5      measured only a 5% decline in PM2 5 concentrations from 1989-1997.
 6
 7      9.3.2.3 Spatial Variability in PM2.5 Concentrations
 8           The 1999 FRM PM2 5 data indicate that, in general, PM2 5 concentrations are highly
 9      correlated among sites within several MSAs (Atlanta, GA; Detroit, MI; Phoenix-Mesa, AZ; and
10      Seattle-Bellevue-Everett, WA), although there are some exceptions to this rule.  These findings
11      are consistent with those of earlier studies in Philadelphia, PA; and Los Angeles, CA.
12      Concentrations of PM25 also tended to be highly correlated on much larger spatial scales in many
13      areas in the United States, supporting the inference that PM2 5 is a regionally distributed pollutant.
14
15      9.3.2.4 Relationships Among Particulate Matter  in Different Size Fractions
16           PM25 to PM10 ratios from the FRM network were generally higher in the eastern (=0.7)
17      than in the central or western (-0.5) United States during  1999. These values are consistent with
18      those found in numerous earlier studies presented in the 1996 PM AQCD.
19           The results of ambient monitoring studies and  receptor modeling studies in the eastern
20      United States indicate that PM2 5 is dominated by secondary components.  Depending on the
21      origin of OC in ambient samples, PM2 5, on average, also may be dominated by secondary
22      components throughout the rest of the United States. Primary constituents represent smaller but
23      still important components of PM25, on average.  Crustal materials constitute the largest fraction
24      of PM(10.2 5) throughout the United States.  Crustal materials in the lower tail of the coarse-mode
25      particles also may be present in the PM2 5-size fraction. Data collected in several airsheds,
26      including the Los Angeles Basin, Bakersfield and Fresno, CA; and Philadelphia, PA, suggest that
27      secondary PM components are more uniformly distributed than are primary components.
28      Compositional data obtained at multiple sites in other urban areas are sparse.
29
        March 2001                               9-20        DRAFT-DO NOT QUOTE OR CITE

-------
  1      9.3.2.5 Short-Term Temporal Variability of Particulate Matter Concentrations
  2           Hour-to-hour changes in PM2 5 concentrations have been obtained at 31 sites by various
  3      continuous monitors. The 1999 nationwide composite circadian variability in PM2 5
  4      concentrations obtained by these monitors indicate two typical intra-day peaks. The first peak
  5      occurs from about 6 to 9 a.m. and the second peak occurs from about 5 to 10 p.m. The amplitude
  6      of these peaks is much smaller than the daily mean concentration.  It also should be noted that
  7      this pattern may not be apparent in the data obtained by any given monitor on any given day.
  8      Although the 98th percentile values  for positive and negative excursions in 24-h PM2 5
  9      concentrations are typically less than 20 /wg/m3, maximum hour-to-hour excursions may be over
10      200 yag/m3 in some locations.
11           The only data sets from which the long-term, day-to-day variability in PM2 5 and PM10
12      concentrations could be assessed, based on daily filter measurements, were obtained in
13      Philadelphia, PA, from 1992 to 1995 and in Phoenix, AZ, from 1995 through 1997.  In the
14      Philadelphia data set, average day-to-day concentration differences obtained were
15      6.8 ± 6.5 Mg/m3 for PM2 5 and 8.6 ± 7.5 Mg/m3 for PM10, whereas maximum day-to-day
16      differences obtained  were 54.7 /ig/m3 for PM2 5 and 50.4 //g/m3 for PM10. In the Phoenix, AZ,
17      data set, average day-to-day PM2 5 concentration differences were 2.9 ± 3.0 /wg/m3, and the
18      maximum day-to-day concentration difference was 23 /wg/m3.
19
20      9.3.3  Sources of Particulate  Matter
21           As shown in Table 9-1, fine and coarse particles have different types of sources. The major
22      sources of fine and coarse PM are summarized in Table 9-2. Because of the complexity of the
23      composition of ambient PM2 5 and PM(IO_2 5), sources are best discussed in terms of individual
24      constituents of both primary and secondary PM2 5 and PM(IO_2 5). Each of these constituents can
25      have anthropogenic and natural sources, as shown in Table 9-2. The distinction between natural
26      and anthropogenic sources is not always obvious. For example, although windblown dust might
27      seem to be the result  of natural processes, highest emission rates are associated with agricultural
28      activities in areas that are susceptible to periodic drought,  such as in the dust bowl region of the
29      mid-western United States. Also, most forest fires in the United States could be classified as
30      human in origin, either through prescribed burning, by accident, or through forest management
31      practices, which allow the buildup of combustible material, thereby increasing the likelihood of
        March 2001                               9-21        DRAFT-DO NOT QUOTE OR CITE

-------
s
p
3
cr
to
o
o
            TABLE 9-2.  CONSTITUENTS OF ATMOSPHERIC PARTICLES AND THEIR MAJOR SOURCES
o
z
o
H
O
C
O
H
tfl
O
?o
n
HH
H
W
Sources
Pnmary (PM < 2. 5 ^m) Primary (PM > 2.5 ^m) Secondary PM Precursors (PM < 2.5 /urn)
Aerosol
species Natural Anthropogenic Natural
SO4" Sea spray Fossil fuel combustion Sea spray
Sulfate
Anthropogenic Natural
— a Oxidation of reduced sulfur
gases emitted by the oceans
and wetlands and SO2 and
H2S emitted by volcanism
and forest fires
Anthropogenic
Oxidation of SO, emitted
from fossil fuel
combustion1'
        Nitrate
        Minerals
                                                                                          Oxidation of NOX produced
                                                                                          by soils, forest fires, and
                                                                                          lighting
                                                                                       Oxidation of NO, emitted
                                                                                       from fossil fuel combustion
                                                                                       and in motor vehicle
                                                                                       exhaust
             Erosion and
             reentramment
Fugitive dust, paved
and unpaved roads, and
agriculture and forestry
Erosion and
reentrainment
Fugitive dust, paved
and unpaved road
dust, and agriculture
and forestry
NH4+
Ammonium

Organic
Carbon (OC)


Elemental
Carbon
(EC)
Metals


Bioaerosols



—


Wildfires



Wildfires


Volcanic
activity

Viruses and
bacteria


—


Prescribed burning,
wood burning, motor
vehicle exhaust, and
cooking
Motor vehicle
exhaust, wood
burning, and cooking
Fossil fuel
combustion, smelting,
and brake wear
—



— —


— Tire and asphalt wear
and paved road dust


— —


Erosion, reentrainment, —
and organic debris

Plant, insect fragments, —
pollen, fungal spores,
and bacterial
agglomerates
Emissions of NH3 from
wild animals and
undisturbed soil
Oxidation of hydrocarbons
emitted by vegetation,
(terpenes, waxes) and
wildfires
—


—


—



Emissions of NH, from
animal husbandry, sewage,
and fertilized land
Oxidation of hydrocarbons
emitted by motor vehicles,
prescribed burning, and
wood burning
—


—


—



"Dash (—) indicates either very minor source or no known source of component.
""Major source of each component shown in boldface type.

-------
  1      fire from whatever cause. As seen in Table 9-2, emissions of crustal material (mineral dust),
  2      organic debris, and sea spray are concentrated mainly in the coarse fraction of PM10 (>2.5 /u.m
  3      AED). A small fraction of this material is in the PM2 5 size range (<2.5 /urn AED). Nevertheless,
  4      the concentrations of crustal material can be appreciable, especially during dust events.
  5      Emissions from combustion sources (mobile and stationary sources, biomass burning, etc.) are
  6      predominantly in the PM2 5 size range.
  7           The results of receptor modeling studies throughout the United States indicate that the
  8      combustion of fossil and biomass fuels is a major source of PM25.  Fugitive dust, found mainly in
  9      the PM(10_2 5) range size, represents the largest source of PM10 in many locations in the western
 10      United States.  Quoted uncertainties in source apportionments of constituents in ambient aerosol
 11      samples typically range from 10 to 50%.  It is apparent that a relatively small number of source
 12      categories, compared to the total number of chemical species that are typically measured in
 13      ambient monitoring-source receptor model studies, are needed to account for most of the
 14      observed mass of PM in these studies.
 15           Although most emphasis in this discussion has been on sources within the United States,
 16      it should be remembered that sources outside the United States also contribute to ambient PM
 17      levels in the United States that can, at times, exceed the ambient NAAQS level for PM. Perry
 18      et al. (1997) have found that the highest concentrations of mineral dust in the PM2 5 fraction are
 19      found in the eastern United States during the summer and not in arid areas of the western
 20      United States.  This dust originates from the Sahara Desert and is then transported across the
 21      Atlantic Ocean. Much of the Saharan dust that reaches the United States is in the PM2 5 size
 22      range.  Large-scale dust storms in the deserts of central Asia also have contributed to PM levels
 23      in the Northwest on an episodic basis.  In  addition, uncontrolled biomass burning in Central
 24      America and Mexico occasionally contributes to elevated U.S. ambient PM levels, having led at
 25      times to brief exceedances of daily PM NAAQS level in Texas.  Wildfires throughout the United
26      States, Canada, Mexico, and Central America all contribute to background concentrations of PM
27      in the United States.
28
29
        March 2001                               9-23        DRAFT-DO NOT QUOTE OR CITE

-------
  1      9.4 HUMAN EXPOSURES TO AMBIENT PARTICULATE MATTER
  2           The concentration of PM in the air inhaled by a person is not necessarily the same as that
  3      measured at a community ambient-air monitoring station. Personal exposure is defined as the
  4      concentration, integrated over a time period, of PM near the breathing zone but not influenced by
  5      exhaled breath. Total personal exposure, including ambient and nonambient PM, may be
  6      measured by a personal exposure monitor (PEM) carried by the person. There are several
  7      reasons why an individual's personal exposure may be different from the ambient concentration.
  8      First, the concentration of PM outside a person's home may be different from the concentration
  9      measured at a monitoring station.  However, for cities where sufficient information is available
10      (e.g., see Section 9.3.2), PM concentrations measured at different pairs of stations (sited to
11      measure community-wide pollution levels rather than individual source contributions) have been
12      found to be highly correlated for PM2 5 and PM10, but not so highly correlated for PM,0_2 5.
13      Although it would be desirable to check the spatial variability of PM indicators in each city
14      where epidemiologic studies are conducted, it seems likely that PM2 5 and PM,0 concentrations
15      are distributed evenly enough so that one site, or the average of several sites, provides an
16      adequate measure of the community average concentration for PM25 and PM10. This may not be
17      the case for PM10.2 5, for specific chemical components, for source contributions, or for sites
18      located near sources.
19           Second, the concentration of ambient PM found indoors is generally less than the
20      concentration of ambient PM outdoors. Ambient air, and the ambient PM it contains, penetrates
21      indoors through open doors and windows and through small openings in the building structure.
22      An equal volume of indoor air moves out of the indoor microenvironment.  Unless the air
23      exchange rate is very high, the ambient PM that penetrates indoors will be removed by deposition
24      more rapidly than it can be replaced. The ratio of ambient PM indoors to ambient PM outdoors,
25      called the infiltration factor, depends on the air exchange rate and, also, on the penetration
26      efficiency and deposition or removal rate, both of which vary with particle aerodynamic size.
27      The infiltration factor is a maximum for particles within the accumulation mode (=0.3 to 0.7 /um
28      AED) and decreases for smaller (ultrafine) or larger (coarse-mode) particles. For a given size
29      particle, the relationship between the indoor and outdoor PM concentration, given by the
30      infiltration factor, will vary with the air exchange rate. For a home closed for heating or air-
31      conditioning, the air exchange rate depends on the temperature difference between the indoor and
        March 2001                               9-24        DRAFT-DO NOT QUOTE OR CITE

-------
  1      outdoor air; the greater the difference, the greater the air exchange rate. If windows are opened
  2      for ventilation or doors are opened frequently, the air exchange rate will be higher.
  3           The relationship between ambient concentration of PM and personal exposure to ambient
  4      PM also is modulated by the time spent outdoors, because, while outdoors, a person is exposed to
  5      the ambient concentration. The ratio of personal exposure to ambient PM to the ambient PM
  6      concentration is called the attenuation factor and is given the symbol a. Both the infiltration
  7      factor and the attenuation factor, a, may be estimated by several techniques.  The most direct
  8      method is to measure the personal exposure and ambient concentration of a chemical species that
  9      has no indoor sources and is in the same size range as the PM component of interest.  Candidates
 10      are sulfate and, in homes with no open combustion, elemental carbon. The ratio of the personal
 11      exposure to the tracer to the ambient concentration of the tracer gives a. In turn, a times the
 12      ambient concentration of the appropriate PM indicator gives the personal  exposure to  that
 13      component of ambient PM.
 14           Personal exposure also contains a component resulting from indoor  sources of PM, which
 15      tend to produce ultrafme and coarse-mode particles rather than accumulation-mode particles.
 16      Important indoor sources are tobacco smoke and other open combustion (ultrafme); cleaning,
 17      sweeping, dusting, vacuuming (coarse); oven cooking (ultrafme); and resuspension caused by
 18      walking on rugs (coarse).  Stove-top cooking produces both ultrafme and  coarse-mode particles.
 19      Vacuum cleaners may produce ultrafme carbon  or copper particles from motor brushes.
 20      Another recently identified indoor source involves PM generated by the reaction of ozone (which
 21      infiltrates with ambient air) with terpenes from air fresheners or cleaning agents.
 22      Indoor-generated ultrafine particles will grow into the accumulation mode unless they are
 23      removed first by deposition or air exchange. Indoor-generated PM would be  expected to have a
 24      lower proportion of transition and toxic metals and highly oxidized and nitrated organic
 25      compounds than ambient air.
 26           Community time-series epidemiology studies evaluate the daily totals of deaths (or other
 27      health outcomes) in the community in relation to concentrations of one or more air pollutants
 28      measured at stationary community ambient air monitoring sites (assumed to be representative of
29      the community average).  There has been some controversy over whether the  ambient
30      concentration should be considered to be a surrogate for total human personal exposure or only
31      for exposure to the ambient-generated component of total personal exposure.  Some exposure

        March 2001                               9-25        DRAFT-DO NOT QUOTE OR CITE

-------
 1     analysts feel that ambient concentrations represent a surrogate for total personal exposure (the
 2     sum of exposure to ambient-generated pollution plus exposure to nonambient exposure).  This
 3     view is difficult to reconcile with epidemiologic studies that find statistically significant
 4     relationships between ambient concentrations and health outcomes, even though correlations of
 5     ambient concentrations with total personal exposures are found to be near zero. On the other
 6     hand, certain other scientists have argued that the ambient concentration represents a surrogate
 7     only for exposure to the ambient-generated component of total PM exposure.
 8          Recent studies of exposure error suggest that, provided ambient-generated and nonambient
 9     PM have equal toxicity, the increase in health outcomes per unit increase in concentration,
10     compared to PE, the increase in health outcome per unit increase in exposure, will not be biased
11     by the nonambient component of exposure if the nonambient component is independent of the
12     ambient concentration. Both logic and experiment suggest that nonambient PM exposures are
13     independent of daily ambient concentrations.  However, the increase in health outcomes per unit
14     increase in ambient concentration will be biased low compared to the increase in health outcomes
15     per unit increase in exposure.  For a constant average ratio of exposure to ambient-generated PM
16     to PM concentration (the attenuation factor, a, discussed earlier), the bias will be given by this
17     ratio which might be expected to  vary from a few tenths (0.1 s) to nearly 1.0 depending on air
18     exchange and indoor removal rates. Thus, it seems reasonable to conclude that community
19     time-series epidemiology studies  provide information on the statistical association of exposure to
20     ambient-generated pollutants with health outcomes, but do not provide any information on the
21     relationship of nonambient exposure with health outcomes.  It is likely that the nonambient
22     component of total personal exposure also has health effects. However, techniques other than
23     community time-series epidemiology must be used to identify relationships between nonambient
24     exposure and health outcomes.
25
26
27     9.5 DOSIMETRY CONSIDERATIONS
28          A basic health effects assessment principle is that dose delivered to the target site, rather
29     than external exposure, is the proximal cause of biological responses. Characterization of an
30     exposure-dose-response continuum for PM (key objective of any dose-response assessment for
31     evaluation of health effects) requires elucidation of mechanistic determinants of inhaled particle
       March 2001                               9-26        DRAFT-DO NOT QUOTE OR CITE

-------
  1      dose, which depend initially on deposition of particles in the respiratory tract. Once deposited on
  2      respiratory tract surfaces, particles undergo absorptive or nonabsorptive removal (clearance)
  3      processes that may result in their removal from airway surfaces and translocation from the
  4      respiratory tract. Clearance depends on initial site of deposition and physicochemical properties
  5      of the particles; both impact translocation mechanisms.  Retained particle burdens are determined
  6      by dynamic relationships between deposition and clearance mechanisms.  The dose from inhaled
  7      particles deposited and retained in the respiratory tract is governed by many factors (e.g.,
  8      exposure concentration and duration, respiratory tract anatomy and ventilatory parameters, and
  9      physicochemical properties of the particles (e.g., particle size, hygroscopicity, solubility).
 10           Particles exist in the atmosphere as aerosols (i.e., airborne suspensions of finely dispersed
 11      solid or liquid particles).  As noted in Chapter 2 and Section 9.2.1, the most commonly used
 12      metric AED, whereby particles of differing geometric size, shape, and density are compared
 13      aerodynamically with the instability behavior (i.e., terminal setting velocity) of particles that are
 14      unit density (1 gm/cm3) spheres.  Importantly, aerosols present in natural and work environments
 15      have polydisperse size distributions (i.e., particles within an aerosol have a range of sizes most
 16      appropriately described by a size distribution).  Aerosol size distributions are frequently modeled
 17      by a sum of lognormal distributions, one for each mode (nuclei, accumulation, and coarse). Two
 18      parameters needed to describe a log normal distribution of aerosol particle sizes are the median
 19      diameter and the geometric standard deviation. When using aerodynamic diameters, the mass
20      median aerodynamic diameter (MMAD) refers to the median of the distribution of mass with
21      respect to the AED, the most commonly used measure of aerosol distribution.
22           As Chapter 7 notes, for dosimetry purposes, the respiratory tract can be divided into three
23      main regions: (1) extrathoracic (ET), (2) tracheobronchial (TB), and (3) alveolar (A).  The ET
24      region consists of head airways (i.e., nasal or oral passages) through the larynx, the areas through
25      which inhaled air first passes.  In humans, inhalation can occur via the nose or mouth or both
26      (i.e., oronasal breathing).  From the ET region, inspired air enters the TB region at the trachea.
27      From the trachea, the conducting airways then undergo branching for several generations.
28      The terminal bronchioles are the most peripheral of the distal conducting airways and these lead,
29      in humans, to respiratory bronchioles, alveolar ducts, alveolar sacs, and alveoli, all comprising
30      the A region.
31

        March 2001                                9-27        DRAFT-DO NOT QUOTE OR CITE

-------
 1      9.5.1  Particle Deposition in the Respiratory Tract
 2           Knowledge of respiratory tract regional deposition patterns for particles of different sizes is
 3      important for understanding possible health effects associated with exposure to ambient PM and
 4      for extrapolating and interpreting data obtained from studies of laboratory animals.  Particles
 5      deposited in various respiratory tract regions are subjected to large differences in clearance
 6      mechanisms and pathways and, consequently, retention times.
 7           Particles deposit in the respiratory tract by five mechanisms: (1) inertial impaction,
 8      (2) sedimentation, (3) diffusion, (4) electrostatic precipitation, and (5) interception.  Sudden
 9      changes in airstream direction and velocity cause inhaled particles to impact onto airway
10      surfaces. The ET and upper TB airways are dominant sites of inertial impaction, a key
11      mechanism for particles with AED >1 /urn. Particles with AED > 0.5 /urn mostly are affected by
12      sedimentation out of the airstream.  Both sedimentation and inertial impaction influence
13      deposition of particles in the same size range and occur in the ET and TB regions, with inertial
14      impaction dominating in the upper airways and gravitational settling (sedimentation) increasingly
15      more dominant in lower conducting airways. Particles with actual physical diameters <1 /urn are
16      increasingly subjected to diffusive deposition due to random bombardment by air molecules,
17      resulting in contact with airway surfaces. Particles circa 0.3 to 0.5 //m in size are small enough
18      to be little influenced by impaction or sedimentation and large enough to be minimally
19      influenced by diffusion, and so, they undergo the least respiratory tract deposition. The
20      interception potential of any particle depends on its physical size; fibers are of chief concern for
21      interception, their aerodynamic size being determined mainly by  their diameter.  Electrostatic
22      precipitation is deposition related to particle charge; effects of charge on deposition are inversely
23      proportional to particle size and airflow rate. This type of deposition is likely small compared to
24      effects of other deposition mechanisms and is generally a minor contributor to overall particle
25      deposition,  but one recent study found it to be a significant TB region deposition mechanism for
26      ultrafine, and some fine, particles.
27           Total  human respiratory tract deposition, as a function of particle size, is depicted in
28      Figure 9-6 for healthy male adults under different ventilation conditions. The ET region acts as
29      an efficient filter that reduces penetration of inhaled particles to the TB and A regions of the
30      lower respiratory tract.  Total respiratory tract deposition increases with particle size for particles
31      >1.0 jj.ro. AED, is at a minimum for particles 0.3 to 0.5 //m, and increases as particle size
        March 2001                               9-28        DRAFT-DO NOT QUOTE OR CITE

-------
            .0
            '55
             o
             Q.
            a
                100
                 80
                 60
40
                 20
                   0
                     ^    O
                             I
                        O  Human (Oral)
                        •  Human (Nasal)
                                           0   T
                          0.01
                            0.1                1.0
                         Particle Diameter (urn)
                       10
       Figure 9-6. Total human respiratory tract deposition (percent deposition of amount
                  inhaled) as a function of particle size. AH values are means with standard
                  deviations as available.  Particle diameters are aerodynamic (MMAD) for those
                  ^0.5 fj,m.

       Source: Modified from Schlesinger (1988).
1      decreases below that range.  The ET deposition is higher with nose breathing than for mouth

2      breathing, with increased ventilation rates associated with increasing levels of physical activity or

3      exercise leading to more oronasal breathing and increased delivery of inhaled particles to TB and

4      A regions in the lung.

5          Hygroscopicity, the propensity of a material for taking up and retaining moisture, is a

6      property of some ambient particle species and affects respiratory tract deposition. Such particles

7      can increase in size in humid air in the respiratory tract and, when inhaled, deposit according to
      March 2001
                              9-29
DRAFT-DO NOT QUOTE OR CITE

-------
  1      their hydrated size rather than their initial size. Compared to nonhygroscopic particles of the
  2      same initial size, deposition of hygroscopic aerosols in different regions varies, depending on
  3      initial size: hygroscopicity generally increases total deposition for particles with initial sizes
  4      larger than =0.5 /urn, but decreases deposition for smaller ones.
  5           Enhanced particle retention occurs on carinal ridges in the trachea and through segmental
  6      bronchi; and deposition "hot spots" occur at airway bifurcations or branching points.  Peak
  7      deposition sites shift from distal to proximal sites as a function of particle size, with greater
  8      surface dose in conducting airways than in the A region for all particle sizes. Whereas both fine
  9      (<2.5 /^m) and coarse (2.5 to 10 jam) inhalable particles deposit to about the same extent on a
10      percent particle mass basis in the trachea and upper bronchi, a distinctly higher percent of fine
11      particles deposit in the A region. However, surface number dose (particles/cm2/day) is much
12      higher for fine particles than for coarse, indicating much higher numbers of fine particles
13      depositing, with the fine fraction contributing upwards of 10,000 times greater particle number
14      per alveolar macrophage.
15           Ventilation rate, gender, age, and respiratory' disease status are all factors that affect total
16      and regional respiratory tract particle deposition. In general, because of somewhat faster
17      breathing rates and likely smaller airway size, women have somewhat greater deposition of
18      inhaled particles than men in upper TB airways, but somewhat lower A region deposition than
19      for men.  Children appear to show four effects:  (1) greater total respiratory tract deposition than
20      adults (possibly as much as 50% greater for those <14 years old than for adults >14 years),
21      (2) distinctly enhanced  ET region deposition (decreasing with age from 1 year), (3) enhanced TB
22      deposition for particles <5 /u.m, and (4) enhanced A region deposition (also decreasing with age).
23      Overall, given that children have smaller lungs and higher minute volumes relative to lung size,
24      they likely receive greater doses of particles per  lung surface area than adults for comparable
25      ambient PM exposures.  This and the propensity for young children to generally exhibit higher
26      activity levels and associated higher breathing rates than adults likely contribute to enhanced
27      susceptibility to ambient particle effects resulting from particle dosimetry factors. In contrast,
28      limited available data on respiratory tract deposition across adult age groups (18 to 80 years) with
29      normal lung function do not indicate age-dependent effects (e.g., enhanced deposition in healthy
30      elderly adults). Altered PM deposition patterns resulting from respiratory disease status may put


        March 2001                                 9-30        DRAFT-DO NOT QUOTE OR CITE

-------
  1     certain groups of adults (including some elderly), as well as certain groups of children, at greater
  2     risk for PM effects.
  3          Both information noted in the 1996 PM AQCD and newly published findings indicate that
  4     respiratory disease status is an especially important determinant of respiratory tract particle
  5     deposition. Of particular importance is the finding that chronic obstructive disease states
  6     contribute to more heterogenous deposition patterns and differences in regional deposition. One
  7     new study indicates that  people with COPD tend to breath faster and deeper than those with
  8     normal lungs (i.e., about 50% higher resting ventilation), and had ca. 50% greater deposition than
  9     age-matched healthy adults under typical breathing conditions and average deposition rates
 10     2.5 times higher under elevated ventilation rates.  Enhanced deposition appears to be associated
 11     more with the chronic bronchitic than the emphysematous component of COPD. In this and
 12     other new studies, fine-particle deposition increased markedly with increased degree of airway
 13     obstruction (ranging up to ca. 100% greater with severe COPD). With increasing airway
 14     obstruction and uneven airflow because of irregular obstruction patterns, particles tend to
 15     penetrate more into remaining better ventilated lung areas, leading to enhanced  focal deposition
 16     at airway bifurcations and alveoli in those A region areas. In contrast, TB deposition increases
 17     with increasingly more severe bronchoconstrictive states, as occur with asthmatic conditions.
 18          Differences between humans and animals in deposition patterns were summarized in the
 19     1996 PM AQCD and by  Schlesinger et al. (1997) and should be considered when relating
 20     biological responses obtained in laboratory animal studies to effects in humans. Various species
 21      used in inhalation toxicology studies serving as the basis for dose-response assessment may not
 22     receive identical doses in a comparable respiratory tract region (i.e., ET, TB, A) when exposed to
 23      the  same aerosol at the same inhaled concentration.
 24           New mathematical  modeling studies evaluate interspecies differences in respiratory tract
 25      deposition.  For example, Hofmann et al. (1996) found total deposition efficiencies for all
 26      particles (0.01, 1, and 10 /um) at upper and lower airway bifurcations to be comparable for rats
27      and humans, but when higher penetration probabilities from preceding airways in the human lung
28      were considered, bronchial deposition fractions were mostly higher for humans. For all particle
29      sizes, deposition at rat bronchial bifurcations was less enhanced on the carinas than in human
30      airways. Numerical simulations  of three-dimensional particle deposition patterns within selected
31      (species-specific) bronchial bifurcations indicated that interspecies differences in morphologic

        March 2001                                9-31        DRAFT-DO NOT QUOTE OR CITE

-------
  1      asymmetry is a major determinant of local deposition patterns. The dependence of deposition on
  2      particle size is similar in rats and humans, with deposition minima in the 0.1- to l-^m size range
  3      for both total deposition and deposition in the TB and A regions, but total respiratory tract and
  4      TB deposition was consistently higher in the human lung. Alveoli region deposition in humans
  5      was lower than in rat for 0.001- to 10-//m particles (deposition of such particles being highest in
  6      the upper bronchial airways), whereas it was higher for 0.1 - and 1 -//m particles in more
  7      peripheral airways (i.e., bronchiolar airways in rat, respiratory bronchioles in humans). In a new
  8      histology study, Nikula et al. (2000) examined particle retention in rats (exposed to diesel soot)
  9      and humans (exposed to coal dust).  In both, the volume density of deposition increased with
10      increasing dose,  hi rats, diesel exhaust particles were found mainly in lumens of the alveolar
11      duct and alveoli, whereas  in humans, retained dust was mainly in interstitial tissue.  Thus, in the
12      two species, different lung cells appear to contact retained particles and may result in different
13      biological responses with  chronic exposure.
14           The probability of any biological effect of PM in humans or animals depends on particle
15      deposition and retention, as well as underlying dose-response relationships.  Interspecies
16      dosimetric extrapolation must consider differences in deposition, clearance, and dose-response.
17      Even similar deposition patterns may not result in similar effects in different species, because
18      dose also is affected by clearance mechanisms and species sensitivity. Total number of particles
19      deposited in the lung may not be the most relevant dose metric by which to compare species;
20      rather, the number of deposited particles per unit surface  area may determine response. Even if
21      deposition is similar in rats and humans, there would be a higher deposition density in the rat
22      because of the smaller surface area of rat lung. Thus, species-specific differences in deposition
23      density are important when attempting to extrapolate health effects observed in laboratory
24      animals to humans.
25
26      9.5.2  Particle Clearance and Translocation
27           Particles depositing on airway surfaces may be cleared from the respiratory tract completely
28      or translocated to other sites within this system by regionally specific clearance mechanisms, as
29      follow: ET region—mucocialiary transport, sneezing, nose wiping and blowing, and dissolution
30      and absorption into blood; TB region—mucociliary transport, endocytosis by macrophages and
31      epithelial cells, coughing,  and dissolution and absorption into blood and lymph;
        March 2001                                9-32        DRAFT-DO NOT QUOTE OR CITE

-------
  1     A region—macrophages, epithelial cells, interstitial, and dissolution and absorption into blood
  2     and lymph.  Clearance routes from various respiratory tract regions are depicted in Chapter 7
  3     (Figures 7-2 and 7-3).
  4          Regionally specific clearance defense mechanisms operate to clear deposited particles of
  5     varying particle characteristics (size, solubility, etc.) from the ET, TB, and A regions and are
  6     variously affected by different disease states. For example, particles are cleared from the ET
  7     region by mucociliary transport to the nasopharynx area, dissolution and absorption into the
  8     blood, or sneezing, wiping or blowing of the nose, but such clearance is slowed by chronic
  9     sinusitis, bronchiectasis, rhinitis, and cystic fibrosis. Also, in the TB region, poorly soluble
 10     particles are cleared mainly by upward mucociliary transport or by phagocytosis by airway
 11     macrophages that move upward on the mucociliary blanket, followed by swallowing.  Soluble
 12     particles in the TB region are absorbed mostly into the blood and some by mucociliary transport.
 13     Although TB clearance is generally fast and much material is cleared in <24 h, the slow
 14     component of TB clearance (likely associated with bronchides 24 h and clearance
 16     half-times of about 50 days.  Bronchial mucous transport is slowed by bronchial carcinoma,
 17     chronic bronchitis, asthma, and various acute respiratory infections; these are disease conditions
 18     that logically would be expected to increase retention of deposited particle material and, thereby,
 19     increase the probability of toxic effects from inhaled ambient PM components reaching the TB
 20     region.  Also, spontaneous coughing, an important TB region clearance mechanism, does not
 21      appear to fully compensate for impaired mucociliary clearance in small airways and may become
 22     depressed with worsening airway disease, as seen in COPD.
 23           Clearance of particles from the A region via alveolar macrophages and their mucociliary
 24      transport is usually rapid (<24 h). However, penetration of uningested particles into the
25      interstitium increases with increasing particle load and results in increased translocation to lymph
26      nodes.  Soluble particles not absorbed quickly into the blood  stream and translocated to
27      extrapulmonary organs (e.g., the heart) within minutes also may enter the lymphatic system, with
28      lymphatic translocation probably being increased as other clearance mechanisms (e.g., removal
29      by macrophages) are taxed or overwhelmed under "particle overload" conditions.  Particles
30      <2 /urn clear to the lymphatic system at a rate independent of  size; particles of this size, more so
31      than those >5.0 /urn, are deposited significantly in the A region. Translocation into the lymphatic

        March 2001                                9-33        DRAFT-DO NOT QUOTE OR CITE

-------
 1      system is quite slow, and elimination from lymph nodes even slower (half-times estimated in
 2      decades). Focal accumulations of reservoirs of potentially toxic materials and their slow release
 3      for years after initial ambient PM exposure may account partially for the higher relative risks
 4      observed in epidemiologic studies to be associated with long-term ambient PM exposure beyond
 5      additive effects of acute PM exposures. Alveolar region clearance rates are decreased in human
 6      COPD sufferers and slowed by acute respiratory infections, and the viability and functioning of
 7      alveolar macrophages are reduced in human asthmatics and in animals with viral lung infections,
 8      this suggests that persons with asthma or acute lung infections are likely at increased risk for
 9      ambient PM exposure effects.
10           Differences in regional and total clearance rates between some species reflect differences in
11      mechanical clearance processes. The importance  of interspecies clearance differences is that
12      retention of deposited particles can differ between species and may result in differences in
13      response to similar PM exposures.  Hsieh and Yu (1998) summarize existing data on pulmonary
14      clearance of inhaled, poorly soluble particles in the rat, mouse, guinea pig, dog, monkey, and
15      human. Two clearance phases "fast" and "slow" in the A region are associated with mechanical
16      clearance along two pathways, the former with the mucociliary system and the latter with lymph
17      nodes. Rats and mice are fast clearers, compared  to other species. Increasing initial lung burden
18      results in an increasing mass fraction of particles cleared by the slower phase. As lung burden
19      increases beyond 1 mg particle/g lung, the fraction cleared by the slow phase increases to almost
20      100% for all species. The rate for the fast phase is similar in all species, not changing with
21      increasing lung burden, whereas the slow phase rate decreases with increasing lung burden.
22      At elevated burdens, the "overload" effect on clearance rate is greater in rats than in humans.
23
24      9.5.3 Deposition and Clearance Patterns  of Particles Administered by
25            Inhalation Versus Intratracheal Instillation
26           Inhalation is the most directly relevant exposure route for evaluating PM toxicity, but many
27      studies deliver particles by intratracheal instillation. Because particle disposition is a determinant
28      of dose, it is important to compare deposition and clearance of particles delivered by instillation
29      versus inhalation. It is difficult to compare particle deposition and clearance among different
30      inhalation and instillation studies because of differences in experimental methods and in
31      quantification of particle deposition and clearance.  Key points from a recent detailed evaluation

        March 2001                              9-34         DRAFT-DO NOT QUOTE OR CITE

-------
  1      (Driscoll et al., 2000) of the role of instillation in respiratory tract dosimetry and toxicology
  2      studies are informative.  In brief, inhalation may result in deposition within the ET region, the
  3      extent of which depends on the size of the particles used, but intratracheal instillation bypasses
  4      this portion of the respiratory tract and delivers particles directly to the TB tree. Although some
  5      studies indicate that short (0 to 2 days) and long (100 to 300 days postexposure) phases of
  6      clearance of insoluble particles delivered either by inhalation or intratracheal instillation are
  7      similar, others indicate that the percent retention of particles delivered by instillation is greater
  8      than for inhalation, at least up to 30 days postexposure. Another salient finding is that inhalation
  9      generally results in a fairly homogeneous distribution of particles throughout the lungs, but
 10      instillation is typified by heterogeneous distribution (especially in the A region) and high levels
 11      of focal particles. Most instilled material penetrates beyond the major tracheobronchial airways,
 12      but the lung periphery is often virtually devoid of particles. This difference is reflected in
 13      particle burdens within macrophages, those from animals inhaling particles being burdened more
 14      homogeneously and those  from animals with instilled particles showing some populations of
 15      cells with no particles and others with heavy burdens, and is likely to impact clearance pathways,
 16      dose to cells and tissues, and systemic absorption.  Exposure method, thus, clearly influences
 17      dose distribution that argues for caution in interpreting results from instillation studies.
 18
 19      9.5.4  Inhaled Particles as Potential Carriers of Toxic Agents
 20           It has been proposed that particles also may act as carriers to transport toxic gases into the
 21      deep lung. Water-soluble gases, which would be removed by deposition to wet surfaces in the
 22      upper respiratory system during inhalation, could dissolve in particle-bound water and be carried
 23      with the particles into the deep lung. Equilibrium calculations  indicate that particles do not
 24      increase vapor deposition in human airways.  However, these calculations do show that soluble
25      gases are carried to higher  generation airways (deeper into the lung) in the presence of particles
26      than  in the absence of particles.  In addition, species such as SO2 and formaldehyde react in
27      water, reducing the concentration of the dissolved gas-phase species and providing a kinetic
28      resistence to evaporation of the dissolved gas. Thus, the concentration of the dissolved species
29      may be greater than that predicted by the equilibrium calculations. Also, certain other toxic
30      species (e.g., nitric oxide [NO], nitrogen dioxide [NO2], benzene, polycyclic aromatic
31      hydrocarbons [PAH], nitro-PAH, a variety of allergens) may be absorbed onto solid particles and
        March 2001                                9-35        DRAFT-DO NOT QUOTE OR CITE

-------
  1      carried into the lungs. Thus, ambient particles may play important roles not only in inducing
  2      direct health impacts of their constituent components but also in facilitating delivery of toxic
  3      gaseous pollutants or bioagents into the lung and may, thereby, serve as key mediators of health
  4      effects caused by the overall air pollutant mix.
  5
  6
  7      9.6 HEALTH EFFECTS OF AMBIENT PARTICULATE MATTER
  8      9.6.1 Introduction
  9           This section evaluates available scientific evidence regarding the physiologic and health
10      effects of exposure to ambient PM. The three main objectives of this evaluation are (1) to
11      summarize and evaluate strengths and limitations of available epidemiologic findings; (2) to
12      assess the biomedical coherence of findings across studied endpoints; and (3) to evaluate the
13      biologic plausibility of available evidence in light of (a) linkages between specific PM
14      components and health effects and (b) various dosimetric, mechanistic, and pathophysiologic
15      considerations.
16           Epidemiologic findings are emphasized first because they provide the strongest body of
17      evidence directly relating ambient PM concentrations to biomedical  outcomes.  Numerous
18      epidemiologic studies have shown statistically significant associations of ambient PM  levels with
19      a variety of human health endpoints, including mortality, hospital admissions, emergency
20      department  visits, other medical visits,  respiratory illness and symptoms measured in community
21      surveys, and physiologic changes in pulmonary function. Associations have been consistently
22      observed between both short- and long-term PM exposure and these endpoints. The general
23      internal consistency of the epidemiologic database and available findings demonstrate  well that
24      notable human health effects are associated with exposures to ambient PM at concentrations
25      currently found in many geographic locations across the United States. However, many
26      difficulties still exist with regard to delineating the magnitudes and variabilities of risk estimates
27      for ambient PM, the ability to attribute  observed health effects  to specific PM constituents, the
28      time intervals over which PM health effects are manifested, the extent to which findings  in one
29      location can be generalized to other locations, and the nature and magnitude of the overall public
30      health risk imposed by ambient PM exposure.

        March 2001                               9-36       DRAFT-DO NOT QUOTE OR CITE

-------
  1           The etiology of most air-pollution-related health outcomes is highly multifactorial, and the
  2      impact of ambient air pollution exposure on these outcomes is often small in comparison to that
  3      of other etiologic factors (e.g., smoking). Also, ambient PM exposure usually is accompanied by
  4      exposure to many other pollutants, and PM itself is composed of numerous physical/chemical
  5      components.  Assessment of the health effects attributable to PM and its constituents within an
  6      already-subtle total air pollution effect is difficult even with well-designed studies. Indeed,
  7      statistical partitioning of separate pollutant effects may not characterize fully the etiology of
  8      effects that actually depend on simultaneous exposure to multiple air pollutants, hi this regard,
  9      several viewpoints existed at the time of the 1996 PM AQCD regarding how best to interpret the
 10      epidemiology data: one saw the PM exposure indicators as surrogate measures of complex
 11      ambient air pollution mixtures, and the reported PM-related effects as representative of those of
 12      the overall mixture; another held that reported PM-related effects are attributable to PM
 13      components (per se) of the air pollution mixture and reflect independent PM effects, and a third
 14      viewpoint holds that PM can be viewed both as a surrogate indicator, as well as a specific cause
 15      of health effects.
 16           Several  other key issues and problems also must be considered when attempting to interpret
 17      the data reviewed in this document. For example, although the epidemiology data provide  strong
 18      support for the associations mentioned above, questions remain regarding potential underlying
 19      mechanisms.  Although much progress has been made toward identification of anatomic sites at
 20      which particles trigger specific health effects and elucidation of biological mechanisms
 21      underlying induction of such effects, this area of scientific inquiry is still at an early stage.
 22      Nevertheless, compared to the lack of much solid evidence available in the 1996 PM AQCD,
 23      there now is a stronger basis for assessing biologic plausibility of the epidemiologic observations
 24      given notable improvement in conceptual formulation of reasonable mechanistic hypotheses and
 25      evidence bearing on such hypotheses. Several hypotheses are discussed later with  regard to
26      possible mechanisms by which ambient PM may exert human health effects, and new evidence is
27      discussed that tends to support a causal relationship between low ambient concentrations of PM
28      and observed  increased mortality or morbidity risks. At the same time, much still remains to be
29      done to identify more confidently specific causal agents among typical ambient PM constituents.
30
        March 2001                                9-37        DRAFT-DO NOT QUOTE OR CITE

-------
 1      9.6.2 Community-Health Epidemiologic Evidence for Ambient Particulate
 2            Matter Effects
 3           In recent years, epidemiologic studies showing associations of ambient air pollution
 4      exposure with mortality, exacerbation of preexisting illness, and pathophysiologic changes have
 5      increased concern about the extent to which exposure to ambient air pollution exacerbates or
 6      causes harmful health outcomes at pollutant concentrations now experienced in the United
 7      States. The PM epidemiology studies assessed in the 1996 PM AQCD implicated ambient PM
 8      as a likely key contributor to mortality and morbidity effects observed epidemiologically to be
 9      associated with ambient air pollution exposures. New studies appearing since the 1996 PM
10      AQCD are important in extending results of earlier studies to many more cities and in confirming
11      earlier findings.
12           In epidemiologic studies of ambient air pollution, small positive estimates of air pollutant
13      health effects have been observed quite consistently, frequently being statistically significant at
14      p < 0.05. If ambient air pollution promotes or produces harmful health effects, relatively small
15      effect estimates from current PM concentrations in the United States and many other countries
16      would generally be expected on biological and epidemiologic grounds. Also, magnitudes and
17      significance levels of observed air pollution-related effects estimates would be expected to vary
18      somewhat from place to place, if the observed epidemiologic associations denote actual effects,
19      because (a) not only would the complex mixture of PM vary from place to place, but also
20      (b) affected populations may differ in characteristics that could affect susceptibility to air
21      pollution health effects.  Such characteristics include sociodemographic factors, underlying
22      health status, indoor-outdoor activities, diet, medical care access, exposure to risk factors other
23      than ambient air pollution (such as extreme weather conditions), and variations in factors (e.g.,
24      air-conditioning) affecting human exposures to ambient-generated PM.
25           Although it has been argued by some that the observed effects estimates for ambient air
26      pollution are not sufficiently constant across epidemiologic studies and that epidemiologic
27      studies are trustworthy only if they show relatively large effects estimates (e.g., large relative
28      risks), these arguments have only limited weight in relation to ambient air pollution studies.
29      Also, in any large population exposed to ambient air pollution, even a small relative risk for a
30      widely prevalent health disorder could result in a substantial public health burden attributable to
31      air pollution exposure.

        March 2001                               9-38        DRAFT-DO NOT QUOTE OR  CITE

-------
  1           As noted above, small health effects estimates generally have been observed for ambient air
  2      pollutants, as would be expected on biological and epidemiologic grounds. In contrast to effects
  3      estimates derived for the 1952 London smog episode with relative risk (RR) exceeding 4.0 (i.e.,
  4      400% increase over baseline) for extremely high (>2 mg/m3) ambient PM concentrations, effects
  5      estimates in most current epidemiology studies at distinctly lower PM concentrations (often
  6      < 100 Mg/m3) are relatively small. The statistical estimates (1) are more often subject to small
  7      (but proportionately large) differences in estimated effects of PM and other pollutants; (2) may
  8      be sensitive to a variety of methodological choices; and (3) sometimes may not be statistically
  9      significant, reflecting low statistical power of the study design to detect a small but real effect.
 10           The ambient atmosphere contains numerous air pollutants, and it is important to continue to
 11      recognize that health effects associated statistically with any single pollutant may actually be
 12      mediated by multiple components of the complex ambient mix.  Specific attribution of effects to
 13      any single pollutant may therefore be overly simplistic. Particulate matter is one of many air
 14      pollutants derived from combustion sources, including mobile sources. These pollutants include
 15      PM, carbon monoxide (CO), sulfur oxides, nitrogen oxides, and ozone, all of which have been
 16      considered  in various epidemiologic studies to date.  Many volatile organic compounds (VOCs)
 17      or semivolatile compounds (SVOCs) also emitted by combustion sources or formed in the
 18      atmosphere have not yet been systematically considered in relation to noncancer health outcomes
 19      usually associated with exposure to criteria air pollutants. In many newly available
20      epidemiologic studies, harmful health outcomes are often associated with multiple combustion-
21      related or mobile-source-related air pollutants, and some investigators have raised the possibility
22      that PM may be a key surrogate or marker for a larger subset of the overall ambient air pollution
23      mix. This possibility takes on added potential significance to the extent that ambient aerosols
24      indeed may not only exert health effects directly attributable to their constituent components, per
25      se, but also  serve as carriers for more efficient delivery of water soluble toxic gases (e.g., O3,
26      NO2, SO2) deeper into lung tissue, as noted earlier in  Section 9.5.5. This suggests that airborne
27      particle effects may be enhanced by the presence of other toxic agents or mistakenly attributed to
28      them if their respective concentrations are highly correlated temporally. Thus, although
29      associations of PM with harmful effects continue to be observed consistently across most new
30      studies, the  newer findings do not fully resolve issues concerning relative contributions to the
31      observed epidemiologic associations of (1) PM acting alone, (2) PM acting in combination with

        March 2001                                9-39        DRAFT-DO NOT QUOTE OR CITE

-------
 1      gaseous co-pollutants, (3) the gaseous pollutants per se, and (4) the overall ambient pollutant
 2      mix.
 3           It seems likely that, for pollutants whose concentrations are not highly correlated, effects
 4      estimates in multipollutant models would be more biologically and epidemiologically sound than
 5      those in single-pollutant models, although it is conceivable that single-pollutant models also
 6      might be credible if independent biological plausibility evidence supported designation of PM or
 7      some other single pollutant as likely being the key toxicant in the ambient pollutant mix
 8      evaluated.  However, neither of these possibilities have been demonstrated convincingly, and
 9      scientific consensus as to optimal interpretation of modeling outcomes for time series air
10      pollution studies has not yet been achieved. Therefore, the choice of appropriate effects
11      estimates to employ in risk assessments for ambient PM effects remains a difficult issue.  Issues
12      related to confounding by co-pollutants, along with issues related to time scales of exposure and
13      response and concentration-response function, importantly apply to new epidemiologic studies
14      relating  concentrations of PM or correlated ambient air pollutants to hospital admissions,
15      exacerbation of respiratory symptoms, and asthma in children, to reduced pulmonary function in
16      children and adults, and to changes in heart rate, and heart rate variability in adults.
17           With considerable new experimental evidence also in hand, it is now possible to
18      hypothesize various ways in which ambient exposure to multiple air pollutants (including not
19      only PM acting alone but also in combination with others) could plausibly be involved in the
20      complex chain of biological events leading to harmful health effects in the human population.
21      The newer experimental evidence, therefore, adds considerable support for interpreting the
22      epidemiologic findings discussed below as being indicative of causal relationships between
23      exposures to ambient PM and consequent associated increased morbidity and mortality risks.
24
25      9.6.2.1  Short-Term Particulate Matter Exposure Effects on Mortality
26           This section focuses primarily on discussion of short-term PM exposure effects on
27      mortality, but also highlights some morbidity effects in relation to the mortality findings.
28      Morbidity effects are discussed more fully after discussion of long-term mortality effects in the
29      section following this one.
30
31
        March 2001                               9-40        DRAFT-DO NOT QUOTE OR CITE

-------
  1     9.6,2.1.1 Summary of Previous Findings on Short-Term Paniculate Matter Exposure-
  2              Mortality Effects
  3          Time series mortality studies reviewed in the 1996 PM AQCD provided strong evidence
  4     that ambient PM air pollution is associated with increased daily mortality.  The 1996 PM AQCD
  5     summarized about 35 PM-mortality time series studies published between  1988 and 1996.
  6     Available information from those studies was consistent with the hypothesis that PM is a causal
  7     agent in the mortality impacts of air pollution. The PM,0 relative risk estimates derived from the
  8     PM,0 studies reviewed in the 1996 PM AQCD suggested that an increase of 50 /ug/m3 in the 24-h
  9     average of PM10 is associated with an increased risk of premature total mortality (total deaths
 10     minus accidents and injuries) mainly on of the order of relative risk (RR) = 1.025 to 1.05 (i.e.,
 11     2.5 to 5.0% excess risk) in the general population, with statistically significant increases being
 12     reported more broadly across the range of 1.5 to 8.5% per 50 /ug/nr1 PMI0.  Higher relative risks
 13     were indicated for the elderly and for those with preexisting respiratory conditions. Also, based
 14     on the then recently published Schwartz et al. (1996a) analysis of Harvard Six City data, the 1996
 15     PM AQCD found the relative risk for excess total mortality in relation to 24-h fine-particle
 16     concentrations to be in the range of RR = 1.026 to 1.055 per 25 yug/m3 PM2 5 (i.e., 2.6 to 5.5%
 17     excess risk  per 25 /ug/m3 PM2 5). Relative risk estimates for morbidity and mortality effects
 18     associated with standard increments in ambient PM10 concentrations and for fine-particle
 19     indicators (e.g., PM25, sulfates,  etc.) were presented in Chapters 12 and 13 of the 1996 PM
 20     AQCD (see Appendix 9A), and those effect estimates are updated below in light of the extensive
 21     newly available evidence discussed in Chapter 6 of this document.
 22          Although numerous  studies reported PM-mortality associations, several important issues
 23     needed to be addressed in  interpreting those relative risks. The 1996 PM AQCD extensively
 24     discussed the following critical  issues: (1) seasonal confounding and effect modification,
 25     (2) confounding by weather, (3) confounding by co-pollutants, (4) measurement error,
 26     (5) functional form and threshold, (6) harvesting and life shortening; and (7) the roles of specific
 27     PM components.
28          Season-specific analyses are often not feasible because of small magnitudes of expected
29     effect size or small sample sizes (low power) available for some studies. Some studies had
 30     earlier suggested possible season-specific variations in PM coefficients, but it was not clear if
31      these were caused by peak variations in PM effects from season to season, varying extent of PM

        March 2001                               9-41        DRAFT-DO NOT QUOTE OR CITE

-------
 1      correlations with other co-pollutants, or weather factors during different seasons. The likelihood
 2      of PM effects being accounted for mainly by weather factors was addressed by various methods
 3      that controlled for weather variables in most studies (including some involving sophisticated
 4      synoptic weather pattern evaluations), and that possibility was found to be very unlikely.
 5           Many early PM studies considered at least one co-pollutant in the mortality regression, and
 6      an increasing number have examined multiple pollutants.  Usually, when PM indices were
 7      significant in single-pollutant models, addition of a co-pollutant  diminished the PM effect size
 8      somewhat, but did not eliminate PM associations. In  multiple-pollutant models performed by
 9      season, the PM coefficients became less stable, again possibly because of varying correlations of
10      PM with co-pollutants among seasonal or smaller sample sizes.  However, in many studies, PM
11      indices showed the highest significance in both single- and multiple-pollutant models. Thus,
12      PM-mortality associations did not appear to be seriously distorted by co-pollutants.
13           Interpretation of the relative significance of each pollutant  in mortality regression in
14      relation to its relative causal strength was difficult, however, because of lack of quantitative
15      information on pertinent exposure measurement errors among the air pollutants.  Measurement
16      errors can influence the size and significance of air pollution coefficients in time  series
17      regression analyses, an issue also important in assessing confounding among multiple pollutants,
18      because the varying extent of such errors among pollutants may influence corresponding relative
19      significance. The 1996 PM AQCD discussed several types of exposure measurement and
20      characterization errors, including site-to-site variability and site-to-person variability. These
21      errors are thought to bias the estimated PM coefficients downward in most cases, but there was
22      insufficient quantitative information available at the time to allow estimation of such bias.
23           The 1996 PM AQCD also reviewed evidence for threshold and various other functional
24      forms of short-term PM mortality associations. Some studies indicated that associations were
25      seen monotonically to even below the PM  standards.  It was considered difficult,  however, to
26      statistically identify a threshold from available data because of low data density at lower ambient
27      PM concentrations, potential influence of measurement error, and adjustments for other
28      covariates.  Thus, use of relative risk (rate  ratio) derived from log-linear Poisson models was
29      deemed adequate.
30           The extent of prematurity of death (i.e., mortality displacement [or harvesting]) in observed
31      PM-mortality associations has important public health policy implications.  At the time of the

        March 2001                                9-42        DRAFT-DO NOT QUOTE OR CITE

-------
  1     1996 PM AQCD review, only a few studies had investigated this issue. Although one of the
  2     studies suggested that the extent of such prematurity might be only a few days, this may not be
  3     generalized because this estimate was obtained for identifiable PM episodes. Insufficient
  4     evidence then existed to suggest the extent of prematurity for nonepisodic periods, from which
  5     most of the recent PM relative risks were derived.
  6          Only a few PM-mortality studies had analyzed fine particles and chemically specific
  7     components of PM. The Harvard Six Cities Study (Schwartz et al., 1996a) analyzed size-
  8     fractionated PM (PM2 5, PM10/15, and PM10/15.2 5) and PM chemical components (sulfates and H+).
  9     The results suggested that PM2 5 was associated most significantly with mortality among the PM
 10     components. Although FT was not significantly associated with mortality in this and earlier
 11     analyses, the smaller sample size for H+ than for other PM components made direct comparison
 12     difficult.  Also,  certain respiratory morbidity studies showed associations between hospital
 13     admissions and  visits with components of PM in the fine-particle range. Thus, the 1996 PM
 14     AQCD concluded that there was adequate evidence to suggest that fine particles play especially
 15     important roles  in observed PM mortality effects.
 16          Overall, then, the outcome of assessment of the above key issues in the 1996 PM AQCD
 17     can be thusly summarized:  (1) observed PM effects are not likely seriously biased by inadequate
 18     statistical modeling (e.g., control for seasonality); (2) observed PM effects are not likely
 19     significantly confounded by weather; (3) observed PM effects may be confounded or modified to
 20     some extent by co-pollutants, and such extent may vary from season to season; (4) determining
 21     the extent of confounding and effect modification by co-pollutants requires knowledge of relative
 22     exposure measurement/characterization error among pollutants (there was not sufficient
 23     information on this); (5) no clear evidence for any threshold for PM-mortality associations was
 24     reported (statistically identifying a threshold from existing data also was considered difficult, if
 25     not impossible); (6) some limited evidence for harvesting, a few days of life-shortening, was
26     reported for episodic periods (no study was conducted to investigate harvesting in nonepisodic
27     U.S. data); and (7) only a relatively limited number of studies suggested a causal role of fine
28     particles in PM-mortality associations, but in light of historical  data, biological plausibility, and
29     results from morbidity studies, a greater role for fine particles than coarse particles was suggested
30     as being likely.
31

        March 2001                               9-43        DRAFT-DO NOT QUOTE OR CITE

-------
 1      9.6.2.1.2 Updated Epidemiologic Findings for Short-Term Ambient Particulate Matter
 2               Exposure Effects on Mortality
 3           With regard to updating the assessment of PM effects in light of new epidemiologic
 4      information published since the 1996 PM AQCD, the most salient key points on relationships
 5      between short-term PM exposure and mortality (drawn from Chapter 6 discussions in this
 6      document) can be summarized as follows.
 7           Since the 1996 PM AQCD, there have been more than 70 new time-series PM-mortality
 8      analyses, several of which investigated multiple cities using consistent data analytical
 9      approaches. With only few exceptions, the estimated mortality relative risks in these studies are
10      generally positive, many are statistically significant, and they generally comport well with
11      previously reported PM-mortality effects estimates delineated in the 1996 PM AQCD.  There are
12      also now numerous additional studies demonstrating associations between short-term (24-h) PM
13      exposures and various morbidity endpoints.
14           Several new studies conducted time series analyses in multiple cities.  The  major advantage
15      of these studies over meta-analyses for multiple "independent" studies is the consistency in data
16      handling and model specifications, thus eliminating variation in results attributable to study
17      design.  Also, many of the cities  included in these studies were ones for which no earlier time
18      series analyses had been conducted. Therefore, unlike regular meta-analysis, they likely do not
19      suffer from omission of negative studies caused by publication bias. Furthermore, any spatial or
20      geographic variability of air pollution effects can be systematically evaluated in such multi-city
21      analyses.
22           PMJO Effect Size Estimates. In the NMMAPS (Samet et al., 2000a,b) analysis of the
23      90 largest U.S. cities, the combined nationwide relative risk estimate was about a 2.3% increase
24      in total mortality per SO-^g/m3 increase in PMI0. The NMMAPS effect size estimates  did vary
25      somewhat by U.S. region (see Figures 6-2 and 6-3), with the largest estimate being for the
26      Northeast (4.5% for a 1-day lag,  the lag typically showing maximum effect size for most U.S.
27      regions). Various other U.S. multi-city analyses, as well as single-city analyses, obtained PM10
28      effect sizes mainly in the range of 2.5 to 5.0% per 50-^g/m3 increase in PM,0. There is some
29      evidence that, if the effects over multiple days are considered, the effect size may be larger.
30      What heterogeneity existed for the estimated PMIO risks across NMMAPS cities  could not be
31      explained with the city-specific explanatory variables (e.g., as the mean levels of pollution and

        March 2001                                9-44         DRAFT-DO NOT QUOTE OR CITE

-------
  1     weather), mortality rate, sociodemographic variables (e.g., median household income),
  2     urbanization, or variables related to measurement error.
  3           Also, the multi-city APHEA study showed generally consistent associations between
  4     mortality and both SO2 and PM indices in western European cities, but not for central and eastern
  5     European cities. The pooled estimate of PM10-mortality relative risks for western European cities
  6     comport well with estimates derived from U.S. data. The contrast between western and
  7     central/eastern Europe results might result from possible differences in representativeness of
  8     exposure measures, air pollution mix or resultant toxicity, proportions of sensitive
  9     subpopulations, climate, etc.
 10           Certain other individual-city studies using similar methodology in analyses for each city
 11     (but not generating combined overall pooled effect estimates) also report variations in PM effect
 12     size estimates between cities and in their robustness to inclusion of gaseous copollutants in
 13     multi-pollutant models.  Thus, one  cannot entirely rule out that real differences may exist in
 14     excess risk levels associated with varying size distributions, number,  or mass of the chemical
 15     constituents of ambient PM; the combined influences of varying co-pollutants present in the
 16     ambient air pollution mix from location to location or season to season; or to variations  in the
 17     relationship between exposure and  ambient PM concentration.
 18          Nevertheless, there still appears to be reasonably good consistency among the results
 19     derived from those several new multi-city studies providing pooled analyses of data combined
 20     across multiple cities (thought to yield the most precise effect size estimates). Such analyses
 21      indicate the percent excess total (nonaccidental) deaths estimated per 50 jUg/m3  increase  in 24-h
 22     PM10 to be 2.3% in the 90 largest U.S. cities (4.5% in the Northeast region); 3.4% in 10  U.S.
 23      cities; 3.5%  in the eight largest Canadian cities; and about 2.0% in western European cities
 24      (using PM,0 = TSP*0.55).  These combined estimates are reasonably consistent with the range of
 25      PM10 estimates previously reported in the 1996 PM AQCD (i.e., 1.5 to 8.5% per 50 Aig/m3 PM)0).
 26      These and other excess risk estimates from many other individual-city studies comport well with
27      a number of new studies confirming increased cause-specific cardiovascular- and respiratory-
28      related mortality, and those noted below as showing ambient PM associations with increased
29      cardiovascular and respiratory hospital admissions  and medical visits.
30
        March 2001                                9-45        DRAFT-DO NOT QUOTE OR CITE

-------
 1           Fine and Coarse Particle Effect Size Estimates. Table 9-3 summarizes effects estimates
 2      (RR values) for increased mortality and/or morbidity associated with variable increments in
 3      short-term (24-h) exposures to ambient fine particles indexed by various fine PM indicators
 4      (PM2 5, sulfates, H+, etc.) in U.S. and Canadian cities.  Table 9-4 shows analogous effect size
 5      estimates for inhalable thoracic fraction coarse particles (i.e., PM^ 5). In both tables, studies
 6      that were highlighted in comparable tables in the 1996 PM AQCD are indicated by italics.
 7           The effect size estimates derived for PM2 5 as an ambient fine particle indicator (especially
 8      those based on directly measured versus estimated PM2 5 levels) generally appear to fall in the
 9      range of 2.0 to 8.5% increase in total (nonaccidental) deaths per 25-^g/m3 increment in 24-h
10      PM2 5 for U.S. and Canadian cities. Cause-specific effects estimates appear to fall mainly in the
11      range of 3.0 to 7.0% per 25 /wg/m3 24-h PM2 5 for cardiovascular or combined cardiorespiratory
12      mortality and 2.0 to 7.0% per 25 /ug/m3 24-h PM2 5 for respiratory mortality in U.S. cities.
13           In the 1996 PM AQCD, there was only one study,  the Harvard Six Cities study, in which
14      the relative importance of fine and coarse particles was examined. That study suggested that fine
15      particles, but not coarse particles, were associated with daily mortality. Now, more than
16      10 studies have analyzed both PM2 5 and PM,0_2 5 for their associations with mortality (see
17      Figure 9-7).  Although some of these studies (e.g., the Santa Clara County, CA, analysis and the
18      eight largest Canadian cities analysis) suggest that PM2 5 is more important than PM,0_2 5 in
19      predicting mortality fluctuations, several others (e.g., the Mexico City and Santiago, Chile
20      studies) seem to suggest that PM10.2 5 may be as important as PM2 5 in certain  locations (some
21      shown to date being drier, more arid areas).  Seasonal dependence of PM components'
22      associations observed in some of the locations (e.g., higher coarse [PM10_2 5] fraction estimates for
23      summer than winter in Santiago, Chile) hint at possible contributions of biogenic materials (e.g.,
24      molds, endotoxins, etc.) to the observed coarse particle effects in at least some locations.
25      Overall, for U.S. and Canadian cities, effect size estimates for the coarse fraction (PMI0.2  5) of
26      those inhalable thoracic particles capable of depositing in TB and A regions of the respiratory
27      tract generally appear to fall in the range of 0.5 to 6.0% excess total (nonaccidental) deaths per
28      25 Aig/m3 of 24-h PMIO_2 5. Respective increases  for cause-specific mortality are 3.0 to 8.0% for
29      cardiovascular and 3.0 to 16.0% for respiratory causes per 25-jUg/m3 increase in 24-h PM10_2 5.
30           Chemical Components of Particulate Matter. Several new studies examined the role of
31      specific chemical components of PM.  Studies of U.S. and Canadian cities showed mortality
        March 2001                               9-46         DRAFT-DO NOT QUOTE OR CITE

-------
   TABLE 9-3. EFFECT ESTIMATES PER VARIABLE INCREMENTS IN 24-HOUR
      CONCENTRATIONS OF FINE PARTICLE INDICATORS (PM2 5, SO=, H+)
                  FROM U.S. AND CANADIAN STUDIES*
Study Location
Indicator
RR(±CI)**per25-A25 ^g/m3) 2.868 (1.126, 7.250)
(<25 ^g/m3) 0.779 (0.610, 0.995)
1 .06 (NS, from figure)
1.003(0.992, 1.015)
1.118(1.013, 1.233)
1.053(1.018, 1.090)

1.043(1.028, 1.059)
1.057(1.001, 1.115)
1.018(0.946, 1.095)
1.030(1.011,1.050)

77.2 (±7.8)
72.2 (±7.4)
15. 7 (±9. 2)
18.7 (±10.5)
20.8 (±9.6)
29.6 (±21.9)
Median 14.7
Means 11. 3-30.5
13(2,105)
61.7(0.78,390.5)
nmol/m3
17.28 (-0.6, 72.6)
18(6,86)
13.0(0,42)
NR
22 (4, 86)
32.5(9.3, 190.1)
16.8(5,48)
15.6 (±9.2)

42.1 (±22.0)
39.9 (±18.0)
37.1 (±19.8)
13.3 (max 86)
March 2001
9-47
DRAFT-DO NOT QUOTE OR CITE

-------
   TABLE 9-3 (cont'd). EFFECT ESTIMATES PER VARIABLE INCREMENTS IN
 24-HOUR CONCENTRATIONS OF FINE PARTICLE INDICATORS (PM2 5, SO;, H+)
                  FROM U.S. AND CANADIAN STUDIES*
Study Location
Toronto, Canada0
Montreal, Canadap
Indicator
Est. PM25
PM25
RR (± CI)** per 25-^g/m3
PM Increase or 1 S-^g/m3
SOJ Increase or 75-nmol/m3 H+ increase
1.048(1.033, 1.064)
1.044(1.025, 1.063)
Reported PM
Levels Mean (Min,
Max)***
18.0(8,90)
17.4(2.2,72.0)
Cause-Specific Mortality
Cardiorespiratory:
Three New Jersey Ctties:M
Newark, NJ
Camden, NJ
Elizabeth, NJ
Total Cardiovascular:
Santa Clara County, CAC
Buffalo, NY)D
Philadelphia, PAF
(seven-county area)
Detroit, MI°
Phoenix, AZH
Los Angeles, CA1
San Bernadino and
Riverside Counties, CAJ
Coachella Valley, CAK
Cerebro vascular:
Los Angeles, CA1
Total Respiratory:
Santa Clara County, CAC
Buffalo, NYD
Philadelphia, PAF
(seven-county area)
Detroit, MIG
San Bernadino and
Riverside Counties, CAJ


PM25
PM25
PM25

PM25
S04=
PM25

PM25
PM25
PM25
Est. PM25
PM25

PM25

PM25
so;
PM25

PM25
Est. PM25


1.051 (1.031, 1.072)
1.062(1.006, 1.121)
1.023(0.950, 1.101)

1.07(p>0.05)
1.040(0.995, 1.088)
1. 028 (p< 0.055)

1.032(0.977,1.089)
1.187(1.057,1.332)
1.266(1.003, 1.048)
1.007(0.997, 1.017)
1.086(0.937,1.258)

1.036(0.994,1.080)

1.13(p>0.05)
1.108(1.007,1.219)
1.014 (p> 0.055)

1.023(0.897, 1.166)
1.021 (0.997, 1.045)


42.1 (±22.0)
39.9 (±18.0)
37.1 (±19.8)

13(2, 105)
61.7(0.78,390.5)
nmol/m3
17.28 (-0.6, 72.6)

18(6,86)
13.0(0,42)
22 (4, 86)
32.5(9.3, 190.1)
16.8(5,48)

22 (4, 86)

13(2,105)
61.7(0.78,390.5)
nmol/m3
17.28 (-0.6, 72.6)

18(6,86)
32.5(9.3, 190.1)
March 2001
9-48
DRAFT-DO NOT QUOTE OR CITE

-------
   TABLE 9-3 (cont'd). EFFECT ESTIMATES PER VARIABLE INCREMENTS IN
  24-HOUR CONCENTRATIONS OF FINE PARTICLE INDICATORS (PM2 5, SO;, H+)
                   FROM U.S. AND CANADIAN STUDIES*
Study Location
COPD:
Los Angeles, CA1
Increased Hospitalization
Ontario, Canada0
Ontario, Canada*
NYC/Buffalo, NY5
Toronto, Canada5
Total Respiratory:
King County, WAT
Toronto, Canada0
Buffalo, NYD
Montreal, Canadav
Montreal, Canada*
St. John, Canada"
Pneumonia:
Detroit, MIF
Respiratory infections:
Toronto, Canadau
COPD:
Atlanta, GAZ
Detroit, MIF
King County WAM
Los Angeles, CABB
Toronto, CanadaY
Indicator

PM25

SO'4
SO",
03
SO"4
H+ (Nmol/m3)
SO°4
PM2S

PM,
PM25
so;
PM25
PM25
PM25

PM25

PM25

PM25
PM25
PM25
PM25
PM25
RR(±CI)**per25-^g/m3
PM Increase or 1 5-//g/m3
SOJ Increase or 75-nmol/m3 H+ increase

1.027(0.966,1.091)

1.03 (1.02, 1.04)
1.03 (1.02, 1.04)
1.03 (1.02, 1.05)
1.05 (1.01, 1.10)
1.16(1.03, 1.30)'
1.12 (1.00, 1.24)
1.15 (1.02, 1.78)

1.058(1.011, 1.110)
1.085(1.034,1.138)
1.082(1.042, 1.128)
1.261 (1.059, 1.503)
1.137(0.998, 1.266)
1.057(1.006, 1.110)

1.125(1.037,1.220)

1.108(1.072,1.145)

1.124(0.921, 1.372)
1.055(0.953,1.168)
1.064(1.009,1.121)
1.051 (1.009,1.094)
1.048(0.998,1.100)
Reported PM
Levels Mean (Min,
Max)***

22 (4, 86)

R = 3.1-8.2
R= 2.0-7.7
NR
28.8 (NR/391)
7. 6 (NR, 48.7)
18.6(NR, 66.0)

NR
16.8(1,66)
61.7(0.78,390.5)
nmol/m3
Summer 93
12.2 (max 31)
18.6(SD9.3)
Summer 93
8.5 (max 53.2)

18(6,86)

18.0 (max 90)

19.4 (±9.35)
18(6,86)
18.1 (3,96)
Median 22 (4, 86)
18.0 (max 90)
March 2001
9-49
DRAFT-DO NOT QUOTE OR CITE

-------
   TABLE 9-3 (cont'd). EFFECT ESTIMATES PER VARIABLE INCREMENTS IN
 24-HOUR CONCENTRATIONS OF FINE PARTICLE INDICATORS (PMZ 5, SO=, IT)
                  FROM U.S. AND CANADIAN STUDIES*
Study Location
Asthma:
Atlanta, GAZ
Seattle, WACC
Seattle, WADD
Toronto, CanadaY
Total Cardiovascular:
Atlanta, GAY
Buffalo, NYD
Los Angeles, CAEE

St. John, Canada"
Toronto, Canada11
Ischemic Heart Disease:
Detroit, MIF
Toronto, CanadaY
Dvsrhvthmias:
Atlanta, GAZ
Detroit, MIF
Toronto, CanadaY
Heart Failure:
Detroit, MIF
Toronto, CanadaY
Cerebrovascular:
Los Angeles, CAEE
Toronto, CanadaY
Peripheral circulation diseases:
Toronto, CanadaY
Indicator

PM25
PM25
Est. PM25
PM25

PM25
so:
PM25

PM25
PM25

PM25
PM25

PM25
PM25
PM25

PM25
PM25

PM25
PM25

PM25
RR(±CI)**per25-^g/m3
PM Increase or 1 5-//g/m3
SC>4 Increase or 75-nmol/m3 H+ increase

1.023(0.852,1.227)
1.087(1.033,1.143)
1.445(1.217, 1.714)
1.064(1.025, 1.106)

1.061(0.969,1.162)
1.015(0.987, 1.043)
(65+) 1.043 (1.025, 1.061)
(<65) 1.035 (1.01 8, 1.053)
1.151 (1.006, 1.110))
1.059(1.018, 1.102)

1.043(0.986, 1.104)
1.080(1.054,1.108)

1.061 (0.874, 1.289)
1.032(0.934, 1.140)
1.061 (1.019, 1.104)

1.091 (1.023,1.162)
1.066(1.025,1.108)

1.015(0.992,1.038)
"NEG" reported

"NEC" reported
Reported PM
Levels Mean (Min,
Max)***

19.4 (±9.35)
16.7(6,32)
4.8(1.2,32.4)
18.0 (max 90)

19.4 (±9.35)
61.7(0.78,390.5)
nmol/m3
Median 22 (4, 86)

Summer 93
8.5 (max 53.2)
16.8(1,66)

18(6,86)
18.0 (max 90)

19.4 (±9.35)
18(6,86)
18.0 (max 90)

18(6,86)
18.0 (max 90)

Median 22 (4, 86)
18.0 (max 90)

1 8.0 (max 90)
March 2001
9-50
DRAFT-DO NOT QUOTE OR CITE

-------
   TABLE 9-3 (cont'd). EFFECT ESTIMATES PER VARIABLE INCREMENTS IN
 24-HOUR CONCENTRATIONS OF FINE PARTICLE INDICATORS (PM2 5, SO;, H+)
                  FROM U.S. AND CANADIAN STUDIES*
Study Location Indicator
Stroke:
Detroit, MIF PM25
Increased Respiratory Symptoms
Southern CaliforniaFF SO°4
SixCitiesCG PM25
(Cough) SO"4
Six Cities00 PM25
(Lower Resp. Symp.) SO°j
Uniontown, PAHH PM2i
(Evening Cough)
Connecticut summer camp" SOJ
State College, PAJJ PM2 ,
(Wheeze)
State College, PAJJ PM2 ,
(Cough)
State College, PAW PM2 ,
(Cold)
Decreased Lung Function
Uniontown, PAHH PM25
Uniontown, PA1* PM25
(Reanalysis)
State College, PA1^ PM25
(Reanalysis)
Connecticut summer camp" SO^
Southwest, VALL PM25
State College, PAJJ PM2 ,
RR (± CI)** per 25-^g/m3
PM Increase or 1 5-,ug/m3
SO; Increase or 75-nmol/m3 H+ increase

1.018(0.947, 1.095)
Odd Ratio (95% CI) per 25-Mg/m3
PM Increase or 15-/ug/m3
SO; Increase or 75-nmol/m3 H+ increase
1.48(1.14, 1.91)
1.24 (1.00, 1.54)
1.86(0.86,4.03)
1.19(0.66,2.15)
1.58(1.18,2.10)
6.82(2.09, 17.35)
1.16(0.10, 13.73)
1.45(1.07, 1.97)

1.71 (1.30,2.25)
1.59(0.94,2.71)

1.61 (1.21,2.17)

1.2.45(1.29,4.64)

PEFR change (L/min) per 25-^g/m3
PM Increase or 1 5-yUg/m3
SO; Increase or 75-nmol/m3 H+ increase
PEFR -1.38 (-2.77, 0.02)
pm PEFR -1 .52, (-2.80, -0.24)

pm PEFR -0.93 (-1.88, 0.01)

PEFR -5.4 (-12.3, 1.52)
am PEFR -1.825 (-3.45, -0.21)
pm PEFR -0.63 (-1.73, 0.44)
Reported PM
Levels Mean (Min,
Max)***

18(6,86)

R = 2-37
18.0 (max 86.0)
2. 5 (max 15.1)
18.1 (max 37 1.1)
nmol/m3
18.0 (max 86.0)
2.5 (max 15.1)
18.1 (max 371.1)
nmol/m3
24. 5 (max 88.1)

7.0(1.1,26.7)
23.5 (max 85.8)

23.5 (max 85.8)

23. 5 (max 85. 8)


24.5 (max 88.1)
24.5 (max 88.1)

23.5 (max 85.8)

7.0(1.1,26.7)
21.62(3.48,59.65)
23.5 (max 85.8)
March 2001
9-51
DRAFT-DO NOT QUOTE OR CITE

-------
    TABLE 9-3 (cont'd).  EFFECT ESTIMATES PER VARIABLE INCREMENTS IN
  24-HOUR CONCENTRATIONS OF FINE PARTICLE INDICATORS (PM2 5, SO,, H+)
                         FROM U.S. AND CANADIAN STUDIES*
Study Location
Philadelphia, PA1™
RR(±CI)**per25-^g/m3
PM Increase or 1 5-/ug/m3
Indicator SO^ Increase or 75-nmol/m3 H+ increase
PM25 am PEFR -3. 18 (-6.64, 0.07)
pmPEFR-0.91 (-4.04,2.21)
Reported PM
Levels Mean (Min,
Max)***
22.2 (IQR 16.2)
* Studies highlighted in the 1996 CD are in italics; new studies in plain text.
** Relative Risk (95% Confidence Interval), except for Fairley (1999) and Lipfert et al. (2000), where insufficient
data were available to calculate confidence intervals so p-value is given in parentheses.
*** Min, Max 24-h PM indicator level shown in parentheses unless otherwise noted as (±S.D.), NR = not reported,
or R = range of values from min-max, no mean value reported.
References:

 ASchwartz et al. (1996a)
 BLaden et al. (2000)
 cFairley(1999)
 DGwynn et al. (2000)
 ELipfert et al. (2000a)
 FLippmann et al. (2000)
 GMar et al. (2000)
 "Smith et al. (2000)
 'Moolgavkar (2000a)
 JOstro(1995)
 KOstro et al. (2000)
 LSchwartz (2000a)
 MTsai  et al. (2000)
NBurnett et al. (2000)
°Burnett et al. (1998a)
pGoldberg et al. (2000)
QBurnettetal.(1994)
RBurnett et al. (1995)
sThurstonetal.(1992, 1994)
TLumley and Heagerty (1999)
"Burnett et al. (1997)
vDelfmo et al. (1997)
wDelfmo et al. (1998)
xStieb et al. (2000)
YBurnett et al. (1999)
zTolbert et al. (2000)
      AAMoolgavkar et al. (2000)
      BBMoolgavkar (2000b)
      ccSheppard et al. (1999)
      DDNorris et al. (1999)
      EEMoolgavkar (2000c)
      FFOstroetal.(1993)
      GGSchwartzetal.(1994)
      HHNeasetal.(1995)
      "Thurston et al. (1997)
      "Neasetal. (1996)
      KKSchwartz and Neas (2000)
      LLNaeher et al. (1999)
      ^Neasetal. (1999)
March 2001
            9-52
DRAFT-DO NOT QUOTE OR CITE

-------
   TABLE 9-4. EFFECT ESTIMATES PER VARIABLE INCREMENTS IN 24-HOUR
       CONCENTRATIONS OF COARSE-FRACTION PARTICLES (PM,0.2 5)
                   FROM U.S. AND CANADIAN STUDIES*
                                    RR(±CI)**per25-^g/m3
                           Reported PM
                         Levels Mean (Min,
Study Location
Indicator
Increase
Max)***
Acute Mortality
Six Cities:4
Portage, Wl
Topeka, KS
Boston, MA
St. Louis, MO
Kingston/Knoxville, TN
Steubenville, OH
Overall Six-City Results
Coachella Valley, CAB
Detroit, MIC
Philadelphia, PAD
Phoenix, AZE
Phoenix, AZF
Santa Clara County, CAG
Eight Canadian Cities"

PM,0.2S
PMI0.25
PM!0.2S
P"flO-2 5
PMI0.2S
PM10.2S
PM,0_25
PM10.25
PMI0.25
PM10.25
PM,o.25
PM25
PM10.25
PM10.2,

7.073 (0.970, 1.058)
0.968 (0.920, 1.015)
1.005 (0.985, 1.030)
1.005 (0.983, 1.028)
1.025 (0.985, 1.066)
1.061 (1.013, 1.111)
1.004(0.999, 1.010)
1.013(0.994, 1.032)
1.040(0.988,1.094)
1. 052 (p> 0.055)
1.030(0.995,1.066)
(>25 ^g/m3) 1.185 (1.069, 1.314)
(<25/j.g/m3) 1.020(1.005, 1.035)
1.02(p>0.05))
1.018(0.992, 1.044)

6.6 (±6.8)
14.5 (±12.2)
8.8 (±7.0)
11.9 (±8.5)
11.2 (±7.4)
16.1 (±13.0)
Median 9.0
17.9(0, 149)
13 (4, 50)
6.80 (-20.0, 28.3)
33.5(5, 187)
NR
1 1 (0, 45)
12.9 (max 99)
Cause-Specific Mortality
Total Cardiovascular:
Coachella Valley, CAB
Detroit, MIC
Philadelphia, PAD
(seven-county area)
Phoenix, AZE
Santa Clara County, CAG
Total Respiratory:
Coachella Valley, CAB
Detroit, MI°
Philadelphia, PAD
(seven-county area)
Santa Clara County, CAG

PM10-25
PM10.25
PM.o-25
PM10.25
PM,o.25

PM10.25
PM10.25
PM,0.25
PMIB.,,

1.026(1.006, 1.045)
1.078(1.000, 1.162)
1. 034 (p> 0.055)
1.064(1.014, 1.117)
1.03(p>0.05)

1.026(1.006, 1.045)
1.074(0.910, 1.269)
1. 030 (p> 0.055)
1.16(p>0.05)

17.9(0, 149)
13(4,50)
6.80 (-20.0, 28.3)
33.5(5,187)
1 1 (0, 45)

17.9(0,149)
13(4,50)
6.80 (-20.0, 28.3)
1 1 (0, 45)
March 2001
9-53
DRAFT-DO NOT QUOTE OR CITE

-------
   TABLE 9-4 (cont'd). EFFECT ESTIMATES PER VARIABLE INCREMENTS IN
  24-HOUR CONCENTRATIONS OF COARSE-FRACTION PARTICLES (PM1M5)
                  FROM U.S. AND CANADIAN STUDIES*
Study Location
Indicator
RR (±CI)** per 25-^g/m3
Increase
Reported PM
Levels Mean (Min,
Max)***
Increased Hospitalization
Total Respiratory:
Toronto, Canada1
Pneumonia:
Detroit, MIC
Respiratory infections:
Toronto, Canada1
COPD:
Atlanta, GAK
Detroit, MIC
Toronto, Canada1
Total Cardiovascular:
Atlanta, GAK
Toronto, Canada1
Ischemic Heart Disease:
Detroit, MIC
Toronto, Canada'
Dvsrhvthmias:
Detroit, MIC
Atlanta, GAK
Toronto, Canada1
Heart Failure:
Detroit, MIC
Toronto, Canada1
Stroke:
Detroit, MIC
Cerebrovascular:
Toronto, Canada1
Peripheral Circulation Diseases:
Toronto, Canada1

PM.o.25

PM,0.25

PM10.25

PM10.25
PMI0.25
PM10.25

PM10.25
PM10.25

PM10.25
PM10.25

PM10.25
PMI0.25
PM10.25

PM,0.25
PM10.25

PM10.25

PMI0.25

PM11L,,

1.125(1.052,1.20)

1.119(1.006, 1.244)

1.093(1.046,1.142)

0.770(0.493,1.202)
1.093(0.958, 1.247)
1.128(1.049,1.213)

1.176(0.954, 1.450)
1.205(1.082, 1.341)

1.105(1.027, 1.189
1.037(1.013, 1.062))

1.002(0.877, 1.144)
1.532(1.021,2.30)
1.051 (0.998, 1.108)

1.052(0.967, 1.144)
1.079(1.023,1.138)

1.049(0.953, 1.155)

"NEG" reported

1.056(1.003, 1.112)

11.6(1,56)

13(4,50)

12.2 (max 68)

9.39 (±4.52)
13(4,50)
12. 2 (max 68)

9.39 (±4.52)
11.6(1,56)

13(4,50)
12.2 (max 68)

13(4,50)
9.39 (±4.52)
12.2 (max 68)

13(4,50)
12. 2 (max 68)

13(4,50)

12.2 (max 68)

12.2 (max 68)
March 2001
9-54
DRAFT-DO NOT QUOTE OR CITE

-------
     TABLE 9-4 (cont'd). EFFECT ESTIMATES PER VARIABLE INCREMENTS IN
    24-HOUR CONCENTRATIONS OF COARSE-FRACTION PARTICLES (PM10.2 5)
                          FROM U.S. AND CANADIAN STUDIES*
Study Location Indicator
Asthma:
Seattle, WAL PM,0.25
Toronto, CanadaJ PM,0.2,
Increased Respiratory Symptoms
Six U.S. CitiesM PM10.25
(Lower Respiratory
Symptoms)
Six U.S. Cities" PM10.25
(Cough)
Southwest VirginiaN PM , 0.2 5
(Runny or Stuffy Nose)
Decreased Lung Function
Southwest Virginia0 PM , 0.2 5
Uniontown, PAM PM,0.25
(Reanalysis)
RR(±CI)**per25-/wg/m3
Increase
1.111(1.028,1.201)
1.111(1.058,1.166)
Odds Ratio (95% CI) per 25-Mg/m3
PM Increase
1.51 (0.94,4.87)
1.77(1.24,2.55)
2.62(1.16,5.87)
PEFR change (L/min) per 25-^g/m3
PM Increase
am PEFR 5.3 (2.6, 8.0)
pm PEFR +1.73 (5.67, -2.2)
Reported PM
Levels Mean (Min,
Max)***
16.2(6,29)
12.2 (max 68)

NR
NR
NR

27.07 (4.89, 69.07)
NR
 State College, PAM
 (Reanalysis)

 Philadelphia, PAP
PM,,
PM,
pm PEFR -0.28 (2.86, -3.45)
am PEFR-4.31 (-11.44,2.75)
    NR
9.5(IQR5.1)
* Studies highlighted in the 1996 CD are in italics; new studies in plain text.
** Relative Risk (95% Confidence Interval), except for Fairley (1999) and Lipfert et al. (2000), where insufficient
data were available to calculate confidence intervals so p-value is given in parentheses.
*** Min, Max 24-h PM indicator level shown in parentheses unless otherwise noted as (±S.D.), NR = not reported,
or R = range of values from min-max, no mean value reported.
References:

 ASchwartz et al., (1996a)
 BOstro et al. (2000)
 cLippmann et al (2000)
 DLipfert et al (2000)
 EMar et al. (2000)
 FSmith et al. (2000)
   °Fairley(1999)
   "Burnett et al. (2000)
   'Burnett et al. (1997)
   JBurnett et al. (1999)
   KTolbert et al. (2000)
                 LSheppardetal.(1999)
                 MSchwartz and Neas (2000)
                 NNaeheretal. (1999)
                 °Zhang et al. (2000)
                 •"Neasetal. (1999)
March 2001
               9-55
          DRAFT-DO NOT QUOTE OR CITE

-------
os
3.
O
O
ON
T1
H
6
o
2
o
H
O
G
O
H
W
O
*>
O
HH
H
CD
                               Percent excess death (total unless otherwise noted) per
                                   25 ug/m3 increase in PM2.s (•) or PM10.2.5 (o).
_
Klemm etal (2000) _
Harvard 6 Cities (recomputed)
Burnett et al (2000) _
8 Canadian Cities
Chock et al (2000)
Pittsburgh, PA
Klemm and Mason (2000)
Atlanta, GA ~
Lipfert et al (2000) _
Philadelphia, PA ~
Lippman et al (2000)
Detroit, Ml ~
Mar et al (2000) _
Phoenix, AZ
Fairley(1999) _
Santa Clara Co
Ostroetal (2000)
Coachella Valley, CA ~
Castillejos et al (2000)
Mexico City, Mexico
Cifuentes et al (2000)
Santiago, Chile
5-4-3-2-10 1 2 34 56 7 8 9 10 11 12 13 14 1
1 I I 1 I 1 1 1 1 1 I 1 1 1 1 1 I 1 I 1


1
	 1 	


^ , 	 .^_


^* }aye>75






1 «g 1 rtay ^


cardi - ubr
_— O— " mortal ity
	 • 	
	




Laq5dayMA> • "" " ^

	 "' Q 	 }Allyear
Laq2dayMA> ^ 	 }Wmtoi

	 1 	 V 	
1
Figure 9-7. Percent excess risks estimated per 25-fj.g/m3 increase in PM2 5 or PM10.2 5 from new studies evaluating both PM2 5
          and PM10_2 5 data for multiple years. All lags = 1 day, unless indicated otherwise.

-------
1      associations with one or more of several specific fine particle components of PM, including H+,
2      sulfate, nitrate, as well as COH; but their relative importance varied from city to city, likely
3      depending, in part, on their concentrations (e.g., no clear associations in those cities where H+
4      and sulfate levels were very low [i.e., circa nondetection limits]). Figure 9-8 depicts relatively
5      consistent estimates of total mortality excess risk resulting from a 5-yUg/m3 increase in sulfate,
6      possibly reflecting impacts of sulfate per se or perhaps sulfate serving as a surrogate for fine
7      particles in general. Sulfate effect size estimates generally fall in the range of 1 to 4% excess
8      total mortality per 5-^g/m3 increase for U.S. and Canadian cities.
9
                           Percent excess death (total mortality, unless otherwise noted)
                                           per 5 |jg/m3 increase in sulfate


Burnett etal. (1998a)
Toronto, Canada ~
Burnett et al. (2000)
8 Largest -
Canadian Cities
Fairley(1999)
Santa Clara, Co.
Gwynn et al. (2000)
Buffalo, NY
Klemm et al. (2000)
Atlanta, GA
Lipfert et al. (2000a)
Philadelphia, PA
Lippman et al. (2000)
Detroit, Ml

Tsai et al. (2000)
3 NJ Cities

-2 0 2 4 6 8 10
1 1 1 1 1 I

	











,



^ Elizabeth
       Figure 9-8.  Relative risks estimated per 5-^g/m3 increase in sulfate from U.S. and
                   Canadian studies in which both PM,S and PMin,* data were available.
      March 2001
9-57
DRAFT-DO NOT QUOTE OR CITE

-------
  1           A significant factor in some western cities is the occasional occurrence of high levels of
  2      windblown crustal particles that constitute the major part of the coarse PM fraction and a
  3      substantial fraction of intermodal fine particles (PM2 5.,).  The small-size tail of the windblown
  4      crustal particles extends into the PM2 5_, size range (intermodal), at times contributing
  5      significantly to PM2 5. Claiborn et al. (2000) report that in Spokane, WA, PM2 5 constitutes about
  6      30% of PM10 on dust event days, but 48% on days preceding the dust event. The intermodal
  7      fraction represents about 51 % of PM2 5 during windblown dust events, about 28% on preceding
  8      days.  However, PM[ in Spokane often shows little change during dust events, when coarse
  9      particles (presumably crustal particles) are transported into the region.  The lack of increased
10      mortality during periods of time with high wind speeds and presumably high crustal material
11      concentrations was shown by Schwartz et al. (1999) for Spokane, and by Pope et al. (1999a) for
12      three cities in the Wasatch front region of Utah. Other recent studies suggest that coarse particles
13      also may be associated with excess mortality as well as fine particles in certain U.S.  locations
14      e.g., in Phoenix, AZ (Smith et al., 2000; Clyde et al., 2000; Mar et al., 2000) the Coachella
15      Valley of California (Ostro et al., 2000), Mexico City  (Castillejos et al., 2000) or Santiago, Chile
16      (Cifuentes et al., 2000).  However, the coarse particle  association with mortality may not be
17      caused by the crustal components. An important advantage of using source profiles  for PM2 5 in
18      western cities is that it allows separation of crustal PM from accumulation-mode PM derived
19      from anthropogenic origins.
20           Several new studies highlighted in Chapter 6 conducted source-oriented evaluations of PM
21      components using factor analysis (see Table 9-5). The results of these studies  (Laden et al.,
22      2000; Mar et al., 2000; Tsai et al., 2000; Ozkaynak et  al., 1996) generally suggest that a number
23      of combustion-related source-types are associated with excess mortality risk, including: regional
24      sulfate; automobile emissions; coal combustion; oil  burning; and vegetative (biomass) burning.
25      In contrast, the crustal factor from fine particles was generally not positively associated with total
26      mortality, with Mar et al. (2000) reporting a negative association between the crustal component
27      of PM2 5 and cardiovascular mortality.
28           However, these source-oriented evaluation results are derived from relatively limited
29      underlying analytic bases and must be viewed with caution at this time. For example, whereas
30      Laden et al. (2000) had 6211 days of every-other-day data from the Harvard Six City Study of
31      eastern/midwest U.S. cities, they had only elements in PM2 5 analyzed by X-ray fluorescence

        March 2001                                9-58         DRAFT-DO NOT QUOTE OR CITE

-------
                  TABLE 9-5.  SUMMARY OF SOURCE-ORIENTED EVALUATIONS OF
                     PARTICULATE MATTER COMPONENTS IN RECENT STUDIES
          Author, City
Source Types Identified (or Suggested) and
Associated Tracers
     Source Types Associated with Mortality.
     Comments.
          Laden et. al., (2000)
          Harvard Six Cities
          1979-1988
          Mar et al. (2000).
          Phoenix, AZ
          1995-1997
          Ozkaynak et al.
          (1996).
          Toronto, Canada.

          Tsai et al. (2000).
          Newark, Elizabeth,
          and Camden, NJ.
          1981-1983.
Soil and crustal material: Si

Motor vehicle emissions: Pb

Coal combustion: Se

Fuel oil combustion: V

Salt: Cl

Note: the trace elements are from PM2 5 samples
PM25 (from DFPSS) trace elements.
Motor vehicle emissions and resuspended road dust:
Mn, Fe, Zn, Pb, OC, EC, CO, and NO2

Soil: Al, Si, and Fe
Vegetative burning: OC and Ks (soil-corrected
potassium)

Local SO, sources: SO2

Regional sulfate: S
PM,I>.25 (from dichot) trace elements:
Soil: Al, Si, K, Ca, Mn, Fe, Sr, and Rb
A source of coarse fraction metals: Zn, Pb, and Cu

A marine influence:  Cl

Motor vehicle emissions: CO, COH, and NO2



Motor vehicle emissions: Pb and CO

Geological (Soil): Mn and Fe

Oil burning: V and Ni

Industrial:  Zn, Cu, and Cd (separately)
Sulfate/secondarv aerosol:  Sulfate
Note:  The trace elements are from PMI5 samples.
     The strongest increase in daily mortality
     was associated with the mobile source
     factor.  The coal combustion factor was
     positively associated with mortality in all
     metropolitan areas, with the exception of
     Topeka. The crustal factor from the fine
     particles was not associated with
     mortality.

     Coal and mobile sources account for the
     majority of fine particles in each city.

     PM^ < factors results:  Soil factor and local
     SO2 factor were negatively associated with
     total mortality.  Regional sulfate was
     positively associated with total mortality
     on the same day, but negatively associated
     on the lag 3 day. Motor vehicle factor,
     vegetative burning factor, and regional
     sulfate factor were significantly positively
     associated with cardiovascular mortality.
     Factors from dichot PM10_25 trace elements
     were not analyzed for their associations
     with mortality because of the small sample
     size (every-third-day samples from June
     1996).

     Motor vehicle factor was a significant
     predictor for total, cancer, cardiovascular,
     respiratory, and pneumonia deaths.

     Oil burning, industry, secondary aerosol,
     and motor vehicle factors were associated
     with mortality.
1       (XRF) spectroscopy (no organic PM or gases) and they used Pb as a tracer to identify a motor

2       vehicle source, Se to identify a coal combustion source, and Si as a tracer for soil.  The "motor

3       vehicle" and "coal combustion" sources were statistically significant for total mortality as well as
        March 2001
                            9-59
DRAFT-DO NOT QUOTE OR CITE

-------
 1      mortality resulting from ischemic heart disease and respiratory diseases (COPD plus pneumonia).
 2      The crustal component had a negative association with total mortality.
 3           The Mar et al. (2000) study had 3 years of pollutant data for Phoenix, AZ.  In addition to
 4      elements determined by XRF, they had pollutant gases (CO, NO2, SO2, and O3) and total,
 5      organic, and elemental carbon. They were able to identify five sources. Motor vehicles (plus
 6      resuspended road dust), vegetative burning, and regional sulfate all had statistically significant
 7      associations with cardiovascular mortality, but soil (indexed by Si and Al, as crustal markers)
 8      had a statistically significant negative association.
 9           Tsai et al. (2000) had only 156 days of data and used measurements of CO, sulfate, and
10      some elements; and they did not have Si, Ca, Al, or Mg as soil tracers nor Se as a tracer of coal
11      combustion, although much of the sulfate probably came from coal combustion. They had three
12      fractions of extractable organic matter, but these did not appear to be useful in determining
13      source factors.  Nevertheless, they were still able to identify motor vehicles, oil burning, and
14      sulfate as statistically significant  (p > 0.05) factors for both total daily deaths and combined
15      cardiovascular and respiratory daily deaths in at least one or another of the three New Jersey
16      cities studied (Newark, Camden, and Elizabeth). Also, an industrial source containing Zn and Cd
17      was statistically significant for total deaths in Newark; and an industrial source containing Cd
18      was marginally statistically significant for cardiorespiratory disease in Elizabeth.
19           Ozkaynak et al.  (1996) had only TSP, coefficient of haze (COH), and gases; however, they
20      reported that a factor with COH,  CO, and NO2 (considered to be representative of motor vehicle
21      emissions) was associated with mortality in Toronto, Canada.
22           None of these studies had measurements of nitrate or semivolatile organic  compounds nor
23      did they use the newest, and most effective, techniques for source apportionment. For example,
24      using positive matrix  factorization, Ramadan et al. (2000) were able to determine eight factors
25      using the same data set as Mar et al. (2000). In spite of these deficiencies, all four studies were
26      able to associate one or more types of morality with motor vehicles, several with coal
27      combustion, and three with sulfate.
28           Factor analyses  also were described briefly in a report by Lippmann et al. (2000). In that
29      study, neither sulfate nor acid aerosols were related significantly to morbidity or mortality, but
30      the concentrations were extremely low (with about 70% of the acid measurements below
31      detection limit).

        March 2001                                9-60       DRAFT-DO NOT QUOTE OR CITE

-------
  1           It is difficult to compare these source-related assessments. They are based on different
  2      regions of the country over different periods of time when the sources of particles and other
  3      urban air pollutants were changing greatly. Furthermore, each of these studies constructed
  4      factors based on city-specific data. Thus, the factors in each study are based on the
  5      idiosyncrasies of the specific data set for each city in the study, so the factors may indeed
  6      represent different  sources in different locations. Nevertheless, although somewhat limited at
  7      this time, the new factor analysis results appear to implicate ambient PM derived from fossil fuel
  8      (oil, coal) combustion and vegetative burning, and secondarily formed sulfates as important
  9      contributors to observed mortality effects, but not crustal particles.
 10           In summary, there is evidence that exposure to particles from several different source
 11      categories and, of different composition and size may have independent associations with health
 12      outcomes. The excess risks from different types of combustion sources (coal, oil, gasoline,
 13      wood, and vegetation) may vary from place to place and from time to time, so that substantial
 14      intra-regional and inter-regional heterogeneity would be expected.  Likewise, although earlier
 15      evaluations in the 1996 PM AQCD seemed to indicate coarse particles and intermodal particles
 16      of crustal composition as not likely being associated with adverse health effects, there are now
 17      some reasonably credible studies suggesting that coarse particles (although not necessarily those
 18      of crustal composition) may sometimes be as associated with excess mortality in at least some
 19      locations.
 20
 21      9.6.2.2 Updated Epidemiologic Findings for Long-Term Particulate Matter Exposure
 22             Effects on Mortality
 23           The 1996 PM AQCD indicated that past epidemiologic studies of chronic PM exposures
 24      collectively indicate increases in mortality to be associated with long-term exposure to airborne
 25      particles of ambient origins (see appendix Table 9A-3). The PM effect size estimates for total
26      mortality from these studies also indicated that a substantial portion of these deaths reflected
27      cumulative PM impacts above and beyond those exerted by acute exposure events. Table 9-6
28      shows long-term exposure effects estimates (RR values) per variable increments in ambient PM
29      indicators in U.S. and Canadian cities, including results from newer analyses since the 1996 PM
30      AQCD.


        March 2001                                9-61        DRAFT-DO NOT QUOTE OR CITE

-------
  TABLE 9-6. EFFECT ESTIMATES PER INCREMENTSA IN LONG-TERM MEAN
  LEVELS OF FINE AND INHALABLE PARTICLE INDICATORS FROM U.S. AND
                        CANADIAN STUDIES
Type of Health
Effect and Location
Increased Total Mortality in
Six City6


ACSStudyc
(151 U.S. SMSA)

Six City Reanalysis0

ACS Study Reanalysis0

Southern CahforniaE



Indicator
Adults
PM,5/IO(20^g/m3)
PM25(20/ug/m3)
SOr.OSvg/m')
PM25(20/ug/m3)
S0:(15^g/m3)
PMl5/10(20//g/m3)
PM25(20,ug/m3)
PM15/lo(20//g/m3)
(SSI)
PM25(20^g/m3)
PM10(50Mg/m3)
PMIO (cutoff =
30 days/year
PM10 (50Mg/m3)
PM,0 (cutoff =
30 days/year
>100,wg/m3)
Increased Bronchitis in Children
Six Cif/
Six City0
24 City"
24 City"
24 City"
24 City"
Southern California'
12 Southern California
communitiesj
(all children)
1 2 Southern California
communitieslc
(children with asthma)
PMlm(50^/m3)
TSP (100/ug/m3)
H+ (100 nmol/m3)
S0:(l5vg/m3)
PM2l(25/ug/m3)
PMlo(50/ug/m3)
S0=(15^g/m3)
PM,0(25Mg/m3)
Acid vapor (1 .7 ppb)
PM10(19,ug/m3)
Acid vapor (1.8 ppb)
Change in Health Indicator per
Increment in PMa
Relative Risk (95% CI)
1.18(1.06-1.32)
1.28(1.09-1.51)
1.46(1.16-2.16)
1.14(1 07-1.21)
1.10(1.06-1.16)
1.19(1.06-1.34)
1.28(1.09-1.51)
1.02(0.99-1.04)
1.14(1.08-1.21)
1.242 (0.955- 1.6 16) (males)
1.082 (1.008- 1.1 62) (males)
0.879 (0.713-1.085) (females)
0.958 (0.899-1.021) (females)
Odds Ratio (95% CI)
3.26(1.13, 10.28)
280(1.17, 7.03)
2.65 (1.22, 5.74)
3.02 (1.28, 7.03)
1.97 (0.85,4.51)
3.29(0.81, 13.62)
1.39(0.99, 1.92)
0.94(0.74, 1.19)
1.16(0.79, 1.68)
1.4(1.1, 1.8)
1.4(0.9,2.3)
1.1 (0.7, 1.6)
Range of City
PM Levels *
Means (//g/m3)

18-47
11-30
5-13
9-34
4-24
18.2-46.5
11.0-29.6
58.7(34-101)
9.0-33.4
51 (±17)

51 (±17)


20-59
39-114
6.2-41.0
18.1-67.3
9 1-173
22.0-28.6
—
28.0-84.9
0.9-3.2 ppb
13.0-70.7
6.7-31.5
1.0-5.0 ppb
March 2001
9-62
DRAFT-DO NOT QUOTE OR CITE

-------
  TABLE 9-6 (cont'd). EFFECT ESTIMATES PER INCREMENTSA IN LONG-TERM
 MEAN LEVELS OF FINE AND INHALABLE PARTICLE INDICATORS FROM U.S.
                      AND CANADIAN STUDIES
Type of Health
Effect and Location
Increased Cough in Children
1 2 Southern California
communitiesj
(all children)
12 Southern California
communitiesK
(children with asthma)
Indicator

PM10 (25 ^g/m3)
Acid vapor (1 .7 ppb)
PM10(19^g/m3)
PM^dS^g/m3)
Acid vapor (1.8 ppb)
Change in Health Indicator per
Increment in PMa
Odds Ratio (95% CI)
1.06(0.93,1.21)
1.13(0.92,1.38)
1.1(0.0.8, 1.7)
1.3(0.7,2.4)
1.4(0.9,2.1)
Range of City
PM Levels *
Means C"g/m3)

28.0-84.9
0.9-3.2 ppb
13.0-70.7
6.7-31.5
1.0-5.0 ppb
Increased Obstruction in Adults
Southern California1"
Decreased Lung Function in
Six Cit/
Six City0
24 City"
24 City"
24 City"
24 City*1
12 Southern California
communities'^
(all children)
12 Southern California
communities1"1
(all children)
12 Southern California
communities0
(4th grade cohort)
12 Southern California
communities0
(4th grade cohort)
PM]0 (cutoff of
42 days/year
>100Aig/m3)
Children
PM15/lo(50^g/m3)
TSP(100jug/m3)
H+ (52 nmoles/m3)
PM2l(15^g/m3)
S0'4(7^g/m3)
PM10(17 Mg/m3)
PM10(25Mg/m3)
Acid vapor (1.7 ppb)
PM10 (25 ^g/m3)
Acid vapor (1 .7 ppb)
PM,0(51.5,ug/m3)
PM25(25.9/ug/m3)
PM10.25(25.6^g/m3)
Acid vapor (4.3 ppb)
PM,0(51.5Aig/m3)
PM25(25.9Mg/m3)
PM10.25(25.6//g/m3)
Acid vapor (4.3 ppb)
1.09(0.92, 1.30)

NS Changes
NS Changes
-3.45% (-4.87, -2.01) FVC
-3.21% (-4.98, -1. 41) FVC
-3.06% (-4.50, -1.60) FVC
-2.42% (-4.30, -.0.51) FVC
-24.9 (-47.2, -2.6) FVC
-24.9 (-65.08, 15.28) FVC
-32.0 (-58.9, -5.1) MMEF
-7.9 (-60.43, 44.63) MMEF
-0.58 (-1.14, -0.02) FVC growth
-0.47 (-0.94, 0.01) FVC growth
-0.57 (-1.20, 0.06) FVC growth
-0.57 (-1.06, -0.07) FVC growth
-1.32 (-2.43, -0.20) MMEF growth
-1.03 (-1.95, -0.09) MMEF growth
-1.37 (-2.57, -0.15) MMEF growth
-1.03 (-2.09, 0.05) MMEF growth
NR

20-59
39-114
6.2-41.0
18.1-67.3
9.1-17.3
22.0-28.6
28.0-84.9
0.9-3.2 ppb
28.0-84.9
0.9-3.2 ppb
NR
NR
March 2001
9-63
DRAFT-DO NOT QUOTE OR CITE

-------
         TABLE 9-6 (cont'd).  EFFECT ESTIMATES PER INCREMENTSA IN LONG-TERM
         MEAN LEVELS OF FINE AND INHALABLE PARTICLE INDICATORS FROM U.S.
                                       AND CANADIAN STUDIES
        Type of Health
        Effect and Location
Indicator
                                              Change in Health Indicator per
                                                    Increment in PMa
                                                     Range of City
                                                      PM Levels *
                                                     Means (/ug/m3)
        Decreased Lung Function in Adults
        Southern Californiap
        (% predicted FEV,,
        females)
PMIO (cutoff of
54.2 days/year
>100,ug/m3)
Southern California1"        PM,0 (cutoff of
(% predicted FEV,, males)   54.2 days/year
                        >100,ug/m3)

Southern California1"        PM]0 (cutoff of
(% predicted FEV,, males   54.2 days/year
whose parents had asthma,   >100 ,ug/m3)
bronchitis, emphysema)

Southern Californiap        SO^ (1.6 /ug/m3)
(% predicted FEV,,
females)
Southern California1"        SO^ (1.6 Afg/m3)
(% predicted FEV,, males)	
                         +0.9 % (-0.8, 2.5) FEV,
                                                         +0.3 % (-2.2, 2.8) FEV,
                                                        -7.2% (-11.5,-2.7) FEV,
                                                             Not reported
                                                         -1.5% (-2.9,-0.1) FEV,
                                                                             52.7(21.3,80.6)
                                                    54.1 (20.0, 80.6)
                                                    54.1 (20.0,80.6)
                                                     7.4(2.7, 10.1)
                                                     7.3(2.0, 10.1)
        *Range of mean PM levels given unless, as indicated, studies reported overall study mean (min, max), or mean
         (±SD); NR=not reported.
        AResults calculated using PM increment between the high and low levels in cities, or other PM increments given
         in parentheses; NS Changes = No significant changes.
       References:
       "Dockery et al. (1993)
       cPopeetal.(1995)
       DKrewski et al. (2000)
       EAbbeyetal. (1999)
       FDockery et al. (1989a)
       GWare etal.( 1986)
       "Dockeryetal. (1996)
       1 Abbey et al. (1995a,b,c)
                         JPetersetal. (1999b)
                         KMcConnell et al. (1999)
                         LBerglund et al. (1999)
                         MRaizenneetal.(1996)
                         NPetersetal. (1999a)
                         °Gauderman et al. (2000)
                         pAbbeyetal. (1998)
1           One of the most important advances since the 1996 PM AQCD is the substantial
2      verification and extension of the findings of the Six City prospective cohort study (Dockery
3      et al., 1993) and the cohort study relating American Cancer Society (ACS) health data to
4      fine-particle data from 50 cities and sulfate data from 151 cities (Pope et al., 1995).  The
5      reanalyses, sponsored by the Health Effects Institute (HEI), included a  data audit, replication of
       March 2001
                   9-64
                                                        DRAFT-DO NOT QUOTE OR CITE

-------
  1     the original investigators' findings, and additional analyses to explore the sensitivity of the
  2     original findings to other model specifications. The investigators of the HEI Reanalysis Project
  3     (Krewski et al., 2000) first performed a data audit, using random samples to verify the accuracy
  4     of the data sets used in the original Six City analyses, including death certificate data, air
  5     pollution data, and socioeconomic data.  In general, the air pollution data were reproducible and
  6     correlated highly with the original aerometric data in Pope et al. (1995).
  7          The reanalyses substantially verified the findings of the original investigators, with PM2 5 or
  8     sulfate relative risk (RR) estimates for total mortality and for cardiopulmonary mortality differing
  9     at most by ±0.02 (±2% excess risk) from the least polluted to the most polluted cities in the
 10     study.  A larger difference was noted for the PM2 5 lung cancer relative risk in the Six Cities
 11     study, 1.37 originally and 1.43 in the reanalysis, neither estimate being statistically significant.
 12     The sensitivity analyses for the Six Cities study found generally similar results with other
 13     individual covariates included. The time-dependent covariate model for total mortality (taking
 14     into account higher postexposures in early years of the study and changes over time to the last
 15     years of the study) had a substantially lower RR than the model without time-dependent
 16     covariates.  Educational level made a large difference, with individuals having less than a high
 17     school education at much greater risk for mortality than those with any postsecondary education.
 18          Among the ecological covariates, sulfates adjusted for artifact had little effect on the risk
 19     estimates for total mortality compared to that without adjustment, but, in the ACS study, the filter
 20     adjustment actually increased the relative risk for all causes and cardiopulmonary mortality,
 21     while substantially reducing the estimated sulfate effect on lung cancer. Inclusion of SO2 as an
 22     additional ecological covariate greatly reduced the estimated PM2 5 and sulfate effects in the ACS
 23     study, whereas a spatial model including SO2 effects caused only a modest reduction of the
 24     estimated PM2 5 and sulfate effects. However, the SO2 effects were reduced greatly when sulfates
 25     were included in the model.  Sulfur dioxide and sulfates often are highly correlated, because of
 26     the formation of secondary sulfates.
 27          Many model selection issues in the prospective cohort studies are analogous to those in the
 28     time series analyses.  One issue of particular concern is whether the exposure indices used in the
29     analyses adequately characterize the exposure of the participants in the study during the months
30     or years preceding death. This question is particularly conspicuous in regard to the Pope et al.
31      (1995) study,  in which PM2 5 and sulfate data were collected in the 1979 to 1982 period from the

        March 2001                               9-65        DRAFT-DO NOT QUOTE OR CITE

-------
 1      EPA AIRS database and the Inhalable Particle Network, largely preceding the collection of the
 2      ACS cohort data by only a few years, and so possibly not adequately reflecting exposure to
 3      presumably much higher PM concentrations occurring long before the cohort was recruited, nor
 4      exposure to presumably lower concentrations during the study. This issue was raised in the 1996
 5      PM AQCD.  However, the Six Cities Study did have air pollution data and repeated survey data
 6      over time, with PM2 5 and sulfate data measured every other day and sometimes daily, and so the
 7      new investigators were able to use the information about time-dependent cumulative PM
 8      concentrations during the course of the study.  Changes in smoking status and body mass index
 9      over the 10 to 12 years of the study had little effect on risk estimates, but taking into account the
10      decrease in particle concentrations from the earlier years to the later years reduced the effect size
11      estimate substantially, although it remained statistically significant. Nevertheless, overall, the
12      reanalyses of the ACS and Harvard Six-Cities studies (Krewski et al., 2000) "replicated the
13      original results, and tested those results against alternative risk models and analytic approaches
14      without substantively altering the original findings of an association between indicators of
15      particulate matter air pollution and mortality."
16           The shape of the relationship of concentration to mortality also was explored.  Preliminary
17      findings suggest some possible nonlineariry, but further study is needed.  Among the most
18      important new findings of the study are spatial relationships between mortality and air pollution,
19      discussed later below.
20           With regard to the role of various PM constituents in the PM-mortality association, past
21      cross-sectional studies generally have found that the fine particle component, as indicated either
22      by PM2 5 or sulfates, was the PM constituent most consistently associated with chronic PM
23      exposure-mortality. Although the relative measurement errors of the various PM constituents
24      must be further evaluated as a possible source of bias in these estimate comparisons, the Harvard
25      Six-Cities study and the latest reported AHSMOG prospective semi-individual study results
26      (Abbey, et al., 1999a,b,c; McConnell et al., 2000) studies are both indicative of the fine mass
27      components of PM likely being associated more strongly with the mortality effects of PM than
28      coarse PM components; and the ACS study, which only evaluated fine particle indicators, further
29      substantiates ambient fine particle effects.
30           Several other new studies report epidemiologic evidence indicating that: (a) PM exposure
31      early in pregnancy (during the first month) may be associated with slowed intrauterine growth

        March 2001                               9-66       DRAFT-DO NOT QUOTE OR CITE

-------
  1      leading to low birth weight events (Dejmek et al., 1999); and (b) early postnatal PM exposures
  2      may lead to increased infant mortality (Woodruff et al.,  1997; Boback and Leon, 1999; Loomis
  3      et al., 1999; Lipfert et al., 2000b).
  4           Recent investigations of the public health implications of effect estimates for long-term PM
  5      exposures also were reviewed in Chapter 6. Life table calculations by Brunekreef (1997) found
  6      that relatively small differences in long-term exposure to airborne PM of ambient origin can have
  7      substantial effects on life expectancy.  For example, a calculation for the 1969 to 71 life table for
  8      U.S. white males indicated that a chronic exposure increase of 10 yUg/m3 PM was associated with
  9      a reduction of 1.31 years for the entire population's life expectancy at age 25.  The new evidence
 10      noted above of infant mortality associations with PM exposure suggests that life shortening in the
 11      entire population from long-term PM exposure could well be significantly larger than estimated
 12      by Brunekreef (1997).
 13
 14      9.6.2.3 Relationships of Ambient Participate Matter Concentrations to Morbidity
 15             Outcomes
 16           New epidemiology studies add greatly to the overall database relating morbidity outcomes
 17      to ambient PM levels.  These include much additional evidence for cardiovascular and
 18      respiratory diseases being related to ambient PM. The newer epidemiology studies expand the
 19      evidence on cardiovascular (CVD) disease and are discussed first below, followed by discussion
 20      of respiratory disease effects with particular emphasis on newly enhanced evidence for
 21      PM-asthma relationships.
 22
 23      9.6.2.3.1  Cardiovascular Effects of Ambient Paniculate Matter Exposures
 24           About 75% of all U.S. deaths occur in persons at least 65 years old, and, of these, nearly
 25      40% are for cardiac causes (nearly 45%, if deaths from cerebrovascular causes are also included).
 26      Thus, if ambient PM exposure indeed produces increased total mortality in the elderly, it would
27      seem possible that cardiovascular (CVD) deaths may be  involved.
28
29      Cardiovascular Hospital Admissions. Just two studies were available for review in the 1996
30      PM AQCD that provided data on acute cardiovascular morbidity outcomes (Schwartz and
31      Morris, 1995; Burnett et al., 1995). Both studies were of ecologic time series design using

        March 2001                               9-67        DRAFT-DO NOT QUOTE OR CITE

-------
  1      standard statistical methods.  Analyzing 4 years of data on the > 65-year-old Medicare population
  2      in Detroit, MI, Schwartz and Morris (1995) reported significant associations between ischemic
  3      heart disease admissions and PM,0, controlling for environmental covariates.  Based on an
  4      analysis of admissions data from 168 hospitals throughout Ontario, Canada, Burnett and
  5      colleagues (1995) reported significant associations between particle sulfate concentrations, as
  6      well as other air pollutants, and daily cardiovascular admissions.  The relative risk because of
  7      sulfate particles was slightly larger for respiratory than for cardiovascular hospital admissions.
  8      The 1996 PM AQCD concluded on the basis of these studies that, "There is a suggestion of a
  9      relationship to heart disease, but the results are based on only two studies and the estimated
 10      effects are smaller than those for other endpoints." The PM AQCD went on to state that acute
 11      impacts on CVD admissions had been demonstrated for elderly populations (i.e., >65), but that
 12      insufficient data existed to assess relative impacts on younger populations.
 13           Although the literature still remains relatively sparse, an important new body of data now
 14      exists that both extends the available quantitative information on relationships between ambient
 15      PM pollution and hospital CVD admissions, and that, more intriguingly, illuminates some of the
 16      physiological changes that may occur on the mechanistic pathway leading from PM exposure to
 17      adverse cardiac outcomes. Figure 9-9 depicts excess risk estimates derived from 10 studies of
 18      acute PM10 exposure effects on CVD admissions in U.S. cities.  Although new studies depicted
 19      in Figure 9-9 have reported generally consistent associations between daily hospitalizations for
20      cardiovascular disease and measures of PM, the data not only implicate PM, but also CO and
21      NO2 as well, possibly because of covarying of PM and these other gaseous pollutants derived
22      from common  emission sources (e.g., motor vehicles). Taken as a whole, this body of evidence
23      suggests that PM is likely an important risk factor for cardiovascular hospitalizations in the
24      United States.
25           For example, in the recently published NMMAPS 14-city analysis of daily CVD hospital
26      admissions in persons 65 and older in relation to PM10 (Samet et al., 2000a,b). The mean risk
27      estimate (for average 0-1 day lag) was a 8.5% increase in CVD admissions per 50 /ug/m3 PM,0
28      (95% CI: 1.0 to 33.0%). No relationship was observed between city-specific risk estimates and
29      measures of socioeconomic status,  including percent living in poverty, percent non-white, and
30      percent with college educations. In another study, remarkably consistent PM10 associations with
31      cardiovascular admissions were observed across eight U.S. metropolitan areas, with a 25 /ug/m3

        March 2001                                9-68        DRAFT-DO NOT QUOTE OR CITE

-------
            Sametetal (2000a,b) -
                14US Cities

                Schwartz (1999) -
                 8 US Counties

              Moolgavkar (2000c) -
                Maricopa, AZ

              Moolgavkar (2000c) -
                  LA.CA

              Moolgavkar (2000c) -
                Cook County

                Linnetal (2000) -
                   LA.CA

                Schwartz (1997) -
                  Tucson.AZ

              Tolbertetal (2000) -
                  Atlanta

        Morris and Naumova (1998) -
               Chicago

            Lippmann et al.(2000) -
Total CVD
                                  i
                                 -15
       Period 1 (MRS Data)	
                         CHF
                                 i   »    i
                                Period 2 (Suqersite Data)
         -10        -50         5        10
      Reconstructed Excess Risk Percentage
               50 ug/m3 Increase in PM,0
       Figure 9-9.   Acute cardiovascular hospitalizations and PM exposure excess risk estimates
                    derived from selected U.S. PM,0 studies. CVD = cardiovascular disease and
                    CHF = congestive heart failure.
1      increase in PM10 associated with between 1.8 and 4.2 percent increases in admissions (Schwartz,

2      1999).  Also, in a study of Los Angeles data from 1992-1995, PM10, CO, and NO2 were all

3      significantly associated with increased cardiovascular admission in single-pollutant models

4      among persons 30 and older (Linn et al., 2000). Moolgavkar (2000c) analyzed PM10, CO, NO2,

5      O3, and SO2 in relation to daily total cardiovascular (CVD) and total cerebrovascular admissions

6      for persons 65 and older from three urban counties (Cook, IL; Los Angeles, CA; Maricopa, AZ),

7      and found that, in univariate regressions, PM10 (and PM2 5 in LA) was associated with CVD
      March 2001
               9-69
DRAFT-DO NOT QUOTE OR CITE

-------
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
admissions in Cook and LA counties but not in Maricopa county. On the other hand, in
two-pollutant models in Cook and LA counties, the PM risk estimates diminished and/or were
rendered nonsignificant.
     The recent NMMAPS study of PM10 concentrations and hospital admissions by persons
65 and older in 14 U.S. cities provides particularly important findings of positive and significant
associations, even when concentrations are below 50 /ug/m3 (Samet et al., 2000a,b).  As noted in
Table 9-7, this study indicates PM10 effects similar to other cities, but with narrower confidence
bands, because of its greater power derived by combining multiple cities in the same analysis.
This allows significant associations to be identified, despite the fact that many of the cities
considered have relatively small populations and that each of the 14 cities had mean PM10 below
50
           TABLE 9-7. PERCENT INCREASE IN HOSPITAL ADMISSIONS PER 10-Aig/m3
                          INCREASE IN 24-HOUR PM,,, IN 14 U.S. CITIES
CVD

Constrained Lag Models
One-day mean3
Previous-day mean
Two-day meanb
PM10 <50 Mg.m3
(2-day mean)b
Quadratic distributed lag
Increase
(Fixed Effect
1.07
0.68
1.17
1.47
1.18
(95%
CI)
COPD
Increase
(95% CI)
Pneumonia
Increase
(95% CI)
Estimates)
(0.93,
1.22)
(0.54,0.81)
(1.01,
(1.18,
(0.96,
1.33)
1.76)
1.39)
1
1
1
2
2
.44
.46
.98
.63
.49
(1.00,
(1.03,
(1.49,
(1.71,
(1.78,
1.89)
1.88)
2.47)
3.55)
3.20)
1.57
1.
1.
2.
1
.31
.98
.84
.68
(1-27,
(1.03,
(1.65,
(2.21,
(1.25,
1.87)
1.58)
2.31)
3.48)
2.11)
Unconstrained Distributed Lag
Fixed effects estimate
Random effects estimate
1.19
1.07
(0.97,
(0.67,
1.41)
1.46)
2
2
.45
.88
(1.75,
(0.19,
3.17)
5.64)
1
2
.90
.07
(1.46,
(0.94,
2.34)
3.22)
        aLag.
        bMean of lag 0 and lag 1.
        Source: Samet et al., 2000a,b.
       March 2001
                                         9-70
DRAFT-DO NOT QUOTE OR CITE

-------
  1     Physiologic Measures of Cardiac Function. Several very recent studies by independent groups
  2     of investigators have also reported longitudinal associations between ambient PM concentrations
  3     and physiologic measures of cardiovascular function. These studies measure outcomes and most
  4     covariates at the individual level, making it possible to draw conclusions regarding individual
  5     risks, as well as to explore mechanistic hypotheses. For example, several studies recently have
  6     reported temporal associations between PM exposures and various electrocardiogram (ECG)
  7     measures of heart beat or rhythm in panels of elderly subjects.  Reduced HR variability is a
  8     predictor of increased cardiovascular morbidity and mortality risks.  Three independent studies
  9     reported decreases in HR variability associated with PM in elderly cohorts, although r-MSSD
 10     (one measure of high-frequency HR variability) showed elevations with PM in one study.
 11     Differences in methods used and results obtained across the studies argue for caution in drawing
 12     any strong conclusions yet regarding PM effects from them, especially in light of the complex
 13     intercorrelations that exist among measures of cardiac physiology, meteorology, and air pollution
 14     (Dockery et al., 1999). Still, the new heart rhythm results, in general, comport well with other
 15     findings of cardiovascular mortality and morbidity endpoints being associated with ambient PM.
 16     Chapter 5 discusses available exposure studies of elderly subjects with CVD, such as the Sarnat
 17     et al. (2000) Baltimore study. Less active groups tend to have lower exposure to nonambient PM
 18     because of reduced personal activity. However, Williams et al. (2000a,b,c) report a very high
 19     pooled correlation coefficient between PM2 5 personal exposure and outdoor concentrations.
 20     These exposure studies tend to enhance the plausibility of panel study findings of impacts on HR
 21     variability being caused by exposure to ambient-generated PM.
 22
 23     Changes in Blood Characteristics. Additional epidemiologic findings (Peters et al., 1997a)
 24     also provide new evidence for ambient PM exposure effects on blood characteristics (e.g.,
 25     increased c-reactive protein in blood) thought to be associated with increased risk of serious
 26     cardiac outcomes (e.g., heart attacks).
 27
28     • Key Conclusions Regarding PM-CVD Morbidity.  Overall, the newly available studies of
29      PM-CVD relationships  appear to support the following conclusions regarding several key
30      issues:
31

        March 2001                               9-71        DRAFT-DO NOT QUOTE OR CITE

-------
  1      • Temporal Patterns of Response. The evidence from recent time series studies of CVD
  2       admissions suggests rather strongly that PM effects are likely maximal at lag 0, with some
  3       carryover to lag 1.
  4
  5      • Physical and Chemical Attributes Related to Participate Matter Health Effects.  The
  6       characterization of ambient PM attributes associated with acute CVD is incomplete.
  7       Insufficient data exist from the time series CVD hospital admissions literature or from the
  8       emerging individual-level studies to provide clear guidance as to which PM attributes, defined
  9       either on the basis of size or composition, determine potency.  The epidemiologic studies
10       published to date have been constrained by the limited availability of multiple PM metrics.
11       Where multiple PM metrics exist, they often are of differential quality because of differences in
12       numbers of monitoring sites and in monitoring frequency.  Until more extensive and consistent
13       data become available for epidemiologic research, the question of PM size and composition, as
14       they relate to acute CVD impacts, will remain open.
15
16      • Susceptible Subpopulations.  Because they lack data on individual subject characteristics,
17       ecologic time series studies provide only limited information on susceptibility factors based on
18       stratified analyses.  The relative impact of PM on cardiovascular (and respiratory) admissions
19       reported in ecologic time series studies is generally somewhat higher than those reported for
20       total admissions. This provides some limited support for the hypothesis that acute effects of
21       PM operate via cardiopulmonary pathways or that persons  with preexisting cardiopulmonary
22       disease have greater susceptibility to PM, or both. Although there is some data from the
23       ecologic time series studies showing larger relative impacts of PM on cardiovascular
24       admissions in adults 65 and over as compared with younger populations, the differences are
25       neither striking nor consistent.  Some individual-level studies of cardiophysiologic function
26       suggest that elderly persons with preexisting cardiopulmonary disease are susceptible to subtle
27       changes in heart rate variability (HRV) in association with PM exposures. However, because
28       younger and healthier populations have not yet been assessed,  it is not possible to say at present
29       whether the elderly have clearly increased susceptibility compared to other groups, as indexed
30       by cardiac pathophysiological indices such  as HRV.
31
        March 2001                               9-72         DRAFT-DO NOT QUOTE OR CITE

-------
  1      • Role of Other Environmental Factors. The ecologic time series morbidity studies published
  2       since 1996 generally have controlled adequately for weather influences. Thus, it is unlikely that
  3       residual confounding by weather accounts for the PM associations observed. With one possible
  4       exception (Pope et al., 1999b), the roles of meteorological factors have not been analyzed
  5       extensively as yet in the individual-level studies of cardiac physiologic function. Thus, the
  6       possibility of confounding in such studies as yet cannot be discounted totally or readily.
  7       Co-pollutants have been analyzed rather extensively in many of the recent time series studies of
  8       hospital admissions and PM. In some studies, PM clearly carries an independent association
  9       after controlling for gaseous co-pollutants. In others, the "PM effects" are reduced markedly
10       once co-pollutants are added to the model. Among the gaseous criteria pollutants, CO has
11       emerged as the most consistently associated with cardiovascular (CVD) hospitalizations. The
12       CO effects are generally robust in the multi-pollutant model, sometimes as much so as PM
13       effects. However, the typically low levels of ambient CO concentrations in most such studies
14       and minimal expected impacts on carboxyhemoglobin levels and consequent associated
15       hypoxic effects thought to underlie CO CVD effects complicate interpretation of the CO
16       findings and argue for the possibility that CO may be serving as a general surrogate for
17       combustion products (e.g., PM) in the ambient pollution mix. See the recently completed EPA
18       CO criteria document (U.S. Environmental Protection Agency, 2000a).
19
20      9.6.2.3.2 Respiratory Effects of Ambient Particulate Matter Exposures
21           The number of studies examining hospitalization and emergency department visits for
22      respiratory-related causes and other respiratory morbidity endpoints has increased markedly since
23      the 1996 PM AQCD.  In addition to evaluating statistical relationships for PMIO, quite a few new
24      studies also evaluated other PM metrics. Those providing estimates of increased risk in U.S. and
25      Canadian cities for respiratory-related morbidity measures (hospitalizations, respiratory
26      symptoms, etc.) in relation to 24-h increments in ambient fine particles (PM2 5) or coarse fraction
27      (PM10_2 5) of inhalable thoracic particles are included in Tables 9-3 and 9-4, respectively.
28
29      Respiratory-Related Hospital Admission/Visits. PM hospital admissions/ visit studies that
30      evaluated excess risks in relation to PMIO measures are still quite informative.  Maximum excess
31      risk estimates for PM10 associations with respiratory-related hospital admissions and visits in

        March 2001                               9-73        DRAFT-DO NOT QUOTE OR CITE

-------
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
U.S. cities are shown in Figure 9-10. Nearly all the studies showed positive, statistically
significant relationships between ambient PM10 and increased risk for respiratory-related doctors'
visits and hospital admissions. Overall, the results substantiate well ambient PM10 impacts on
respiratory-related hospital admissions/visits.  The excess risk estimates fall most consistently in
the range of 5 to 25.0% per 50 /^g/m3 PM10 increment, with those for asthma hospital admissions
and doctor's visits being higher than for COPD and pneumonia hospitalization. Other, more
limited, new evidence (not depicted in Figure 9-10) shows excess risk estimates for overall
respiratory-related or COPD hospital admissions falling in the range of 5 to 15.0% per 24-h
25 Mg/m3 increment in PM2 5 or PM10.2 5.  Larger estimates are found for asthma admissions or
physician visits, ranging up to ca. 40 to 50% for children <18 yr old in one study.
              Tolbert et al. (2000) Atlanta -
               Norris et al (2000) Seattle -
              Norns et al (2000) Spokane -
               Norris etal (1999) Seattle -
         Choudhury et al  (1997) Anchorage -
         Nauenberg and Basu (1999) LA.CA -
            Sheppard etal (1999) Seattle -
            Zonobetti et al. (2000) Chicago -
         Sametelal (2000a,b) 14 US Cities -
             Moolgavkar (2000b) Phoenix -
              Moolgavkar (2000b) LA.CA -
             Moolgavkar (2000b) Chicago -
          Moolgavkar et af (2000a) King C -
          Moolgavkar etal (1997) Minn-SP -
            Moolgavkar et al (1997) Birm  -
              Chen et al (2000) Reno.NV -
            Zanobetti et al (2000) Chicago -
         Sametetal. (2000a,b) 14 US Cities -
                                  -25









\


Asthma Visits

, ,



Asthma Hospital Admissions
, |
'
w
* '
.n COPD Hospital Admissions
l-»-H
»H

w Pneumonia Hospital Admissions
                                            25      50      75       100
                                                 Excess Risk, %
                     125
150
       Figure 9-10.  Maximum excess risk in selected studies of U.S. cities relating PM,0 estimate
                     of exposure (50 /wg/m3) to respiratory-related hospital admissions and visits.
        March 2001
                                            9-74
DRAFT-DO NOT QUOTE OR CITE

-------
  1           Of particular note in Figure 9-10, are the large effect size estimates now being reported for
  2      asthma hospitalizations and visits.  Very importantly, these hospital admission/visit studies and
  3      other new studies on respiratory symptoms and lung function decrements in asthmatics are
  4      emerging as possibly indicative of ambient PM likely being a notable contributor to exacerbation
  5      of asthma. Additional evidence for PM-asthma effects is also emerging from panel studies of
  6      lung function and respiratory symptoms, and these are discussed in more detail below.
  7           New panel studies of lung function and respiratory symptoms in asthmatic subjects have
  8      been conducted by more than 10 research teams in various locations world-wide. As a group, the
  9      studies examine health outcome effects that  are similar, such as pulmonary peak flow rate
 10      (PEFR); and the studies typically characterize the clinical-symptomatic aspects in a sample of
 11      mild to moderate asthmatics (mainly children aged 5 to 16 yrs) observed in their natural setting.
 12      Their asthma typically is being treated to keep them symptom free (with "normal" pulmonary
 13      function rates, and activity levels) and to prevent recurrent exacerbations of asthma. Severity of
 14      their asthma  is characterized by symptom, pulmonary function, and medication use and would be
 15      classified to include mild intermittent to mild persistent asthma suffers (National Institutes of
 16      Health, 1997).  As a group, they may thusly differ from asthmatics examined in studies of
 17      hospitalization or doctor visits for acute asthmatic episodes, who may have more severe asthma.
 18           Most studies reported ambient PM10 results, but PM2 5 was examined in two studies.  Other
 19      ambient PM measures (BS and SO4) also were used. For these studies, mean PM10 levels range
 20      from a low of 13 Aig/m3 in Finland to a high of 167 /ug/m3 in Mexico City. The Mexico City
 21      level is over three times more than each of the other levels and is unique compared to the others.
 22      Related 95% CI for these means or  ranges show 1-day maximums above 100 Aig/m3 in four
 23      studies, with two of these above 150 jUg/m3.  Hence, these studies mainly evaluated different PM
 24      metrics indexing PM concentrations in the range found in U.S. cities (see Chapter 3). All the
 25      studies controlled for temperature, and several controlled for relative humidity.
 26           Many panel studies are analyzed using  a design that takes advantage of the repeated
 27      measures on the same subject.  Study subject number (N) varied from 12 to 164, with most
28      having N >50; and all gathered adequate subject-day data to provide sufficient power for their
29      analyses.  Linear models often are used for lung function and logistic models for dichotomous
30      outcomes.  Meteorological variables are used as covariates; and medication use is also sometimes
31      evaluated as a dependent variable or treated as an important potential confounder. However,

        March 2001                               9-75        DRAFT-DO NOT QUOTE OR CITE

-------
 1     perhaps the most critical choice in the model is selection of the lag for the pollution variable.
 2     Presenting lag periods with only the strongest associations introduces potential bias, because the
 3     biological basis for lag structure may be related to effect. No biological bases for pertinent lag
 4     periods are known, but some hypotheses can be proposed. Acute asthmatic reactions can occur
 5     4 to 6 h after exposure and, thus, 0-day lag may be more appropriate than 1-day lags for that
 6     acute reaction. Lag 1 may be more relevant for morning measurement of asthma outcome from
 7     PM exposure the day before, and longer term lags (i.e., 2 to 5 days) may represent the outcome of
 8     a more prolonged inflammatory mechanism; but too little information is now available to
 9     predetermine appropriate lag(s).
10           Chapter 7 noted that people with asthma tend to have greater TB deposition then do healthy
11     people, but this data was not derived from the younger age group studied in most asthma panel
12     studies. The Peters et al. (1997b) study is unique for two reasons: (1) they studied the size
13     distribution of the particles in the range 0.01 to 2.5 ptm and (2) examined the number of particles.
14     They reported that asthma-related health effects of 5-day means of the number of ultrafme
15     particles were larger than those of the mass of the fine particles. In contrast, Pekkanen et al.
16     (1997) also examined a range of PM sizes, but PMIO was more consistently associated with PEF.
17     Delfino et al. (1998) is unique in that they report larger effects for 1- and 8-h maximum PM10
18     than for the 24-h mean.
19           The results for the asthma panels of the peak flow analysis consistently show small
20     decrements for both PM10 and PM2 5. The effects using 2- to 5-day lags averaged about the same
21     as did the 0 to 1 day lags. Stronger relationships often were found with ozone.  The analyses
22     were not able to clearly separate co-pollutant effects.  The effects on respiratory symptoms in
23     asthmatics also tended to be positive.  Most studies showed increases in cough, phlegm,
24     difficulty breathing, and bronchodilator use. The only endpoint more strongly related to longer
25     lag times was bronchodilator use, which was observed in three studies. The peak flow
26     decrements and respiratory symptoms are indicators for asthma episodes.
27           For PMIO, nearly all of the point estimates showed decreases, but most were not statistically
28     significant, as shown in Figure 9-11 as an example of PEF outcomes. Lag 1 may be more
29     relevant for morning measurement of asthma outcome from the previous day. The figure
30     presents studies that provided this data.  The results were consistent for both AM and PM peak
31     flow analyses. Similar results were found for the PM2 5 studies, although there were fewer

       March 2001                               9-76       DRAFT-DO NOT QUOTE OR CITE

-------
            Romieuetal. (1996)
                 (Mexico)
         Pekkannen et al. (1997)
               (Finland)
             Gielenetal. (1997)
               (Netherlands)
            Romieuetal. (1997) -
                 (Mexico)
                              -10                -505
                                        Change in Pulmonary Function, L/min

       Figure 9-11. Selected acute pulmonary function change studies of asthmatic children.
                    Effect of 50 Aig/m3 PM10 on morning Peak flow lagged 1 day.
 1     studies. Several studies included PM2 5 and PM10 independently in their analyses of peak flow.
 2     Of these, Gold et al. (1999), Naeher et al. (1999), Tiittanen et al. (1999), Pekkanen et al. (1997),
 3     and Romieu et al. (1996) all found similar results for PM2 5 and PM10.  The study of Peters et al.
 4     (1997b) found slightly larger effects for PM2 5. The study of Schwartz and Neas (2000) found
 5     larger effects for PM2 5 than for the coarse mode. Naeher et al. (1999) found that H+ was related
 6     significantly to a decrease in morning PEF. Thus, there is no evidence here for a stronger effect
 7     of PM2 5 when compared to PM10.  Also,  of studies that provided analyses that attempted to
 8     separate out effects of PM10 and PM2 5 from other pollutants, Gold et al. (1999) studied possible
 9     interactive effects of PM2 5 and ozone on PEF; they found independent effects of the two
10     pollutants, but the joint effect was slightly less than the sum of the independent effects.
11          The effects on respiratory symptoms in asthmatics also tended to be positive, although
12     much less consistent than the lung function effects. Most studies showed increases in cough,
13     phlegm, difficulty breathing, and bronchodilator use (although generally not statistically
       March 2001
9-77
DRAFT-DO NOT QUOTE OR CITE

-------
 1
 2
 3
 4
 5
 6
 7
significant), as shown in Figure 9-12 for cough as an example.  Three studies included both PMIO
and PM2 5 in their analyses. The studies of Peters et al. (1997c) and Tiittanen et al. (1999) found
comparable effects for the two measures. Only the Romieu et al. (1996) found slightly larger
effects for PM2 5. These studies also give no good evidence for a stronger effect of PM2 5 when
compared to PM,0.
           Vedaletal. (1998)
               (Canada)
         Romieu etal. (1997)
              (Mexico)
          Gielen etal. (1997)
            (Netherlands)
         Peters etal. (1997c)
          (Czech Republic)
                                     M
                            01234567
                                              Odds Ratios for Cough

       Figure 9-12.  Odds ratios for cough for a 50-/^g/m3 increase in PM,0 for selected asthmatic
                    children studies, with lag 0 with 95% CI.
 8          The results of PM10 peak flow analyses for nonasthmatic populations were inconsistent.
 9     Fewer studies reported results in the same manner as the asthmatic studies.  Many of the point
10     estimates showed increases rather than decreases.  PM2 5 studies found similar results. The
11     effects on respiratory symptoms in nonasthmatics were similar to those in asthmatics: most
12     studies showed that PM10 increases cough, phlegm, and difficulty breathing, but these increases
13     were generally not statistically significant. Schwartz and Neas (2000) found that PM10.2 5 coarse
       March 2001
                                        9-78
DRAFT-DO NOT QUOTE OR CITE

-------
   1      particles were significantly related to cough.  Tiittanen et al. (1999) found that 1-day lag of
  2      PMI0.2 5 was related to morning PEF, but not evening PEF. Neas et al. (1999) found no coarse
  3      mode effects of PEF in non-asthmatic subjects.
  4
  5      9.6.2.3.3  Long-Term Paniculate Matter Exposure Effects on Lung Function and Respiratory
  6               Symptoms
  1           In the 1996 PM AQCD, the available respiratory disease studies were limited in terms of
  8      conclusions that could be drawn. At that time, three studies based on a similar type of
  9      questionnaire administered at three different times as part of the Harvard Six-City and 24-City
 10      Studies provided data on the relationship of chronic respiratory disease to PM. All three studies
 11      suggest a chronic PM exposure effect on respiratory disease. The analysis of chronic cough,
 12      chest illness,  and bronchitis tended to be significantly positive for the earlier surveys described
 13      by Ware et al. (1986) and Dockery et al. (1989).  Using a design similar to the earlier one,
 14      Dockery et al. (1996) expanded the analyses to include 24 communities in the United States and
 15      Canada. Bronchitis was found to be higher (odds ratio = 1.66) in the community with highest
 16      exposure of strongly acidic particles when compared with the least polluted community. Fine
 17      PM sulfate was also associated with higher reporting of bronchitis (OR = 1.65, 95% CI  1.12,
 18      2.42).
 19          The studies by Ware et al. (1986), Dockery et al. (1989), and Neas et al. (1994) all had
 20      good monitoring data and well-conducted standardized pulmonary function testing over many
 21     years, but showed no effect on children of PM pollution indexed by TSP,  PM15, PM2 5, or
 22     sulfates. In contrast, the latest 24-city analyses reported by Raizenne et al. (1996) found
 23     significant associations of effects on FEV, or FVC in U.S. and Canadian children with both
 24     acidic particles and other PM indicators.  Overall, the available studies provided limited evidence
 25     suggestive of pulmonary lung function decrements being associated with chronic exposure to PM
 26     indexed by various measures (TSP, PMIO, sulfates, etc.).
 27          A number of studies have been published since 1996, which evaluate the effects of
 28     long-term PM exposure on lung function and respiratory symptoms, as presented in Chapter 6.
29      The methodology in the long-term studies varies much more than the methodology in the short-
30      term studies. Some studies reported highly significant results (related to PM), whereas others
31      reported no significant results. Of particular note are several studies reporting associations

        March 2001                                9-79        DRAFT-DO NOT QUOTE OR CITE

-------
  1      between long-term PM exposures (indexed by various measures) or changes in such exposures
  2      over time and chronic bronchitis rates, consistent with bronchitis results from the Dockery et al.
  3      (1996) study noted above.
  4           Unfortunately, the cross-sectional studies often are potentially confounded, in part, by
  5      unexplained differences in geographic regions; and it is difficult to separate out results consistent
  6      with a PM gradient from any other pollutants or factors having the same gradient.  The studies
  7      that looked for a time trend also are confounded by other conditions that changed over time.  The
  8      most credible cross-sectional study remains that described by Dockery et al. (1996) and Raizenne
  9      et al. (1996).  Whereas most studies include two to six communities, this study included 24
10      communities and is considered to provide the most credible estimates of long-term PM exposure
11      effects on lung function and respiratory symptoms.
12
13      9.6.2.4 Methodological Issues
14           Chapter 6 discussed several still important methodological issues related to assessment of
15      the overall PM epidemiologic database.  These include, especially, issues related to model
16      specifications and consequent adequacy of control for potentially confounding of PM effects by
17      co-pollutants, evaluations of possible source relationships to pollutant effects that may be useful
18      in sorting out better effects attributable to PM versus other  co-pollutants or both, and other issues
19      such as lag  structure. Key points are discussed concisely below.
20
21      9.6.2.4.2  Time Series Studies: Confounding by Co-Pollutants in Individual Cities
22           The co-pollutant issue was discussed at length in the  1996 document and still remains an
23      important issue. It must be recognized that there are large differences in concentrations of
24      measured gaseous co-pollutants (and presumably unmeasured pollutants as well) in different
25      parts of the United States, as well as the rest of the world; and the concentrations are often
26      correlated with concentrations of PM and its components because of commonality in source
27      emissions, wind speed and direction, atmospheric processes, and other human activities and
28      meteorological conditions. Large sources in the United States include motor vehicle emissions
29      (gasoline combustion, diesel fuel combustion, evaporation, particles generated by tire wear, etc.),
30      coal combustion, fuel oil combustion,  industrial processes,  residential wood burning, solid waste
31      combustion, and so on.  Thus, one might reasonably expect some large correlations among PM

        March 2001                               9-80        DRAFT-DO NOT QUOTE OR CITE

-------
   1      and co-pollutants, but possibly with substantial differences in relation by season in different
  2      cities or regions. Statistical theory suggests that PM and co-pollutant effect size estimates will be
  3      highly unstable and often insignificant in multi-pollutant models when collinearity exists. Many
  4      recent studies demonstrate this effect, for both hospital admissions (Moolgavkar, 2000b) and
  5      mortality (Moolgavkar, 2000a; Chock et al., 2000). Because the problem seems largely insoluble
  6      in studies in single cities, the new multi-city studies (Samet et al., 2000a,b; Schwartz, 1999;
  7      Schwartz and Zanobetti,  2000) have provided important new insights. See discussions of
  8      NMMAPS analysis in Chapter 6 and below for discussion of issues related to control for co-
  9      pollutant effects. Overall, although such issues may warrant further evaluation, it now appears
 10      unlikely that such confounding accounts for the vast array of effects attributed to  ambient PM
 11      based on the rapidly expanding PM epidemiology database.
 12          Numerous new studies have reported associations not only between PM, but also gaseous
 13      pollutants (O3, SO2, NO2, and CO), and mortality. In many of these studies, simultaneous
 14      inclusion of one or more  gaseous pollutants in regression models did not markedly affect PM
 15      effect size estimates, as was generally the case in  the NMMAPS analyses for 90 cities (see
 16      Figure 9-13).  On the other hand, some studies reporting positive and statistically significant
 17      effects for gaseous copollutants (e.g., O3, NO2, SO2, CO) found varying degrees of robustness of
 18      their effects estimates or those of PM in multipollutant models.  Thus, although it is likely that
 19      there are independent health effects of PM and gaseous pollutants, there is not yet sufficient
 20      evidence by which to confidently separate out fully the relative contributions of PM versus those
 21       of other gaseous pollutants or by which to quantitate modifications of PM effects by other co-
 22     pollutants, including possible synergistic interactions that may vary seasonally or from location
 23      to location.  Overall, it appears, however, that ambient PM and O3 can be most clearly separated
 24      out as likely having independent effects, their concentrations often not being highly correlated.
 25      More difficulty is encountered, at times, in sorting out whether NO2, CO, or SO2 are exerting
 26      independent effects in cities where they tend to be highly correlated with ambient PM
 27      concentrations, possibly because of derivation of important PM constituents from  the same
 28      source (e.g., NO2, CO, PM from mobile sources) or a gaseous pollutant (e.g.,  SO2) serving as a
29      precursor for a significant PM  component (e.g., sulfate).
        March 2001                                9-81        DRAFT-DO NOT QUOTE OR CITE

-------
                     PM10
                     PM10
                     PM10 + O3 + N02
                     PM10 + O3+S02
                     PM10
                    0.0           0.2
                  % Change in Mortality per 10 gg/m3 Increase in PM
                               10
      Figure 9-13.  Marginal posterior distributions for effect of PM10 on total mortality at lag 1,
                   with and without control for other pollutants, for the 90 cities.  The numbers
                   in the upper right legend are the posterior probabilities that the overall
                   effects are greater than 0.
      Source: Samet et al. (2000a,b).
1     9.6.2.4.3 Time Series Studies: Model Selection for Lags, Moving Averages, and Distributed
2              Lags
3          A number of different approaches have been used to evaluate the temporal dependence of
4     mortality or morbidity on time-lagged PM concentrations, including unweighted moving
5     averages of PM concentrations over one or more days, general weighted moving averages, and
6     polynomial distributed moving averages. Unless there are nearly complete daily data, each
7     different lag will be using a different set of mortality data corresponding to spaced PM
8     measurement; for example, for lag 0 with every-sixth-day PM measurements, the mortality data
      March 2001
9-82
DRAFT-DO NOT QUOTE OR CITE

-------
  1     are on the same day as the PM data, for lag 1 the mortality data are on the next day after the PM
  2     data, and so on. Although this effect is likely to be small, it should nonetheless be kept in mind.
  3          The issue of dealing with lag structure, which may not necessarily be the same for all cities
  4     or for all regions, can be illustrated by NMMAPS findings. As shown in Table 9-8, the rank
  5     ordering of effects by lag days differs somewhat among NMMAPS regions.  The combined data
  6     set suggests that lag 1 provides the best fit, but with some regional differences.  This raises the
  7     question as to whether a single lag model should be assumed to characterize a diverse set of
  8     regional findings. Because the particle constituents, co-pollutants, susceptible subpopulations,
  9     and meteorological covariates are likely to differ substantially from one region to another, the
 10     timing of the largest mortality effects also may be presumed to differ in at least some cases. This
 11     undoubtedly contributes to the variance of the estimated effects.
 12
 13
          TABLE 9-8. PERCENT INCREASE IN MORTALITY PER 10 yug/m3 PM10 IN SEVEN
        	 U.S. REGIONS (from Figure 23 in NMMAPS II)       	
         Region                                Rank Order of Effects by Lags
         Northwest                             lag 0 < lag 1  = lag 2
         Southwest                             lag 0 < lag 1  < lag 2
         Southern California                     lag 0 < lag 1, lag 1 > lag 2, lag 0 < lag 2
         Upper Midwest                         lag 0 > lag 1, lag 0 > lag 2, lag 1 < lag 2
         Industrial Midwest                      lag 0 < lag 1, lag 1 > lag 2
         Northeast                              lag 0 < lag 1, lag 1 » lag 2
         Southeast                              lag 0 « lag 1, lag 1  > lag 2
         Combined   	lag 0 < lag 1, lag 1 > lag 2
14          The distributed lag models used in the NMMAPS II morbidity studies are a noteworthy
15     methodological advance. The fitted distributed lag models showed significant heterogeneity
16     across cities for COPD and pneumonia, however (see Table 15 therein), again raising the
       March 2001                              9-83        DRAFT-DO NOT QUOTE OR CITE

-------
  1      question of how heterogeneous effects can best be combined so as not to obscure potentially real
  2      city-specific or region-specific differences.
  3           Only three cities with nearly complete daily PM10 data were used to evaluate more general
  4      multi-day lag models (Chicago, Minneapolis/St. Paul, Pittsburgh), and these show somewhat
  5      different patterns of effect, with lag 0 < lag 1 and lag 1 » lag 2 for Chicago, lag 0 = lag 1 > lag 2
  6      for Minneapolis, and lag 0 < lag 1 = lag 2 for Pittsburgh. The 7-day distributed lag model is
  7      significant for Pittsburgh, but less so in the other cities. The remaining data are limited
  8      intrinsically in what they can reveal about temporal structure.
  9
 10      9.6.2.4.4 Time Series Studies: Model Selection for Concentration-Response Functions
 11           Given the number of analyses that needed to be performed, it is not surprising that most of
 12      the NMMAPS studies focused on linear concentration-response models.  More recent studies
 13      (Daniels et al., 2000) for the 20 largest U.S. cities have found posterior mean effects of 2 to 2.7%
 14      excess risk of total daily mortality per 50 /ug/m3 24-h PM10 at lags 0, 1, 0+1 days; 2.4 to 3.5%
 15      excess risk of cardiovascular and respiratory mortality; and 1.2 to 1.7% for other causes of
 16      mortality.  The posterior 95% credible regions are all significantly greater than 0.  However, the
 17      threshold models gave distinctly different estimates of 95% credible regions for the threshold for
 18      total mortality (15 /ug/m3 at lag  1, range 10 to 20), cardiovascular and respiratory mortality
 19      (15 //g/m3 at lag 0+1, range 0 to 20), and other causes of mortality (65 /ug/m3 at lag 0+1,  range
20      50 to 75 /^g/ni3).
21           Another problem is that the shape of the relationship between mortality and PM10 may
22      depend, to some extent, on the associations of PM10 with gaseous co-pollutants. The association
23      is not necessarily linear, and is indeed likely to have both seasonal and secular components that
24      depend on the city location.  Thus, further elaborations of these models may be desirable.
25
26
27      9.6.2.4.5 Effects of Exposure Error in Daily Time Series Epidemiology
28           There has been considerable controversy over how to deal with the nonambient component
29      of personal exposure. Recent biostatistical analyses of exposure error have indicated that the
30      nonambient component will not bias the statistically calculated risk in community time-series
31      epidemiology, provided that the nonambient component of personal exposure is independent of

        March 2001                               9-84        DRAFT-DO NOT QUOTE OR CITE

-------
   1      the ambient concentration. Consideration of the random nature of nonambient sources and recent
   2      studies, in which estimates of a, ambient-generated PM divided by ambient PM concentrations,
   3      have been used to estimate separately the ambient-generated and nonambient components of
  4      personal exposure, support the assumption that the nonambient exposure is independent of the
   5      ambient concentration.  Therefore, it is reasonable to conclude that community time series
  6      epidemiology describes statistical associations between health effects and exposure to ambient-
  7      generated PM, but does not provide any information on possible health effects resulting from
  8      exposure to nonambient PM (e.g., indoor-generated PM).
  9           From the point of view of exposure error, it is also significant to note that,  although
 10      ambient concentrations of a number of gaseous pollutants (O3, NO2, SO2) often are found to be
 11      highly correlated with various PM parameters, personal exposures to these gases are not
 12      correlated highly with personal exposure to PM indicators.  The correlations of the ambient
 13      concentrations of these gases also are not correlated highly with the personal exposure to these
 14      gases. Therefore, when significant statistical associations are found between these gases and
 15      health effects, it could be that these gases may, at times, be serving as surrogates for PM rather
 16      than being causal themselves. Pertinent information on CO has not been reported.
 17           The attenuation factor, a, is a useful variable.  For relatively constant a, the risk because of
 18      a personal exposure to 10 //g/m3 of ambient PM is equal I/a times the risk from a concentration
 19      of 10 yWg/m3 of ambient PM, where a varies from a low of 0.1 to 0.2 to a maximum of 1.0.  (The
 20      health risk  for an interquartile change in ambient concentration of PM is the same as that for an
 21      interquartile change in exposure to ambient PM). Differences in a among cities,  reflecting
 22      differences in air-exchange rates (e.g., because of variation in seasonal temperatures and in extent
 23      of use of air conditioners) and differences in indoor/outdoor time ratios, may, in part, account for
 24     any differences in risk estimates based on statical associations between ambient concentrations
 25     and health effects for different cities or regions.  If a were 0.3 in city A, but 0.6 in city B, and the
 26     risks for an increase  in personal exposure of 10 //g/m3 were identical, then a regression of health
 27     effects on ambient concentrations would yield a health risk for city B that would be twice that
 28      obtained for city A.
29           A number of exposure analysts have discussed the PM exposure paradox (i.e., that
30      epidemiology yields  statistically significant associations between ambient concentrations and
31      health effects even though there is a near zero correlation between ambient concentrations and

        March 2001                                9-85         DRAFT-DO NOT QUOTE OR CITE

-------
  1      personal exposure in many studies). Several explanations have been advanced to resolve this
  2      paradox.  First, personal exposure contains both an ambient-generated and a nonambient
  3      component.  Community time series epidemiology yields information only on the ambient-
  4      generated component of exposure.  Therefore, the appropriate correlation to investigate is the
  5      correlation between ambient concentration and personal exposure to ambient-generated PM, not
  6      between ambient concentrations and total personal exposure (i.e., the sum of ambient-generated
  7      and nonambient PM).  Second, biostatistical analysis of exposure error indicates that if the risk
  8      function is linear in the PM indicator, the average of the sum of the individual risks (risk function
  9      times individual exposure) may be replaced by the risk function times the community average
10      exposure. Thus, the appropriate correlation (of ambient concentrations and ambient-generated
11      exposure) is not the pooled correlation of different days and different people but the correlation
12      between the daily ambient concentrations and the community average daily personal exposure to
13      ambient-generated PM. Because the nonambient component is not a function  of the ambient
14      concentration, its average will tend to be similar each day. Therefore, the correlation coefficient
15      will depend on a but not on the nonambient exposure.  These types of correlation yield high
16      correlation coefficients.
17           A few studies have conducted simulation analyses of effects of measurement errors on the
18      estimated PM mortality effects. These studies suggest that ambient PM excess risk effects are
19      more likely underestimated than overestimated, and that spurious PM effects (i.e., qualitative
20      bias such as change in the sign of the coefficient) because  of transferring of effects from other
21      covariates require extreme conditions and are therefore very unlikely. The error because the
22      difference between the average personal exposure and the  ambient concentration is likely the
23      major source of bias in the estimated relative risk. One study also suggested that apparent linear
24      exposure-response curves are unlikely to be artifacts of measurement error.
25           In conclusion, for time-series epidemiology, ambient concentration is a useful surrogate for
26      personal exposure to ambient-generated PM, although the  risk per unit ambient PM
27      concentration is biased low by the factor a compared to the risk per unit exposure to ambient-
28      generated PM.  Epidemiologic studies of statistical associations between long-term effects and
29      long term ambient concentrations compare health outcome rates across cities with different
30      ambient concentrations. Ordinarily, PM exposure measurement errors are not  expected to
31      influence the interpretation of findings from either the community time-series or long-term

        March 2001                               9-86        DRAFT-DO NOT QUOTE OR CITE

-------
  1      epidemiologic studies that have used ambient concentration data if they include sufficient
  2      adjustments for seasonality and key personal and geographic confounders. When individual level
  3      health outcomes are measured in small cohorts, to reduce exposure misclassification errors, it is
  4      essential that better real-time exposure monitoring techniques be used and that further speciation
  5      of indoor-generated, ambient, and personal PM mass be accomplished.  This should enable
  6      measurement (or estimation) of both ambient and nonambient components of personal exposure
  7      and evaluation of the extent to which personal exposure to ambient-generated PM, personal
  8      exposure to nonambient PM, or total personal exposure (to ambient-generated plus nonambient
  9      PM) contribute to observed health effects.
 10
 11       9.6.3  Coherence of Reported Epidemiologic Findings
 12           Interrelationships Between Health Endpoints.  Considerable coherence exists across
 13       newly available epidemiologic study findings.  For example, it was earlier noted that effects
 14       estimates for total (nonaccidental) mortality generally fall in the range of 2.5 to 5.0% excess
 15       deaths per 50 /ug/m3 24-h PM10 increment. These estimates comport well with those found for
 16       cause-specific cardiovascular- and respiratory-related mortality.  Furthermore, larger effect sizes
 17       for cardiovascular (in the range of 3 to 6% per 50 /^g/m3 24-h PM10 increment) and respiratory (in
 18      the range of 5 to 25% per 50 Aig/m3 24-h PM,0) hospital admissions and visits are found, as
 19      would be expected versus those for PM10-related mortality. Also, several independent panel
 20      studies, evaluating temporal associations between PM exposures and measures of heart beat
 21      rhythm in elderly subjects, provide generally consistent indications of decreased heart rate (HR)
 22      variability being associated with ambient PM exposure (decreased HR variability being an
 23      indicator of increased risk for serious cardiovascular outcomes, e.g., heart attacks). Other studies
 24      point toward changes in blood characteristics (e.g., increased C-reactive protein levels) related to
 25      increased risk of ischemic heart disease as also being associated with ambient PM exposures.
 26
 27           Spatial Interrelationships. Both the NMMAPS and Cohort Reanalyses studies had a
 28      sufficiently  large number of cities to allow considerable resolution of regional PM effects within
29      the "lower 48" states, but this approach was taken much farther in the Cohort Reanalysis studies
30      than in NMMAPS.  There were 88 cities with PM10 effect size estimates in NMMAPS; 50 cities
31      with PM2 5 and 151 cities with sulfates in Pope et al. (1995) and in the reanalyses using the
        March 2001                               9-87        DRAFT-DO NOT QUOTE OR CITE

-------
  1      original data; and, in the additional analyses by the cohort study reanalysis team, 63 cities with
  2      PM25 data and 144 cities with sulfate data. The relatively large number of data points allowed
  3      estimation of surfaces for elevated long-term concentrations of PM2 5, sulfates, and SO2 with
  4      resolution on a scale of a few tens to hundreds of kilometers.  Information drawn from the maps
  5      presented in Figures 16-21 in Krewski et al. (2000) is summarized below.
  6           The patterns are similar, but not identical.  In particular, the modeled PM2 5 surface
  7      (Krewski, Figure 18) has peak levels in the industrial midwest, including the Chicago and
  8      Cleveland areas, the upper Ohio River Valley, and around Birmingham, AL. Lower, but
  9      elevated, PM2 5 is found almost everywhere else east of the Mississippi, as well as in southern
10      California.  This is rather similar to the modeled sulfate surface (Krewski, Figure 16), with the
11      absence of a peak in Birmingham and an emerging sulfate peak in Atlanta. The only region with
12      elevated SO2 concentrations is the Cleveland-Pittsburgh area. A preliminary evaluation is that
13      secondary sulfates in particles derived from local SO2 is more likely to be important in the
14      industrial midwest, south from the Chicago-Gary region and along the upper Ohio River region.
15      This intriguing pattern may be related to the combustion of high-sulfur fuels in the subject areas.
16           The overlay of mortality and air pollution is also of interest.  The spatial overlay of long-
17      term PM2 5 and mortality (Krewksi, Figure 21) is highest for the upper Ohio River region, but
18      also includes a significant association over most of the industrial midwest from Illinois to the
19      eastern noncoastal parts of North Carolina, Virginia, Pennsylvania, and New York.  This is
20      reflected,  in diminished form, by the sulfates map (Krewski, Figure 19) where the peak sulfate-
21      mortality associations occur somewhat east of the peak PM2 5-mortality associations. The SO2
22      map (Krewski, Figure 20) shows peak associations similar to, but slightly east of, the peak
23      sulfate associations.  This suggests that, although SO2 may be an important precursor of sulfates
24      in this region, there may be other considerations (e.g., metals) in the association between PM2 5
25      and long-term mortality, embracing a wide area of the midwest and northeast (especially
26      noncoastal areas).
27           It should be noticed that, although a variety of spatial modeling approaches were discussed
28      in the NMMAPS methodology report (NMMAPS Part I, pp. 66-71), the primary spatial analyses
29      in the 90-city study (NMMAPS, Part II) were based on a simpler seven-region breakdown of the
30      contiguous 48 states. The 20-city results reported for the spatial model in NMMAPS I show a
31      much smaller posterior probability of a PM10 excess risk of short-term mortality, with a spatial

        March 2001                              9-88        DRAFT-DO NOT QUOTE OR CITE

-------
   1      posterior probability versus a nonspatial probability of a PM10 effect of 0.89 versus 0.98 at lag 0,
   2      of 0.92 versus 0.99 at lag 1, and of 0.85 versus 0.97 at lag 2. The evidence that PM10 is
   3      associated with an excess short-term mortality risk is still moderately strong with a spatial model,
   4      but much less strong than with a nonspatial model. In view of the sensitivity of the strength of
   5      evidence to the spatial model, the model assumptions warrant additional study. Even so, there is
   6      a considerable degree of coherence between the long-term and short-term mortality findings of
   7      the studies, with stronger evidence of a modest but significant short-term PM10 effect and a larger
   8      long-term fine particle (PM2 5 or sulfate) effect in the industrial midwest. The short-term effects
  9      are larger but less certain in southern California and the northeast, whereas the long-term effects
 10      seem less certain there.  Possible differences should be further explored.
 11
 12      9.6.4 Toxicologic Insights on Biological Plausibility
 13           Toxicological studies can play an integral role in answering key questions regarding
 14      biological plausibility of health effects associated with ambient PM. The materials presented
 15      below focus on the progress that toxicological studies have made towards answering the
 16      following two key questions.
 17        (1) What are the potential mechanisms by which PM causes health effects?
 18        (2) What specific  component or components of ambient PM cause health effects?
 19
 20      9.6.4.1  Mechanisms of Action
 21          Various studies  using particulate matter having diverse physicochemical characteristics
 22     have shown that these characteristics have a great impact on the specific response that is
 23     observed. Thus, there may, in fact, be multiple biological mechanisms responsible for observed
 24     morbidity/mortality because of exposure to ambient PM, and these mechanisms may be highly
 25     dependent on the type of particle in the exposure atmosphere. However, it should be noted that
 26     many controlled exposure studies used concentrations of PM that were much higher than those
 27     occurring in ambient air. Thus, some of the effects elicited may not occur with exposure to lower
 28     levels. Clearly, controlled exposure studies as yet have not been able to unequivocally determine
29     the particle characteristics and the toxicological mechanisms by which ambient PM may affect
30      biological systems. There is growing toxicological and epidemiological evidence that both the
31      cardiovascular and respiratory systems are affected by ambient PM. Nonetheless, understanding
        March 2001                               9-89        DRAFT-DO NOT QUOTE OR CITE

-------
  1     how participate air pollution causes and exacerbates cardiovascular or respiratory diseases
  2     remains an important goal.  The pathophysiological mechanisms involved in PM-associated
  3     cardiovascular and respiratory health effects remain unclear, but progress has been made since
  4     the 1996 PM AQCD was written. This section summarizes current hypotheses and reviews the
  5     toxicological evidence for potential pathophysiological mechanisms.
  6
  7     9.6.4.1.1 Direct Respiratory System Effects
  8          Emerging new toxicological evidence for three key mechanisms hypothesized as underlying
  9     direct effects of PM on the respiratory system is summarized below.
 10
 11     Lung Injury and Inflammation.  In the last few years, numerous studies have shown that
 12     instilled and inhaled ROFA, a product of fossil fuel combustion, can cause substantial lung injury
 13     and inflammation.  The toxic effects of ROFA largely result from its high content of soluble
 14     metals, and the pulmonary effects of ROFA can be reproduced by equivalent exposures to
 15     soluble metal salts. In contrast, controlled exposures of animals to  sulfuric acid aerosols, acid
 16     coated carbon, and sulfate salts cause little lung injury or inflammation even at high
 17     concentrations.  Inhalation of concentrated ambient PM (which contains only small amounts of
 18     metals) by laboratory animals at concentrations in the range of 100  to 1000 /^g/m3 have been
 19     shown in some (but not all) studies to cause mild pulmonary injury and inflammation. Rats with
 20     SO2-induced bronchitis and monocrotaline-treated rats have a greater inflammatory response to
 21      concentrated ambient PM than healthy rats.  These studies suggest that exacerbation of
 22     respiratory disease by ambient PM may be caused, in part, by lung injury and inflammation.
 23
 24      Increased Susceptibility to Respiratory Infections. There are no published studies on the effects
 25      of inhaled concentrated ambient PM on host susceptibility to infectious agents.  In vivo exposure
 26      of mice to acid-coated carbon particles at a mass concentration of 10,000 ^g/m3 causes decreased
27      phagocytic activity of alveolar macrophages even in the absence of lung injury (Ohtsuka et al.,
28      2000).  More studies are needed on the effects of concentrated ambient PM on the pulmonary
29      immune defense system.
30
        March 2001                               9-90        DRAFT-DO NOT QUOTE OR CITE

-------
  1     Increased Airway Reactivity and Asthma Exacerbation. The strongest toxicologic evidence
  2     supporting this hypothesis is from studies on diesel particulate matter (DPM). Diesel particulate
  3     matter has been shown to increase production of antigen-specific IgE in mice and humans
  4     (summarized in Section 8.2.4.3). In vitro studies have suggested that the organic fraction of
  5     DPM is involved in the increased IgE production. The ROFA leachate also enhances antigen-
  6     specific airway reactivity in mice (Goldsmith et al., 1999), indicating that soluble metals also can
  7     enhance an allergic response. However, in this same study, exposure of mice to concentrated
  8     ambient PM did not affect antigen-specific airway reactivity.  It is premature to conclude from
  9     this one experiment that concentrated ambient PM does not exacerbate allergic airways disease
 10     because the chemical composition of the PM (as indicated by studies with DPM and ROFA) may
 11     be more important than the mass concentration.
 12
 13     9.6.4.1.2 Systemic Effects Secondary to Lung Injury
 14          When the 1996 PM AQCD was written, it was thought that cardiovascular-related
 15     morbidity and mortality most likely would be secondary to impairment of oxygenation or some
 16     other consequence of lung injury and inflammation.  There is some toxicological evidence for the
 17     following mechanisms for adverse systemic effects secondary to lung injury.
 18
 19     Impairment  of Heart Function by Lowering Blood Oxygen Levels and Increasing the Work of
 20     Breathing. Instillation of ROFA has been shown to cause a 50% mortality rate in
 21      monocrotaline- treated rats (Watkinson et al., 2000). Although blood oxygen levels were not
 22     measured in this study, there were ECG abnormalities consistent with severe hypoxemia in about
 23      half of the rats that subsequently died.  Given the severe inflammatory effects of instilled ROFA
 24      and the fact that monocrotaline-treated rats have increased lung permeability as well as
 25      pulmonary hypertension, it is plausible that instilled ROFA can cause severe hypoxemia leading
 26      to death in this rat model. However, results from studies in which animals (normal and
27      compromised) were exposed to concentrated ambient PM (at concentrations many times higher
28      than would be encountered in the United States) indicate that ambient PM is unlikely to cause
29      severe disturbances in blood oxygenation or pulmonary function. However, even a modest
30      decrease in oxygenation can have serious consequences in individuals with ischemic heart
31      disease.  For example, reducing arterial blood saturation from 98 to 94% by either mild hypoxia

        March 2001                              9-91        DRAFT-DO NOT QUOTE  OR CITE

-------
 1     or by exposure to 100 ppm CO significantly reduced the time to onset of angina in exercising
 2     volunteers (Kleinman et al., 1998). Thus, more information is needed on the effects of PM on
 3     arterial blood gases and pulmonary function to fully address the above hypothesis.
 4
 5     Lung Inflammation and Cytokine Production Leading to Systemic Hemodynamic Effects.
 6     It has been suggested that systemic effects of particulate air pollution may be caused by
 7     activation of cytokine production in the lung (Li et al., 1997).  In support of this idea,
 8     monocrotaline-treated rats exposed to inhaled ROFA showed increased pulmonary cytokine gene
 9     expression, bradycardia, hypothermia, and increased arrhythmias (Watkinson et al., 2000).
10     However, spontaneously hypertensive rats had a similar cardiovascular response to inhaled
11     ROFA (except they also developed ST segment depression) with no increase in pulmonary
12     cytokine gene expression.  Studies in dogs exposed to concentrated ambient PM  showed minimal
13     pulmonary inflammation and no positive staining for IL-8, IL-1, or TNF in airway biopsies.
14     However, the time of onset of ischemic ECG changes following coronary artery occlusion
15     decreased significantly (Godleski et al., 2000).  Thus, there is not a clear-cut link between
16     changes in cardiovascular function and production of cytokines in the lung.  Because human and
17     animal exposure studies of ambient PM are using increasingly sophisticated and  sensitive
18     measures of cardiac function, basic information on the effects of mild pulmonary injury on these
19     cardiac endpoints is needed to understand the mechanisms by which inhaled PM may affect the
20     heart.
21
22     Increased Risk of Heart Attacks and Strokes Because of Increasing Blood Coagulability
23     Secondary to Lung Inflammation. There is abundant evidence linking risk of heart attacks and
24     strokes to small prothrombotic changes in the blood coagulation system; and some new
25     epidemiologic evidence (discussed earlier above) indicates that ambient PM may affect blood
26     coagulation and/or other blood characteristics related to increased risk of serious cardiac
27     outcomes. However, there is no published experimental evidence as of yet that moderate lung
28     inflammation increases blood coagulability; a high dose (8,300 Mg/kg) of instilled ROFA did
29     cause increased levels of fibrinogen, but no effect was seen at lower doses (Gardner et al., 2000).
30     Also, exposure of dogs to concentrated ambient PM also had no effect on fibrinogen levels
31     (Godleski et al., 2000).  The coagulation system is as multifaceted and complex as the immune

       March 2001                              9-92        DRAFT-DO NOT QUOTE OR CITE

-------
  1     system, and there are many other sensitive and clinically significant parameters that should be
  2     examined in addition to fibrinogen. Thus, it is premature to draw any strong conclusions about
  3     ambient PM exposure effects on cardiovascular morbidity or mortality being mediated via PM
  4     effects on blood coagulability or other blood characteristics.
  5
  6     Particulate Matter and Lung Interactions Potentially Affecting Hematopoiesis.  Instillation of
  7     fine carbon particles (20,000 /wg/rabbit) stimulated release of PMNs from the bone marrow
  8     (Terashima et al., 1997).  In support of this hypothesis, Gordon and colleagues reported that the
  9     percentage of PMNs in the peripheral blood increased in rats exposed to ambient PM in some but
 10     not all exposures. On the other hand, Godleski et al. (2000) found no changes in peripheral
 11     blood counts of dogs exposed to concentrated ambient PM.  Thus,  direct evidence that ambient
 12     concentrations of PM can affect hematopoiesis is still needed.
 13
 14     9.6.4.1.3 Direct Effects on the Heart
 15          Changes in heart rate and heart rate variability associated with ambient PM exposure have
 16     been reported in animal studies (Godleski et al., 2000; Gordon et al., 2000), in several human
 17     panel studies (described in Chapter 6), and in a reanalysis of data from the MONICA  study
 18     (Peters et al., 2000). Some of these studies included endpoints related to respiratory effects, but
 19     few significant adverse respiratory changes were detected. This raises the possibility that
 20     ambient PM may have effects on the heart that are independent of adverse changes in the lung.
 21      There is precedent for this idea:  tobacco smoke (a mixture of combustion-generated gases and
 22     particles) causes cardiovascular disease by mechanisms independent of its lung effects.
 23
 24      Heart Rate Variability. Epidemiological studies have linked fine particulate air pollution with
 25      cardiopulmonary morbidity and mortality (Schwartz and Morris, 1995; Burnett et al.,  1995;
 26      Morris et al., 1995; Schwartz, 1997), but the underlying biologic mechanisms remain unclear.
 27      Recently, attention has focused on possible effects on heart rate (HR) variability as a potential
 28      mechanism underlying cardiovascular morbidity and mortality effects associated with ambient
29      PM. During recent decades, a large clinical database has developed describing a significant
30      relationship between autonomic dysfunction and sudden cardiac death. Moreover, low HR
31      variability has been implicated as a marker for a number of pathophysiological conditions

        March 2001                               9-93       DRAFT-DO NOT QUOTE OR CITE

-------
  1      including myocardial infarction (Task Force of the European Society of Cardiology and the
  2      North American Society of Pacing and Electrophysiology 1996; Bigger et al., 1992; Hayano
  3      et al., 1990; Kleiger et al., 1987; Martin et al., 1987; Singer et al., 1988). This is further
  4      elaborated in Appendix 6-B.
  5           Some studies (Liao et al., 1999; Pope et al., 1999b) provide new evidence for relationships
  6      between ambient PM and decreased HR variability.  Pope et al. (1999b) reported an association
  7      between particulate air pollution, heart rate, and HR variability. A relationship between PM and
  8      HR variability also is supported by laboratory animal studies. Combustion particles instilled into
  9      rat lungs produce arrhythmias and a doubling of mortality (Watkinson et al.,  1998).
 10      Concentrated ambient air particles breathed by dogs elicited electrocardiographic changes,
 11      including T-wave alterans and arrhythmias (Godleski et al., 1998).
 12
 13      Autonomic Control of the Heart and Cardiovascular System. There is growing evidence for
 14      the idea that inhaled particles could affect the heart through the autonomic nervous system.
 15      Activation of neural receptors in the lung is a logical area to investigate. Studies in conscious
 16      rats have  shown that inhalation of wood smoke causes marked changes  in sympathetic and
 17      parasympathetic input to the cardiovascular system that are mediated by neural reflexes
 18      (Nakamura and Hayashida, 1992).  Although research on airway neural  receptors and neural -
 19      mediated reflexes is a well-established discipline, the cardiovascular effects of stimulating airway
20      receptors continue to receive less attention than the pulmonary effects.  Previous studies of
21      airway reflex-mediated cardiac effects usually have employed very high doses of chemical
22      irritants, and the results may not be applicable to air pollutants. There is a need for basic
23      physiological studies to examine cardiovascular system effects when airway and alveolar neural
24      receptors  are stimulated in a manner relevant to air pollutants.
25
26      Uptake of Particles and Distribution of Soluble Substances into the Systemic Circulation.
27      Drugs can be delivered rapidly and efficiently to the systemic circulation by inhalation (as occurs
28      with nicotine from inhaled cigarette smoke). This implies that the pulmonary vasculature absorbs
29      inhaled materials, including charged substances such as small proteins and peptides. It is likely
30      that soluble materials absorbed onto airborne PM find their way into the bloodstream, but it is


        March 2001                               9-94       DRAFT-DO NOT QUOTE OR CITE

-------
  1      not clear whether the particulate materials themselves enter the blood.  It is anticipated that more
  2      information will be available on this important question in the next few years.
  3
  4      9.6.4.2  Links Between Specific Particulate Matter Components and Health Effects
  5           Key to enhancing confidence in the biological plausibility of ambient PM health effects is
  6      the need to identify those components of airborne PM responsible for the health effects and for
  7      placing susceptible individuals at risk.  The plausibility of epidemiologically demonstrated
  8      associations between ambient PM and increases in morbidity and mortality has been questioned
  9      because associations with health effects have been observed at very low PM concentrations.
 10      To date, toxicology studies on PM have provided only limited evidence for specific PM
 11      components being likely responsible for cardiovascular or respiratory effects of ambient PM.
 12      The latest available experimental information concerning potential contributions of individual
 13      physical and chemical factors of particles to cardiorespiratory effects is summarized below.
 14
 15      Acid Aerosols.  There is relatively little new information on the effects of acid aerosols, and the
 16      basic conclusions of the 1996 PM AQCD remain unchanged. It previously was concluded that
 17      acid aerosols cause little or no change in pulmonary function in healthy subjects, but asthmatics
 18      may experience small decrements in pulmonary function.  These conclusions are further
 19      supported by a recent study by Linn and colleagues (1997), in which healthy children (and
 20      children with allergy or asthma) were exposed to sulfuric acid aerosol (100 /ug/m3) for 4 h. There
 21      were no significant effects on symptoms or pulmonary function when the entire group was
 22      analyzed, but the allergy group had a significant increase in symptoms after the acid aerosol
 23      exposure (albeit to distinctly higher than typical ambient acid concentrations).
 24           Although pulmonary effects of acid aerosols have been the subject of extensive research,
 25      the cardiovascular effects of acid aerosols have received much less attention. However,
 26      inhalation of acetic acid fumes has been reported to cause reflex mediated increases in blood
27      pressure in normal and spontaneously hypertensive rats (Zhang et al., 1997). Thus, acid
28      components should not be ruled out as possible mediators of PM health effects.  In particular, the
29      cardiovascular effects of acid aerosols (at realistic concentrations) need further investigation.
30


        March 2001                                9-95         DRAFT-DO NOT QUOTE OR CITE

-------
  1      Metals. The previous 1996 PM AQCD mainly relied on data related to occupational exposures
  2      to evaluate the potential toxicity of metals in particulate air pollution.  Since that time, in vivo
  3      and in vitro studies using ROFA or soluble transition metals have contributed substantial new
  4      information on the health effects of particle-associated soluble metals. Although there are some
  5      uncertainties about differential effects of one transition metal versus another, water soluble
  6      metals  leached from ROFA, albeit at high concentrations, consistently have been shown to cause
  7      cell injury and inflammatory changes in vitro and in vivo.
  8           Even though it is clear that combustion particles that have a high content of soluble metals
  9      can cause lung injury and even death in compromised animals, it has not been established that the
10      small quantities of metals associated with relatively low concentrations of ambient particles are
11      sufficient to cause health effects.  In studies in which various ambient and emission source
12      particulates were instilled into rats, the soluble metal content did appear to be the primary
13      determinant of lung injury.  However, one published study compared the effects of inhaled
14      ROFA  (at 1 mg/m3) to concentrated ambient PM (of 475 to 900 Aig/m3) in normal and
15      SO2-induced bronchitic rats. A statistically significant increase in at least one lung injury marker
16      was seen in bronchitic rats in only one out of four of the concentrated ambient PM exposures,
17      and inhaled ROFA had no effect, even though the content of soluble iron, vanadium, and nickel
18      was much higher in the ROFA sample. Thus, the potential roles of metals in contributing to
19      health effects of ambient PM remains to be more clearly established.  There has been increasing
20      attention focused in recent years on the possibility of ultrafme particles playing a major role in
21      observed ambient PM health effects due to large absolute number counts and/or  surface area of
22      ultrafine particles deposited in the lung.
23
24      Ultrafine Particles.  When this subject was reviewed in the 1996 PM AQCD, it  was not known
25      whether the pulmonary toxicity of freshly generated ultrafine Teflon particles was because of
26      particle size or a result of absorbed fumes. Subsequent studies with other types of ultrafine
27      particles have shown that the chemical constituents of ultrafmes substantially modulate their
28      toxicity. Inhalation of MgO particles, for example,  produces far fewer respiratory effects than
29      does ZnO (Kuschner et al.,  1997). Also, inhalation exposure of normal rats to ultrafine carbon
30      particles generated by electric arc  discharge caused  minimal lung inflammation (Elder et al.,
31      2000) compared to ultrafine Teflon or metal particles.  On the other hand, instillation of ultrafine

        March 2001                               9-96         DRAFT-DO NOT QUOTE OR CITE

-------
  1      carbon black caused substantially more inflammation than did the same dose of fine particles of
  2      carbon black, suggesting that ultrafme particles cause more inflammation than larger particles (Li
  3      et al., 1997). However, the chemical constituents of the two carbon black sizes were not
  4      analyzed and it is uncertain that the chemical composition was the same. As with acid aerosols,
  5      studies of ultrafme particles have focused largely on effects in the lung, but it is possible that
  6      inhaled ultrafme particles may have systemic effects that are independent of lung effects.  It is
  7      also important to note that at least one very recent new epidemiology study (Wichmann et al.,
  8      2000) provides interesting new evidence implicating both ultrafine (nuclei-mode) and
  9      accumulation-mode fine particles in PM-mortality relationships.
 10
 11      Diesel Exhaust Particulate Matter. As described in Section 8.2.4.2, there is growing
 12      toxicological evidence that diesel exhaust particulate matter (DPM) exacerbates the allergic
 13      response to inhaled antigens.  The organic fraction of diesel exhaust has been linked to
 14      eosinophil degranulation and induction of cytokine production suggesting that the organic
 15      constituents of DPM are responsible for the immune effects. It is not known whether the
 16      adjuvant-like activity of DPM is unique or whether other combustion-related particles have
 17      similar effects. It is important to compare the immune effects of other source-specific emissions,
 18      as well as concentrated ambient PM, to DPM to determine the extent to which exposure to diesel
 19      exhaust may contribute to the incidence and severity of allergic rhinitis and asthma. Other types
 20      of noncancer and carcinogenic (especially lung cancer) effects are of concern with regard to
 21      DPM exposures, as discussed in a separate EPA Health Assessment Document for Diesel
 22      Exhaust (U.S. Environmental Protection Agency, 2000b).
 23
 24      Organic Compounds.  Published research on the acute effects of particle-associated organic
25      carbon constituents is conspicuous by its relative absence, except for diesel exhaust particles.
26      Like metals, organics are common constituents of combustion-generated particles and are found
27      in ambient PM samples over a wide geographical range. Organic carbon constituents comprise a
28      substantial portion of the mass of ambient PM (10 to 60% of the total dry mass [Turpin, 1999]).
29      The organic fraction of ambient PM has been evaluated for its mutagenic effects. Although the
30      organic fraction of particulate matter is a poorly characterized heterogeneous mixture of a widely


        March 2001                               9-97        DRAFT-DO NOT QUOTE OR CITE

-------
 1      varying number of different compounds, strategies have been proposed for examining the health
 2      effects of potentially important organic constituents (Turpin, 1999).
 3
 4      Bioaerosols. Recent studies support the conclusion of the 1996 PM AQCD that bioaerosols, at
 5      the concentrations present in the ambient environment, do not likely account for the health
 6      effects of ambient PM. Dose response studies in healthy volunteers exposed to 0.55 and 50 /u.g
 1      endotoxin, by inhalation, showed the threshold for pulmonary and systemic effects for endotoxin
 8      to be between 0.5 and 5.0 yUg (Michel et al., 1997). Available  information suggests that ambient
 9      concentrations of endotoxin are very low and do not exceed 0.5 ng/m3. Also, Monn and Becker
10      (1999) found cytokine induction by human monocytes, characteristic of endotoxin activity, in the
11      coarse size fraction of outdoor PM but not in the fine fraction.
12
13      Concentrated Ambient Particle Studies (CAPS). Ambient particle studies are potentially among
14      the most relevant in improving our understanding of the susceptibility of individuals to PM and
15      underlying mechanisms of toxicity. New studies have used collected urban PM for intratracheal
16      administration to healthy and compromised animals, and some recent work with inhaled
17      concentrated ambient PM has reported cardiopulmonary changes in rodents and dogs at high
18      concentrations of fine PM. Thus, despite difficulties in extrapolating from the bolus delivery
19      used in such studies, they are contributing some new evidence enhancing the plausibility of
20      health effects of fine particles observed in epidemiologic studies.
21
22      Animal Models of Susceptibility.  Progress has been made in understanding the role of
23      individual susceptibility to ambient PM effects. Studies have shown consistently that animals
24      with compromised health, either genetic or induced, are more susceptible to instilled or inhaled
25      particles, although the increased animal-to-animal variability in these models has created
26      problems.  Moreover, because PM seems to affect broad categories of disease states ranging from
27      altered cardiac rhythms to pulmonary infection, it can be difficult to know what disease models
28      to use in understanding the biological plausibility of the adverse health effects of PM. Thus, the
29      identification of susceptible animal models has been slow, but, overall, it represents solid
30      progress when one considers the numbers of people necessary in epidemiology studies to develop
31      the statistical power to detect small increases in morbidity and mortality.

        March 2001                                9-98        DRAFT-DO NOT QUOTE OR CITE

-------
  1     9.7 RISK FACTORS AND POTENTIALLY SUSCEPTIBLE POPULATION
  2         GROUPS
  3     9.7.1  Introduction
  4          The 1996 PM AQCD identified several population groups as likely being at increased risk
  5     for experiencing health impacts of ambient PM exposure. Elderly individuals (>65 years) were
  6     most clearly identified, along with those having preexisting cardiovascular or respiratory disease
  7     conditions.  The latter likely include smokers and ex-smokers as individuals comprising large
  8     percentages of cardiovascular and respiratory disease (e.g., COPD) sufferers. Individuals with
  9     asthma also were, albeit more tentatively, identified as a likely susceptible population group as
 10     well, as were children.  The new studies appearing since the 1996 PM AQCD, as assessed earlier
 11     in this document and chapter, provide considerable additional evidence substantiating all of the
 12     above named groups as likely being at increased risk for ambient PM-related morbidity or
 13     mortality effects.  Information related to factors contributing to  such increased susceptibility or
 14     useful in placing the potential public health impacts in perspective is presented below.
 15
 16     9.7.2  Preexisting Disease as a Risk Factor for Particulate Matter Health
 17            Effects
 18          Earlier available information reviewed in the 1996 PM AQCD has now been extensively
 19     augmented by new studies that substantiate well that preexisting disease conditions are among
 20     the most important key risk factors for ambient PM health effects. Cardiovascular- and
 21      respiratory-related diseases have been shown to be of greatest concern, thus far, in relation to
 22     increasing risk for PM mortality and morbidity. Table 9-9 shows the numbers of U.S. cases
 23      reported for COPD, asthma, heart disease, and hypertension.
 24
 25      9.7.2.1  Ambient Particulate Matter Exacerbation of Cardiovascular Disease Conditions
 26          Exacerbation of heart disease has been epidemiologically associated, not only with ambient
 27      PM, but also with other combustion-related ambient pollutants such as CO. Thus, while leaving
28      little doubt that ambient PM exposures importantly affect CVD  mortality and morbidity, the
29      quantitation of the proportion of risk for such exacerbation specifically attributable to ambient
30      PM exposure is difficult. Recent studies (e.g.,  concentrated ambient particle studies [CAPS])
31      have demonstrated cardiovascular effects in response to ambient particle exposures, and
        March 2001                               9-99        DRAFT-DO NOT QUOTE OR CITE

-------
tr
TABLE 9-9. INCIDENCE OF SELECTED CARDIORESPIRATORY DISORDERS BY AGE AND
                        BY GEOGRAPHIC REGION, 1996
      (reported as incidence per thousand population and as number of cases in thousands)
0
o








1
0
o


a
!>
H
1
a
o
o
H
o
a
o
H
tn
O
73
o
H
W

Chronic Condition/Disease
COPD*
Incidence/ 1,000 persons
No. cases x 1,000
Asthma
Incidence/1,000 persons
No. cases x 1,000
Heart Disease
Incidence/ 1,000 persons
No. cases x ] ,000

HD-ischemic
Incidence/1,000 persons
No. cases x 1,000
HD-rhythmic
Incidence/ 1 ,000 persons
No. cases x 1,000

Hypertension
Incidence/ 1,000 persons
No. cases x 1,000


All Ages

60.4
15,971

55.2
14,596

78.2
20,653


29
7,672

33
8,716


107.1
28,314


Under 45

50.6
9,081

58.9
10,570

33.1
5,934


2.5
453

24.3
4,358


30.1
5,391

Age
45-64

72.3
3,843

48.6
2,581

116.4
6,184


51.6
2,743

40.7
2,164


214.1
11,376


Over 65

95.9
3,047

45.5
1,445

268.7
8,535


140.9
4,476

69.1
2,195


363.5
11,547


Over 75

99.9
1,334

48.0
641

310.7
4,151


154.6
2,065

73.1
977


373.8
4,994

Regional
NE MW S W

57.8 67.6 59.4 56.6


61.8 56.6 51.8 52.9


88.5 78.0 77.0 70.4



28.9 30.0 30.7 25.0


40.2 34.0 28.1 32.9



109.3 108.2 113.5 93.7


'Total chronic bronchitis and emphysema.

Source: Adams et al. (1999).


































-------
  1     studies utilizing other techniques also have produced various results suggesting some plausible
  2     mechanisms for cardiovascular effects.  However, much remains to be resolved with regard to
  3     delineation of dose-response relationships for the induction of such effects and the extrapolation
  4     of such to estimate effective human equivalent exposures to ambient PM (or specific constituent)
  5     concentrations.
  6          Schwartz (1999) has argued that independent effects of both PM and other pollutants are
  7     biologically plausible. In the case of PM10, Schwartz's plausibility argument draws on the
  8     emerging literature, which has demonstrated effects of ambient PM on pulmonary inflammation
  9     in laboratory animals and human volunteers (Gilmour et al., 1996; Salvi et al., 1999), toxicity of
 10     transition metals carried by combustion- generated particles (Costa and Dreher,  1999), effects on
 11     cardiac dysfunction in animals with preexisting cardiopulmonary disease (Godleski et al., 1996;
 12     Watkinson et al., 1998), and new epidemiologic evidence of associations between ambient  PM
 13     and physiologic changes in cardiac function (Pope et al., 1999a,b; Liao et al., 1999; Peters et al.,
 14     1999b; Gold et  al., 1998, 2000) and plasma viscosity (Peters et al., 1997a) in humans.  For  CO,
 15     his argument is based on well-established effects of CO on oxygen transport by hemoglobin,
 16     although such an impact typically is observed only at much higher CO concentrations than those
 17     seen in these ambient studies.  Although much more research is needed to clarify and confirm the
 18     hypothesized linkages among these new findings, these arguments provide an initial framework
 19     for such a linkage.
 20          One very recently published HEI report on an epidemiologic study conducted by Goldberg
 21      et al. (2000) in Montreal, Canada, provides especially interesting new information regarding
 22     types of medical conditions potentially putting susceptible individuals  at increased risk for PM-
 23      associated mortality effects, and obtained results suggestive of other diseases with cardiovascular
 24      complications being affected by ambient PM.  First, the immediate causes of death, as listed on
 25      death certificates, were evaluated in relation to various ambient PM indices (TSP, PM10
 26      estimated PM2 5, COH, sulfates, and extinction coefficients) lagged for 0 to 4 days. Significant
 27      associations were seen between each of the PM measures and total nonaccidental deaths,
28      respiratory diseases, and diabetes, with an approximate 2%  increase in excess nonaccidental
29      mortality being observed per 9.5 /^g/m3 interquartile increase in 3-day mean estimated PM2 5
30      exposure. When underlying clinical conditions identified in the decedents' medical records were
31      then evaluated in relation to ambient PM measures, all three measures (COH, sulfate, and

        March 2001                               9-101        DRAFT-DO NOT QUOTE OR CITE

-------
 1      estimated PM2 5) were associated with acute lower respiratory disease, congestive heart failure,
 2      and any cardiovascular disease.  Predicted PM2 5 and COH also were reported to be associated
 3      with cancer, chronic coronary artery disease, and any coronary artery disease, whereas sulfate
 4      was associated with acute and chronic upper respiratory disease. None of the three PM measures
 5      were related to airways disease, acute coronary artery disease, or hypertension.  These results
 6      both tend to confirm previous findings identifying those with preexisting cardiopulmonary
 7      diseases as being at increased risk for ambient PM effects and implicate another possible risk
 8      factor, diabetes (which involves  cardiovascular complications as it progresses), as a potential
 9      susceptibility condition putting individuals at increased risk for ambient PM effects.
10           To the extent that observed associations of ambient PM with heart disease exacerbation
11      prove to be causal and specific to PM, they would be of genuine public health concern. In the
12      U.S. in 1997, there were about 4,188,000 hospital discharges with heart disease as the first-listed
13      diagnosis (Lawrence and Hall, 1999). Among these, about 2,090,000 (50%) were for ischemic
14      heart disease, 756,000 (18%) for myocardial infarction or heart attack (a subcategory of ischemic
15      heart disease), 957,000 (23%) for congestive heart failure, and 635,000 (15%) for cardiac
16      dysrhythmias. Also, there were 726,974 deaths from heart disease (Hoyert et al.,  1999). Even a
17      small percentage reduction in admissions or deaths from heart disease would predict a large
18      number of avoided cases.
19
20      9.7.2.2 Ambient Particulate Matter Exacerbation of Respiratory Disease Conditions
21           Many investigators also have observed associations  of short-term fluctuations in ambient
22      PM with daily frequency of respiratory illness. In most cases, exacerbation of preexisting
23      respiratory illness has been assessed, although some cases of acute respiratory infection may be
24      considered as occurrence of new illness, especially in young people. Symptoms of acute
25      respiratory distress in children have been linked to elevated PM concentrations in studies in the
26      United States and other countries, with asthmatics apparently more susceptible than
27      nonasthmatics. However, some  studies also have found associations between child respiratory
28      symptoms or reduced lung function and other pollutants (such as O3) in addition to PM or no
29      significant relationship with air pollution. The credibility of ambient PM plausibly being linked
30      to exacerbation of preexisting respiratory disease (e.g., asthma) is enhanced by newly reported
31      dosimetry data noted earlier, which show greater lung  deposition of l-/ini particles in people

        March 2001                               9-102        DRAFT-DO NOT QUOTE OR CITE

-------
  1
  2
  3
  4
  5
  6
  7
  8
  9
10
11
12
13
14
15
16
17
18
19
with varying degrees of airway obstruction than in healthy subjects.  The increased deposition
was greatest for COPD patients and asthmatics, but smokers also showed increased deposition as
well.
     In the United States in 1997, there were 3,475,000 hospital discharges for respiratory
diseases:  38% for pneumonia, 14% for asthma, 13% for chronic bronchitis, 8% for acute
bronchitis, and the remainder not specified (Lawrence and Hall, 1999).  Of the 195,943 deaths
recorded as caused by respiratory diseases, 44% resulted from acute infections, 10% for
emphysema and bronchitis, 2.8% for asthma, and 42% for unspecified COPD (Hoyert et al.,
1999).  Again, even a small percentage reduction in respiratory-related diseases could calculate
out to a large number of avoided cases.

9.7.3  Aged-Related At-Risk Population Groups:  The Elderly and Children
     Why are the very young and the very old apparently  among those most affected by PM air
pollution? One major factor in increased susceptibility to  air pollution is the presence of a
preexisting illness, as shown by Zanobetti and  Schwartz (2000). The youngest children have the
highest rates of respiratory illnesses, as shown  in the Table 9-10, which may be an important
factor in their apparently greater susceptibility  to the adverse effects of PM air pollution.
              TABLE 9-10. NUMBER OF ACUTE RESPIRATORY CONDITIONS PER
                    100 PERSONS PER YEAR, BY AGE: UNITED STATES, 1996
Type of Acute Condition
Respiratory Conditions
Common Cold
Other Acute Upper Respiratory
Infections
Influenza
Acute Bronchitis
Pneumonia
Other Respiratory Conditions
All
Ages
78.9
23.6
11.3
36.0
4.6
1.8
1.7
Under 5
Years
1294
48.6
13.1
53.7
*7.2
*3.9
*2.9
5-17
Years
101.5
33.8
15.0
443
4.3
*1 7
*2.4
18-24
Years
86.0
23.8
16.1
40.5
*3.9
*1.4
*0.4
25-44
Years
76.9
187
11.6
38.1
5.1
*1.3
*2.0
45
Total
53.3
16.1
7.0
23.3
38
*2.0
*1 1
Years and Over
45-64
Years
55.9
16.4
7.5
26.1
3.5
*0.9
*1 5
65 Years
and Over
49.0
157
6.1
18.6
*4.4
*3.8
*0 5
        Source: Adams et al. (1999).
       March 2001
                                       9-103
DRAFT-DO NOT QUOTE OR CITE

-------
 1           In addition to their higher incidences of preexisting respiratory conditions, several other
 2      factors may render children and infants more susceptible to PM exposures, including a greater
 3      amount of time spent outdoors, greater activity levels and breathing rates, higher doses per body
 4      weight and lung surface area, and potential irreversible effects on children's developing lungs.
 5      For example, PM doses on a per kilogram body weight basis are much higher for children than
 6      for adults. This is displayed graphically in the Figure 9-14, which indicates that the amount of
 7      air inhaled per kilogram body weight increases dramatically as  age decreases below adult levels,
 8      with the inhalation rate (in cubic meters per kilogram a day) of a 10-year-old being roughly twice
 9      that of a 30-year-old person, and this estimate does not consider higher personal exposure
10      concentrations that a child is usually exposed to as a result of higher activity levels.  Thus, on a
11      per unit body weight basis, children receive higher doses of air pollution than adults, consistent
12      with lung deposition information discussed earlier in this chapter.
13           Child-adult dosage disparities are even greater when viewed on a per lung area basis. This
14      may be more important than body weight if the number of particle "hits" per unit lung surface is
15      an important health impact metric, which it may well be for ultrafme particles.  A newborn infant
16      has approximately 10 million alveoli versus some 300 million as an adult.  The alveolar surface
17      area increases from approximately 3 m2  at birth to about 75 m2  in adulthood, causing the dose
18      delivered per lung surface area for infants and children to be much higher than in adults, even
19      given the same personal exposures (which is not the case, as they generally have greater PM,0
20      personal exposures than adults, as noted above). Thus, observed high PM air pollution-hospital
21      admissions associations for infants may result from PM doses that are significantly higher in
22      children than in adults, when one considers children's higher personal  exposures, their greater
23      activity rates, and  their smaller body weights and lung surface areas.
24           As discussed by Plopper and Fanucchi (2000), the limited experimental and epidemiologic
25      studies currently available identify the early postneonatal period of lung development as a time of
26      high susceptibility for lung damage created by exposure to environmental toxicants.  This is
27      likely the reason for the above noted high rate of respiratory infectious diseases in young
28      children.  In  addition to their diminished immune status, infants are growing rapidly, and  some
29      recent (though limited) evidence supports the hypothesis that environmental pollution can
30      significantly alter  development of the respiratory system at that period of life.  In experimental
31      animals, for  example, elevated neonatal susceptibility to lung-targeted toxicants has been

        March 2001                               9-104       DRAFT-DO NOT QUOTE OR CITE

-------
              0.6
              0.5
              0.4
         03
              0.3
         CD
         -E   0.2

              0.1
                                                     40
              T~
               50
           60
70
80
                                                  Age (y)
       Figure 9-14. Inhalation rates on a per body-weight basis for males (•) and females (±) by
                   age (Layton, 1993).
1     reported at doses "well below the no-effects level for adults" (Plopper and Fanucchi, 2000;
2     Fanucchi and Plopper, 1997). In addition, acute injury to the lung during early postnatal
3     development causes a failure of normal repair processes, including down-regulation of cellular
4     proliferation at sites of injury (Smiley-Jewel et al., 2000, Fanucchi et al., 2000). Both infants'
5     diminished defenses and pollution-induced impairment of repair mechanisms therefore can
6     coincide during infancy, making the neonatal and postneonatal period one of potentially
7     especially elevated susceptibility to damage by environmental toxicants like PM.
8          Other information reviewed earlier in this document and chapter highlighted new evidence
9     pointing toward enhanced asthma symptoms, pulmonary function  decrements, and asthma-
      March 2001
9-105
DRAFT-DO NOT QUOTE OR CITE

-------
 1      related doctors' visits and hospital admissions being associated with ambient PM exposures.
 2      Generally higher activity levels in children and other factors related to attaining adequate medical
 3      control of asthma in children may put asthmatic children (especially physically active mild to
 4      moderate asthmatics) at particular risks for untoward effects of ambient PM among pediatric
 5      population groups.
 6           These and other types of health effects in children are emerging as a more important area of
 7      concern than in the 1996 PM AQCD.  Unfortunately, relatively little is known about the
 8      relationship of PM to the most serious health endpoints (low birth weight, preterm birth, neonatal
 9      and infant mortality, emergency hospital admissions and mortality in older children). Also, little
10      is yet known about involvement of PM exposure in the progression from less serious childhood
11      conditions, such as asthma and respiratory symptoms, to more serious disease endpoints later in
12      life. This is an important health issue because childhood illness or death may cost a very large
13      number of productive life-years. Lastly new epidemiologic studies of ambient PM associations
14      with increased non-hospital medical visits (physician visits) and asthma effects suggest likely
15      much larger health impacts and costs to society due to  ambient PM effects on children than just
16      those indexed by mortality and/or hospital admissions/visits.
17           In contrast to information noted above for children, elderly adults do not appear to be put at
18      increased risk because of difference in lung deposition, clearance, or retention of inhaled
19      particles associated with aging, per se. However, the possible gradual focal accumation of
20      previously inhaled PM material at bifurcations and carinal ridges in TB airways and release of
21      previously accumulated inhaled PM-derived materials  from lymph nodes could contribute to
22      enhanced susceptibility of elderly adults, especially those residing for long periods of time in
23      high PM exposure areas.
24           Probably of much more importance in placing elderly adults at increased risk for PM
25      effects is the higher propensity for such individuals to have preexisting cardiovascular or
26      respiratory disease conditions. Increased breathing rates due to compromised (e.g., obstructed)
27      lungs and airways or altered particle deposition patterns resulting from such conditions could be
28      among important factors increasing the risk for the elderly.
29
        March 2001                               9-106       DRAFT-DO NOT QUOTE OR CITE

-------
  1      REFERENCES

  2      Abbey, D. E.; Ostro, B. E.; Petersen, F.; Burchette, R. J. (1995) Chronic respiratory symptoms associated with
  3              estimated long-term ambient concentrations of fine participates less than 2.5 microns in aerodynamic
  4              diameter (PM25) and other air pollutants. J. Exposure Anal. Environ. Epidemiol. 5:  137-159.
  5      Abbey, D. E.; Burchette, R. J.; Knutsen, S. F.; McDonnell, W.  F.; Lebowitz, M. D.; Enright, P. L. (1998) Long-term
  6              particulate and other air pollutants and lung function in nonsmokers. Am. J. Respir. Crit. Care Med.
  7              158:289-298.
  8      Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. J.; Knutsen, S. F.; Beeson, W. L.; Yang, J. X. (1999)
  9              Long-term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir.
 10              Crit. Care Med. 159: 373-382.
 11      Adams, P. F.; Hendershot, G. E.; Marano, M. A. (1999) Current estimates from the National Health Interview
 12              Survey, 1996. Vital Health Stat. 10(200).
 13      Bigger, J. T., Jr.; Fleiss, J. L.; Steinman, R. C.; Rolnitzky, L. M.; Kleiger, R. E.; Rottman, J.  N. (1992) Frequency
 14              domain measures of heart period variability and mortality after myocardial infarction. Circulation
 15              85:164-171.
 16      Bobak, M.; Leon, D. A. (1999) Pregnancy outcomes and outdoor air pollution: an ecological study in districts of the
 17              Czech Republic 1986-8. Occup. Environ. Med. 56: 539-543.
 18      Broday, D. M.; Georgopoulos, P. G. (2000) Growth and deposition  of hygroscopic particulate matter in the human
 19              lungs. Aerosol Sci. Technol.: submitted.
 20      Brunekreef, B. (1997) Air pollution and life expectancy: is there a relation? Occup. Environ. Med. 54: 781-784.
 21      Burnett, R. T.; Dales, R. E.; Raizenne, M. E.; Krewski, D.; Summers, P. W.; Roberts, G. R.;  Raad-Young, M.;
 22              Dann, T.;-Brook, J. (1994) Effects of low ambient levels of ozone and sulfates on the  frequency of
 23              respiratory admissions to Ontario hospitals. Environ. Res. 65: 172-194.
 24      Burnett, R. T.; Dales, R.; Krewski, D.; Vincent, R.; Dann, T.; Brook, J. R. (1995) Associations between ambient
 25              particulate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. Am.  J.
 26              Epidemiol. 142: 15-22.
 27      Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997) The role of particulate size and chemistry in the
 28              association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
 29              Environ. Health Perspect. 105: 614-620.
 30      Burnett, R. T.; Cakmak, S.; Raizenne, M. E.; Stieb, D.; Vincent, R.; Krewski, D.; Brook, J. R.; Philips, O.;
 31              Ozkaynak, H. (1998) The association between ambient carbon monoxide levels and daily mortality in
 32              Toronto, Canada. J. Air Waste Manage. Assoc. 48: 689-700.
 33      Burnett, R. T.; Smith-Doiron, M.; Stieb, D.; Cakmak, S.; Brook, J. R. (1999) Effects of particulate and gaseous air
 34              pollution on cardiorespiratory hospitalizations. Arch. Environ. Health 54: 130-139.
 35      Burnett, R. T.; Brook, J.; Dann, T.; Delocla, C.; Philips, O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
 36              (2000) Association between particulate- and gas-phase components of urban air pollution and daily
 37              mortality in eight Canadian cities. In: Grant, L. D., ed.  PM2000: particulate matter and health. Inhalation
 38              Toxicol. 12(suppl.4): 15-39.
 39      Castillejos, M.; Borja-Aburto, V. H.; Dockery, D. W.; Gold, D. R.; Loomis, D. (2000) Airborne coarse particles and
 40              mortality. In: Inhalation Toxicology: proceedings of the third colloquium on particulate air pollution and
 41              human health; June, 1999; Durham, NC. Inhalation Toxicology 12(suppl. 1): 61-72.
 42      Chen, L.; Yang, W.; Jennison, B. L.; Omaye, S. T. (2000) Air particulate pollution and hospital admissions for
 43              chronic obstructive pulmonary disease in Reno, Nevada. Inhalation Toxicol. 12: 281-298.
 44      Chock, D. P.; Winkler, S.; Chen, C. (2000) A study of the association between daily mortality and ambient air
45              pollutant concentrations in Pittsburgh, Pennsylvania. J. Air Waste Manage. Assoc. 50: 1481-1500.
46      Choudhury, A. H.; Gordian, M. E.; Morris, S. S. (1997) Associations between respiratory illness and PMIO air
47              pollution. Arch. Environ. Health 52:  113-117.
48      Cifuentes, L. A.; Vega, J.; Kopfer, K.; Lave, L. B. (2000) Effect of the fine fraction of particulate matter versus the
49              coarse mass and other pollutants on daily mortality in Santiago, Chile. J. Air Waste Manage. Assoc.
 50              50:  1287-1298.
51       Claiborn, C. S.; Finn, D.; Larson, T. V.; Koemg, J. Q. (2000) Windblown dust contributes to high PM2 5
 52              concentrations. J. Air Waste Manage. Assoc. 50: 1440-1445.
53      Clark, W. E.; Whitby, K. T. (1975) Measurements of aerosols produced by the photochemical oxidation of SO2 in
54             air.  J. Colloid Interface Sci. 51: 477-490.


         March 2001                                    9-107         DRAFT-DO NOT QUOTE OR CITE

-------
  1       Clyde, M. A.; Guttorp, P.; Sullivan, E. (2000) Effects of ambient fine and coarse particles on mortality in Phoenix,
  2               Arizona. J. Exposure Anal. Environ. Epidemiol.: submitted.
  3       Code of Federal Regulations. (199 la) Appendix J to Part 50-reference method for the determination of paniculate
  4               matter as PM10 in the atmosphere. C. F. R. 40: §50.
  5       Code of Federal Regulations. (1991b) Ambient air monitoring reference and equivalent methods. C. F. R. 40: §53.
  6       Costa, D. L.; Dreher, K. L. (1999) What do we need to know about airborne particles to make effective risk
  7               management decisions? A toxicology perspective. Hum. Ecol. Risk Assess. 5: 481-491.
  8       Daniels, M.; Dominici, F.; Samet, J. M.; Zeger, S. L. (2000) Estimating particulate matter-mortality dose-response
  9               curves and threshold levels: an analysis of daily time-series for the 20 largest US cities. Am. J. Epidemiol.
 10               152:397-406.
 11       Dejmek, J.;  Selevan, S. G.; Benes, I.; Solansky, I.; Sram, R. J. (1999) Fetal growth and maternal exposure to
 12               particulate matter during pregnancy. Environ. Health Perspect. 107: 475-480.
 13       Delfino, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M. R. (1997) Effects of air pollution
 14               on  emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Cnt. Care Med.
 15               155:568-576.
 16       Delfino, R. J.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H. (1998) Symptoms in pediatric asthmatics and air pollution:
 17               differences in effects by symptom seventy, anti-inflammatory medication use and particulate averaging
 18               time. Environ. Health Perspect. 106: 751-761.
 19       Dockery, D. W.; Speizer, F. E.; Stram, D. O.; Ware, J. H.; Spengler, J. D.;  Ferris, B. G., Jr. (1989) Effects of
 20               inhalable particles on respiratory health of children. Am. Rev. Respir. Dis. 139: 587-594.
 21       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.
 22               (1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med.
 23               329: 1753-1759.
 24       Dockery, D. W.; Cunningham, J.; Damokosh, A. I.; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
 25               Raizenne, M.; Speizer,  F. E. (1996) Health effects of acid aerosols on North American children:
 26               respiratory symptoms. Environ. Health Perspect. 104: 500-505.
 27       Dockery, D. W.; Pope, C. A., Ill; Kanner, R. E.; Villegas, G. M.; Schwartz, J. (1999) Daily changes in oxygen
 28               saturation and pulse rate associated with particulate air pollution and barometric  pressure. Cambridge, MA:
 29               Health Effects Institute; research report no. 83.
 30       Driscoll, K.  E.; Costa, D. L.; Hatch, G.; Henderson, R.; Oberdorster, G.; Salem,  H.; Schlesinger, R. B. (2000)
 31               Intratracheal instillation as an exposure technique for the evaluation of respiratory tract toxicity: uses and
 32               limitations. Toxicol. Sci. 55: 24-35.
 33       Elder, A. C. P.; Gelein, R.; Finkelstein J. N.; Cox, C.; Oberdorster, G. (2000) Endotoxin priming affects the lung
 34               response to ultrafine particles and ozone in young and old rats. In: Inhalation Toxicology: proceedings of
 35               the third colloquium on particulate air pollution  and human health; June, 1999; Durham, NC. Inhalation
 36               Toxicology 12(suppl. 1): 85-98.
 37       Fairley, D. (1999) Daily mortality and air pollution in Santa Clara County, California: 1989-1996. Environ. Health
 38               Perspect. 107:637-641.
 39       Fanucchi, M. V.; Plopper, C. G.  (1997) Pulmonary developmental responses to toxicants. In: Roth, R. A., ed.
 40               Toxicology of the Respiratory System,  v. 8 of Comprehensive Toxicology. New  York, NY: Pergamon,
 41               p. 203-220.
 42       Fanucchi, M. V.; Wong, V. J.; Hinds, D.; Tarkington, B.  K.; Van Winkle, L. S.;  Evans, M. J.; Plopper, C. G. (2000)
 43               Repeated episodes of exposure to ozone alters postnatal development of distal conducting airways in infant
44               rhesus monkeys. Am. J. Respir. Crit. Care Med.  161: A615.
45       Federal Register. (1987) Revisions to the national ambient air quality standards for particulate matter. F. R.  (July 1)
46               52:24,634-24,669.
47       Fitz-Simons, T. S.; Mathias, M.;  Rizzo, M. (2000) Analyses of 1999 PM data for the PM NAAQS review.
48               U.S. Environmental Protection Agency, Office of Air Quality Planning  and Standards; September 29.
49       Gardner, S. Y.; Lehmann, J. R.; Costa, D. L. (2000) Oil fly ash-induced elevation of plasma fibrinogen levels in
 50              rats. Toxicol. Sci. 56: 175-180.
 51        Gauderman, W. J.; Mcconnell, R.; Gilliland, F.; London,  S.; Thomas, D.; Avol, E.; Vora,  H.; Berhane, K.;
 52               Rappaport, E. B.; Lurmann, F.; Margolis, H. G.; Peters, J. (2000) Association between air pollution and
53              lung function growth in southern California children. Am. J. Respir. Crit. Care Med. 162: 1383-1390.
         March 2001                                    9-108         DRAFT-DO NOT QUOTE OR CITE

-------
  1       Gielen, M. H.; Van Der Zee, S. C.; Van Wijnen, J. H.; Van Steen, C. J.; Brunekreef, B. (1997) Acute effects of
  2               summer air pollution on respiratory health of asthmatic children. Am. J. Respir. Crit. Care Med. 155: 2105-2108.
  3       Gilmour, P. S.; Brown, D. M.; Lindsay, T. G.; Beswick, P. H.; MacNee, W.; Donaldson, K. (1996) Adverse health
  4               effects of PM|0 particles: involvement of iron in generation of hydroxyl radical. Occup. Environ. Med.
  5               53:817-822.
  6       Godleski, J. J.; Sioutas, C.; Katler, M.; Koutrakis, P. (1996) Death from inhalation of concentrated ambient air
  7               particles in animal models of pulmonary disease. Am. J. Respir. Crit. Care Med. 153: A15.
  8       Godleski, J. J.; et al. (1998) Increased cardiac vulnerability during exposure to inhaled environmental particles
  9               [abstract].  Presented at: Health Effects Institute annual meeting; May; Boston, MA. Boston, MA: Health
 10               Effects Institute.
 11       Godleski, J. J.; Verrier, R. L.; Koutrakis, P.; Catalano, P. (2000) Mechanisms  of morbidity and mortality from
 12               exposure to ambient air particles. Cambridge, MA: Health Effects Institute; research report no. 91.
 13       Gold, A.; Litonjua,  J.; Schwartz, M.; Verrier, R.; Milstein, A.; Larson, E.; Lovett, B. (1998) Cardiovascular
 14               vulnerability to particulate pollution. Presented at: 1998 International conference [American Thoracic
 15               Society]; April; Chicago, IL. Am. J. Respir. Crit. Care Med. 157: A261.
 16       Gold, D. R.; Damokosh, A. I.; Pope, C. A., Ill; Dockery, D. W.; McDonnell, W. F.; Serrano, P.; Retama, A.;
 17               Castillejos, M. (1999) Particulate and ozone pollutant effects on the respiratory function of children in
 18               southwest Mexico City. Epidemiology  10: 8-16.
 19       Gold, D. R.; Litonjua, A.; Schwartz, J.; Lovett, E.; Larson, A.; Nearing, B.; Allen, G.; Verrier, M.; Cherry, R.;
 20               Verrier, R. (2000) Ambient pollution and heart rate variability. Circulation 101: 1267-1273.
 21       Goldberg, M.  S.; Bailar, J. C., Ill;  Burnett, R. T.; Brook, J. R.; Tamblyn, R.; Bonvalot, Y.; Ernst, P.; Flegel,  K. M.;
 22               Singh, R. K.; Valois, M.-F. (2000) Identifying subgroups of the general population that may be susceptible
 23               to short-term increases in particulate air pollution: a time-series study in Montreal, Quebec. Cambridge,
 24               MA:  Health Effects Institute; research report 97.  Available:
 25               http://www.healtheffects.org/pubs-research.htm [15 February,  2001].
 26       Goldsmith, C.-A. W.; Hamada, K.; Ning, Y. Y.; Qin, G.; Catalano, P.; Murthy, G. G. K.; Lawrence,  J.; Kobzik, L.
 27               (1999) The effects of environmental aerosols on airway hyperresponsiveness in a murine model of asthma.
 28               Inhalation Toxicol. 11: 981-998.
 29       Gordon, T.; Nadziejko, C.; Chen, L. C.; Schlesinger, R. (2000) Effects of concentrated ambient particles in rats and
 30               hamsters: an exploratory study. Cambridge, MA:  Health Effects Institute; research report no. 93.
 31       Gwynn, R. C.; Burnett, R. T.; Thurston, G. D. (2000) A time-series analysis of acidic particulate matter and daily
 32               mortality and morbidity in the Buffalo,  New York, region. Environ. Health Perspect. 108: 125-133.
 33       Hayano, J.; Sakakibara, Y.; Yamada, M.; Ohte, N.; Fujmami, T.; Yokoyama, K.; Watanabe, Y.; Takata, K. (1990)
 34               Decreased  magnitude of heart rate spectral components in coronary artery disease. Its relation to
 35               angiographic severity. Circulation 81: 1217-1224.
 36       Hofmann, W.; Balashazy, I.; Heistracher, T.; Koblmger, L. (1996) The significance of particle deposition patterns in
 37               bronchial airway bifurcations for extrapolation modeling. Aerosol Sci. Technol. 25: 305-327.
 38       Hoyert, D. L.;  Kochanek, K. D.; Murphy, S. L. (1999) Deaths: final data for 1997. Atlanta, GA: U.S. Centers for
 39               Disease Control and Prevention, National Center  for Health Statistics. (National vital statistics reports:
 40               v. 47, no. 19).
 41       Hsieh, T. H.; Yu, C. P. (1998) Two-phase pulmonary clearance of insoluble particles in mammalian species.
 42               Inhalation Toxicol. 10: 121-130.
 43       Kamens, R.; Jang, M.; Chien, C.-J.; Leach, K. (1999) Aerosol formation from the reaction of a-pinene and ozone
 44              using a gas-phase kinetics-aerosol partitioning model. Environ. Sci. Technol. 33: 1430-1438.
 45       Keywood, M. D.; Ayers, G. P.; Gras, J. L.; Gillett, R. W.; Cohen, D. D.  (1999) Relationships between size
46              segregated mass concentration data and ultrafine particle number concentrations in urban areas.
47              Atrhos. Environ. 33: 2907-2913.
48      Kim, Y. J.; Boatman, J. F.; Gunter, R. L.; Wellman, D. L.;  Wilkison, S.  W. (1993) Vertical distribution of
49              atmospheric aerosol  size distribution over south-central New Mexico. Atmos. Environ. Part A
 50              27: 1351-1362.
 51       Kittelson, D. B. (1998) Engines and nanoparticles: a review. J. Aerosol Sci. 29: 575-588.
 52      Kleiger, R. E.;  Miller, J. P.; Bigger J. T. Jr.; Moss, A. J.; Multicenter Post-infarction Research Group. (1987)
53               Decreased heart rate variability and its association with increased mortality after acute myocardial
54              infarction. Am. J. Cardiol. 59: 256-262.
         March 2001                                     9-109         DRAFT-DO NOT QUOTE OR CITE

-------
  1       Kleinman, M. T.; Leaf, D. A.; Kelly, E.; Caiozzo, V.; Osann, K.; O'Niell, T. (1998) Urban angina in the mountains:
  2               effects of carbon monoxide and mild hypoxemia on subjects with chronic stable angina. Arch. Environ.
  3               Health 53: 388-397.
  4       Klemm, R. J.; Mason, R. M., Jr. (2000) Aerosol research and inhalation epidemiological study (ARIES): air quality
  5               and daily mortality statistical modeling-interim results. J. Air. Waste Manage. Assoc. 50: 1433-1439.
  6       Klemm, R. J.; Mason, R. M., Jr.; Heilig, C. M.; Neas, L. M.; Dockery, D. W. (2000) Is daily mortality associated
  7               specifically with fine particles? Data reconstruction and replication of analyses. J. Air Waste Manage.
  8               Assoc. 50: 1215-1222.
  9       Krewski, D.; Burnett, R.  T.; Goldberg, M. S.; Hoover, K.; Siemiatycki, J.; Jerrett, M.; Abrahamowicz, M.; White,
10               W. H.. (2000) Reanalysis of the Harvard Six Cities study and the American Cancer Society study of
11               particulate air pollution and mortality. A special report of the Institute's Particle Epidemiology Reanalysis
12               Project. Cambridge, MA: Health Effects Institute.
13       Kuschner, W. G.; Wong, H.; D'Alessandro, A.; Quinlan, P.; Blanc, P. D. (1997) Human pulmonary responses to
14               experimental inhalation of high concentration fine and ultrafine magnesium oxide particles. Environ.
15               Health Perspect. 105: 1234-1237.
16       Laden, F.; Neas, L. M.; Dockery, D. W.; Schwartz, J. (2000) Association of fine particulate matter from different
17               sources with daily mortality in six U.S. cities. Environ. Health Perspect. 108: 941-947.
18       Lawrence, L.; Hall, M. J. (1999) 1997 summary: national hospital discharge survey. Hyattsville, MD: U.S.
19               Department of Health &  Human Resources, Centers for Disease Control and Prevention, National Center
20               for Health Statistics. (Advance data  from vital and health statistics: no. 308).
21       Layton, D. W. (1993) Metabolically consistent breathing rates for use in dose assessments. Health Phys. 64: 23-36.
22       Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1997) In vivo and in vitro proinflammatory effects of
23               particulate air pollution (PM10). In: Dnscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth
24               international meeting on  the toxicology of natural and man-made fibrous and non-fibrous particles;
25               September 1996; Lake Placid,  NY. Environ. Health Perspect. Suppl. 105(5): 1279-1283.
26       Liao, D.; Creason, J.; Shy, C.; Williams, R.; Watts, R.; Zweidinger, R. (1999) Daily variation of particulate air
27               pollution and poor cardiac autonomic control in the elderly. Environ. Health Perspect. 107:  521-525.
28       Linn, W. S.;  Gong, H., Jr.; Shamoo, D.  A.; Anderson, K. R.; Avol, E. L. (1997) Chamber exposures of children to
29               mixed ozone, sulfur dioxide, and sulfuric acid. Arch. Environ. Health 52: 179-187.
30       Linn, W. S.;  Szlachcic, Y.;  Gong, H., Jr.; Kinney, P. L.; Berhane, K. T. (2000) Air pollution and daily hospital
31               admissions in metropolitan Los Angeles. Environ. Health Perspect. 108: 427-434.
32       Lipfert, F. W.; Morris, S. C.; Wyzga, R. E. (2000a) Daily mortality in the Philadelphia metropolitan area and
33               size-classified particulate matter. J. Air Waste Manage. Assoc.: 1501-1513.
34       Lipfert, F. W.; Zhang, J.; Wyzga,  R. E.  (2000b) Infant mortality and air pollution: a comprehensive analysis of U.S.
35               data for 1990. J. Air Waste Manage. Assoc. 50: 1350-1366.
36       Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of particulate matter components with daily
37               mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report
38               no. 95.
39       Loomis, D.; Castillejos, M.; Gold, D. R.; McDonnell, W.; Borja-Aburto, V. H. (1999) Air pollution and infant
40               mortality in Mexico City. Epidemiology 10: 118-123.
41       Lumley, T.; Heagerty, P. (1999) Weighted empirical adaptive variance estimators for correlated data regression.
42               J. R. Stat. Soc. B 61 (part 2): 459-477.
43       Lundgren, D. A.; Burton, R. M. (1995)  Effect of particle size distribution on the cut point between fine and coarse
44               ambient mass fractions. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on particulate
45               air pollution and human mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol. 7: 131-148.
46       Mar, T. F.; Norris, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations  between air pollution and mortality in
47               Phoenix, 1995-1997. Environ.  Health Perspect. 108: 347-353.
48       Martin, G. J.; Magid, N. M.; Myers, G.; Barnett, P. S.; Schaad, J. W.; Weiss, J. S.; Lesch, M.; Singer, D. H. (1987)
49               Heart rate variability and sudden death secondary to coronary artery disease during ambulatory
50               electrocardiographic monitoring. Am. J. Cardiol. 60: 86-89.
51       McConnell, R.; Berhane, K.; Gilliland,  F.; London, S.  J.; Vora, H.; Avol, E.; Gauderman, W. J.; Margolis, H. G.;
52               Lurmann, F.; Thomas, D. C.; Peters, J. M. (1999) Air  pollution and bronchitic symptoms in southern
53               California children with asthma. Environ. Health Perspect. 107: 757-760.
54       Michel, O.; Nagy, A.-M.; Schroeven, M.; Duchateau, J.; Neve, J.; Fondu, P.; Sergysels, R. (1997) Dose-response
55               relationship to inhaled endotoxin in  normal subjects. Am. J. Respir. Crit. Care Med. 156: 1157-1164.


         March 2001                                     9-110         DRAFT-DO NOT QUOTE OR CITE

-------
   1      Monn, C.; Becker, S. (1999) Cytotoxicity and induction of proinflammatory cytokines from human monocytes
   2              exposed to fine (PM2 5) and coarse particles (PM10 2 5) in outdoor and indoor air. Toxicol. Appl. Pharmacol.
   3              155:245-252.
   4      Moolgavkar, S. H. (2000a) Air pollution and daily mortality in three U.S. counties. Environ. Health Perspect.
   5              108:777-784.
   6      Moolgavkar, S. H. (2000b) Air pollution and hospital admissions for chronic obstructive pulmonary disease in three
   7              metropolitan areas in the United States. In: Grant, L. D., ed. PM2000: paniculate matter and health.
   8              Inhalation Toxicol. 12(suppl. 4): 75-90.
   9      Moolgavkar, S. H. (2000c) Air pollution and hospital admissions for diseases of the circulatory system in three U.S.
 10              metropolitan areas. J. Air Waste Manage Assoc. 50: 1199-1206.
 11      Moolgavkar, S. H.; Luebeck, E. G.; Anderson, E. L. (1997)  Air pollution and hospital admissions for respiratory
 12              causes in Minneapolis-St. Paul and Birmingham. Epidemiology 8: 364-370.
 13      Moolgavkar, S. H.; Hazelton, W.; Luebeck, G.; Levy, D.; Sheppard, L. (2000) Air pollution, pollens, and
 14              admissions for chronic respiratory disease in King County, Washington. In: Inhalation Toxicology:
 15              proceedings of the third colloquium on particulate air pollution and human health; June, 1999; Durham,
 16              NC. Inhalation Toxicology 12(suppl. 1): 157-171.
 17      Morawska, L.;  Thomas, S.; Bofinger, N.; Wainwright, D.; Neale, D. (1998)  Comprehensive characterization of
 18              aerosols in a subtropical urban atmosphere: particle size distribution and correlation with gaseous
 19              pollutants. Atmos. Environ. 32: 2467-2478.
 20      Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
 21               evidence of an increased effect at low temperatures. Environ. Health Perspect. 106: 649-653.
 22      Morris, R. D.; Naumova, E. N.; Munasinghe, R. L. (1995) Ambient air pollution and hospitalization for congestive
 23               heart failure among elderly people in seven large US cities. Am. J. Public Health 85: 1361-1365.
 24       Naeher, L. P.; Holford, T. R.;  Beckett, W. S.; Belanger, K.; Triche, E. W.; Bracken, M. B.; Leaderer, B. P. (1999)
 25               Healthy women's PEF variations with ambient summer concentrations of PM10, PN2 5, SO42, H+, and O3.
 26               Am. J. Respir. Crit. Care Med. 160:  117-125.
 27       Nakamura, T.; Hayashida, Y. (1992) Autonomic cardiovascular responses to smoke exposure in conscious rats. Am.
 28               J. Physiol. 262(5 pt. 2): R738-745.                                                            .    ,
 29       National Institutes of Health. (1997) Guidelines for the diagnosis and management of asthma: expert panel report 2.
 30               Bethesda, MD:  U.S. Department of Health and Human Services, National Heart, Lung, and Blood
 31               Institute; publication  no. 97-4051.
 32       Nauenberg, E.;  Basu, K. (1999) Effect of insurance coverage on the relationship between asthma hospitalizations
 33               and exposure to air pollution. Public Health Rep. 114: 135-148.
 34       Neas, L. M.; Dockery, D. W.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr.;  Speizer, F.  E. (1994) Concentration of
 35               indoor particulate matter as a determinant of respiratory health in children. Am. J. Epidemiol.
 36               139: 1088-1099.
 37       Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Tollerud, D. J.;  Speizer, F. E. (1995) The association of ambient air
 38              pollution with twice daily peak expiratory flow rate measurements in children. Am. J. Epidemiol
 39               141:111-122.
 40       Neas, L. M.; Dockery, D. W.; Burge, H.; Koutrakis, P.; Speizer,  F. E. (1996) Fungus spores, air pollutants, and
 41              other determinants of peak expiratory flow rate in children. Am.  J. Epidemiol. 143: 797-807.
 42       Neas, L. M.; Schwartz, J.; Dockery, D. (1999) A case-crossover  analysis of air pollution and mortality in
 43              Philadelphia.  Environ. Health Perspect. 107: 629-631.
 44       Nikula, K. J.; Vallyathan, V.; Green, F. H. Y.; Hahn, F. F. (2000) Influence of dose on the distribution of retained
 45              particulate material in rat and human lungs. Presented at: PM2000: particulate matter and health—the
 46              scientific basis for regulatory decision-making, specialty conference & exhibition; January;  Charleston,
 47              SC. Pittsburgh, PA: Air & Waste Management Association.
48      Norris, G.; Young-Pong, S. N.; Koemg, J. Q.; Larson, T. V.;  Sheppard, L.; Stout, J. W. (1999) An association
49              between fine particles and asthma emergency department visits for children in Seattle. Environ. Health
 50              Perspect. 107:489-493.
 51      Norris, G.; Larson, T.;  Koenig, J.; Claiborn, C.; Sheppard, L.; Finn, D. (2000) Asthma aggravation, combustion, and
52              stagnant air. Thorax 55:  466-470.
53      Ohtsuka, Y.; Clarke, R. W.; Mitzner, W.; Brunson, K.; Jakab, G. J.; Kleeberger, S. R. (2000) Interstrain variation in
54              munne susceptibility to inhaled acid-coated particles. Am. J. Physiol. L469-L476.



         March 2001                                    9-111         DRAFT-DO NOT QUOTE OR CITE

-------
  1       Ostro, B. D.; Lipsett, M. J.; Wiener, M. B.; Seiner, J. C. (1991) Asthmatic responses to airborne acid aerosols. Am.
 2               J. Public Health 81: 694-702.
 3       Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Krupnick, A.; Harrington, W. (1993) Air pollution and respiratory
 4               morbidity among adults in Southern California. Am. J. Epidemiol. 137: 691-700.
 5       Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Braxton-Owens, H.; White, M. C. (1995) Air pollution and asthma
 6               exacerbations among African-American children in Los Angeles. In: Phalen, R. F.; Bates, D. V., eds.
 7               Proceedings of the colloquium on particulate air pollution and human mortality and morbidity, part II;
 8               January 1994; Irvine, CA. Inhalation Toxicol. 7:  711-722.
 9       Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coachella
10               Valley, California: a follow-up study. J. Exposure Anal. Environ. Epidemiol. 10: 412-419.
11       Ozkaynak, H.; Xue, J.; Zhou, H.; Raizenne, M. (1996) Associations between daily mortality and motor vehicle
12               pollution in Toronto, Canada. Boston, MA: Harvard University School of Public Health, Department of
13               Environmental Health; March 45.
14       Pekkanen, J.; Timonen, K. L.; Ruuskanen, J.; Reponen, A.; Mirme, A. (1997) Effects of ultrafme and fine particles
15               in urban air on peak expiratory flow among children with asthmatic  symptoms. Environ. Res. 74: 24-33.
16       Perry, K. D.; Cahill, T. A.; Eldred, R. A.; Dutcher, D. D.;  Gill, T. E. (1997) Long-range transport of North African
17               dust to the eastern United States. J. Geophys. Res. [Atmos] 102: 11,225-11,238.
18       Peters, A.; Doring, A.; Wichmann, H.-E.; Koenig, W. (1997a) Increased plasma viscosity during an air pollution
19               episode: a link to mortality? Lancet 349:  1582-1587.
20       Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997b) Respiratory effects are associated with the
21               number of ultrafme particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
22       Peters, A.; Dockery, D. W.; Heinrich, J.; Wichmann, H. E. (1997c) Short-term effects of particulate air pollution on
23               respiratory morbidity in asthmatic children. Eur. Respir. J. 10: 872-879.
24       Peters, J. M.; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
25               Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999a) A study of twelve southern California communities
26               with differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit.
27               Care Med. 159:760-767.
28       Peters, J. M.; Avol, E.; Gauderman, W. J.; Linn, W. S.; Navidi, W.; London,  S. J.; Margolis, H.; Rappaport, E.;
29               Vora,  H.; Gong, H., Jr.; Thomas, D. C. (1999b) A study of twelve southern California communities with
30               differing levels and types of air pollution. II. Effects on pulmonary function. Am. J. Respir. Crit. Care
31               Med. 159:768-775.
32       Plopper, C. G.; Fanucchi, M. V. (2000) Do urban environmental pollutants exacerbate childhood lung diseases?
33               Environ. Health Perspect.  108: A252-A253.
34       Pope, C. A., Ill; Thun, M. J.; Namboodiri, M.  M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
35               (1995) Particulate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am.  J.
36               Respir. Crit. Care Med. 151: 669-674.
37       Pope, C. A., Ill; Hill, R. W.; Villegas, G. M. (1999a) Particulate air pollution and daily mortality on Utah's Wasatch
38               Front. Environ. Health Perspect.: 107: 567-573.
39       Pope, C. A.; Verrier, R. L.; Lovett, E. G.; Larson, A. C.; Raizenne, M. E.; Kanner, R. E.; Schwartz, J.; Villegas,
40               G. M.; Gold, D. R.; Dockery, D. W. (1999b) Heart rate variability associated with particulate air pollution.
41               Am. Heart J. 138: 890-899.
42       Raizenne, M.; Neas, L. M.; Damokosh, A. I.; Dockery, D. W.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer,
43               F. E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ.
44               Health Perspect. 104:  506-514.
45       Ramadan, Z.; Song, X.-H.; Hopke, P. K. (2000) Identification of sources of Phoenix aerosol by positive matrix
46               factorization.  J. Air Waste Manage. Assoc. 50: 1308-1320.
47       Romieu, L; Meneses, F.; Ruiz, S.; Sienra, J. J.; Huerta, J.; White, M. C.; Etzel, R. A. (1996) Effects of air pollution
48               on the respiratory health of asthmatic children living in Mexico City. Am. J. Respir. Crit. Care Med.
49               154:300-307.
50       Romieu, I.; Meneses, F.; Ruiz, S.; Huerta, J.; Sienra, J. J.; White, M.; Etzel, R.;  Hernandez, M. (1997) Effects of
51               intermittent ozone exposure on peak expiratory flow and respiratory symptoms among asthmatic children
52               in Mexico City. Arch. Environ. Health 52: 368-376.
53       Salvi, S.; Blomberg, A.; Rudell, B.; Kelly, F.;  Sandstrom, T.; Holgate, S. T.;  Frew, A. (1999) Acute inflammatory
54               responses in the airways and peripheral blood after short-term exposure to diesel exhaust in healthy human
55               volunteers. Am. J. Respir. Crit. Care Med. 159: 702-709.


         March 2001                                    9-112        DRAFT-DO NOT QUOTE OR CITE

-------
  1       Samet, J. M.; Dominici, F.; Zeger, S. L.; Schwartz, J.; Dockery, D. W. (2000a) National morbidity, mortality, and
  2               air pollution study. Part I: methods and methodologic issues. Cambridge, MA: Health Effects Institute;
  3               research report no. 94.
  4       Samet, J. M.; Zeger, S. L.; Dominici, F.; Curriero, F.; Coursac, I.; Dockery, D. W.; Schwartz, J.; Zanobetti, A.
  5               (2000b) The national morbidity, mortality, and air pollution study. Part II: morbidity, mortality, and air
  6               pollution in the United States. Cambridge, MA: Health Effects Institute; research report no. 94.
  7       Sarnat, J. A.; Koutrakis, P.; Suh, H. H. (2000) Assessing the relationship between personal particulate and gaseous
  8               exposures of senior citizens living in Baltimore, MD. J. Air Waste Manage. Assoc. 50: 1184-1198.
  9       Schlesmger, R. B. (1988) Biological disposition of airborne particles: basic principles and application to vehicular
 10               emissions. In:  Watson, A. Y.; Bates, R. R.; Kennedy, D., eds. Air pollution, the automobile, and public
 11               health. Washington, DC: National Academy Press; pp. 239-298.
 12       Schlesinger, R. B.; Ben-Jebria, A.; Dahl, A. R.; Snipes, M. B.; Ultman, J. (1997) Disposition of inhaled toxicants.
 13               In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press; pp. 493-550.
 14       Schwartz, J. (1997) Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology
 15               8:371-377.
 16       Schwartz, J. (1999) Air pollution and hospital admissions for heart disease in eight U.S. counties. Epidemiology
 17               10: 17-22.
 18       Schwartz, J. (2000) Harvesting and long term exposure effects in the relation between air pollution and mortality.
 19               Am. J. Epidemiol. 151: 440-448.
 20       Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit,
 21               Michigan. Am. J. Epidemiol. 142: 23-35.
 22       Schwartz, J.; Neas, L. M. (2000) Fine particles are more strongly associated than coarse particles with acute
 23               respiratory health effects in schoolchildren. Epidemiology. 11: 6-10.
 24       Schwartz, J.; Zanobetti, A. (2000) Using meta-smoothing to estimate dose-response trends across multiple studies,
 25               with application to air pollution and daily death. Epidemiology 11: 666-672.
 26       Schwartz, J.; Dockery,  D. W.; Neas, L. M.; Wypij, D.; Ware, J. H.; Spengler, J. D.; Koutrakis, P.; Speizer, F. E.;
 27               Ferris, B. G., Jr. (1994) Acute effects of summer air pollution on respiratory symptom reporting in
 28               children. Am.  J. Respir. Crit. Care Med. 150: 1234-1242.
 29       Schwartz, J.; Dockery,  D. W.; Neas, L. M. (1996) Is daily mortality associated specifically with fine particles?
 30               J. Air Waste Manage. Assoc. 46:  927-939.
 31       Schwartz, J.; Norris, G.; Larson, T.; Sheppard, L.; Claiborne, C,; Koenig, J. (1999) Episodes of high coarse particle
 32               concentrations are not associated  with increased mortality. Environ. Health Perspect.  107: 339-342.
 33       Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koemg, J. Q. (1999) Effects of ambient air pollution on
 34               nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
 35       Singer, D. H.; Martin, G. J.; Magid, N.; Weiss, J. S.; Schaad, J. W.; Kehoe, R.; Zheutlin, T; Fintel, D. J.; Hsieh,
 36               A. M.; Lesch,  M. (1988) Low heart rate variability and sudden cardiac death. J. Electrocardiol.
 37               21(suppl.): S46-S55.
 38       Smiley-Jewell, S. M.; Liu, F. J.; Weir, A. J.; Plopper, C. G. (2000) Acute injury to differentiating Clara cells in
 39               neonatal rabbits results in  age-related failure of bronchiolar epithelial repair. Toxicol. Pathol. 28: 267-276.
 40       Smith, R. L.; Spitzner, D.; Kim, Y.; Fuentes, M. (2000) Threshold dependence of mortality effects for fine and
 41                coarse particles in Phoenix, Arizona. J. Air Waste Manage. Assoc. 50: 1367-1379.
 42       Stieb, D. M.; Bevendge, R. C.; Brook, J. R.; Smith-Doiron, M.; Burnett, R. T.; Dales, R. E.; Beaulieu, S.; Judek,  S.;
 43               Mamedov, A. (2000) Air pollution, aeroallergens and cardiorespiratory emergency department visits in
 44               Saint John, Canada. J. Exposure Anal. Environ. Epidemiol. 10: 461-477.
 45       Task Force of the European Society of Cardiology and the North American Society of Pacing and
46               Electrophysiology. (1996) Heart rate variability: standards of measurement, physiological interpretation
47               and clinical use. Circulation 93: 1043-1065.
48        Terashima, T.; Wiggs, B.; English,  D.; Hogg, J. C.; Van Eeden, S. F. (1997) Phagocytosis of small carbon particles
49                (PM]0) by alveolar macrophages stimulates the release of polymorphonuclear leukocytes from bone
 50                marrow. Am. J. Respir.  Crit. Care Med. 155: 1441-1447.
 51       Thurston, G. D.; Ito, K.; Kinney, P. L.; Lippmann, M. (1992) A multi-year study of air pollution and respiratory
 52               hospital admissions in three New York State metropolitan areas: results for 1988 and  1989 summers.
53               J. Exposure Anal. Environ. Epidemiol. 2: 429-450.
         March 2001                                     9-113         DRAFT-DO NOT QUOTE OR CITE

-------
  1       Thurston, G. D.; Ito, K.; Hayes, C. G.; Bates, D. V.; Lippmann, M. (1994) Respiratory hospital admissions and
  2               summertime haze air pollution in Toronto, Ontario: consideration of the role of acid aerosols. Environ.
  3               Res. 65:271-290.
  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       Tiittanen, P.; Timonen, K. L.; Ruuskanen, J.; Mirme, A.; Pekkanen, J. (1999) Fine particulate air pollution,
  7               resuspended road dust and respiratory health among symptomatic children. Eur. Respir. J. 13: 266-273.
  8       Tolbert, P. E.; Klein, M.; Metzger, K. B.; Peel, J.; Flanders, W. D.; Todd, K.; Mulholland, J. A.; Ryan, P. B.;
  9               Frumkin, H. (2000) Interim results of the study of particulates and health in Atlanta (SOPHIA). J.
10               Exposure Anal. Environ. Epidemiol. 10: 446-460.
11       Tsai, F. C.; Apte, M. G.; Daisey, J. M. (2000) An exploratory analysis of the relationship between mortality and the
12               chemical composition of airborne particulate matter. Inhalation Toxicol. 12(suppl.): 121-135.
13       Turpin, B. J. (1999) Options for characterizing organic particulate matter. Environ. Sci. Technol. 33: 76A-79A.
14       U.S. Environmental Protection Agency. (1996) Air quality criteria for particulate matter. Research Triangle Park,
15               NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF.  3v.
16       U.S. Environmental Protection Agency. (2000a) Air quality criteria for carbon monoxide. Research  Triangle Park,
17               NC: National Center for Environmental Assessment; report no. EPA/600/P-99/001 F. Available:
18               www.epa.gov/ncea/co/ [2000, October 6].
19       U.S. Environmental Protection Agency. (2000b) Health assessment document for diesel exhaust [external review
20               draft]. Research Triangle Park, NC: Office of Health and Environmental Assessment, Environmental
21               Criteria and Assessment Office; report nos. EPA/600/8-90/057E.
22       Vedal, S.; Petkau, J.; White, R.; Blair, J. (1998) Acute effects of ambient inhalable particles in asthmatic and
23               nonasthmatic children. Am. J. Respir. Crit. Care Med.  157: 1034-1043.
24       Ware, J. H.; Ferris, B. G., Jr.; Dockery, D. W.; Spengler, J. D.; Stram, D. O.; Speizer, F. E. (1986) Effects of
25               ambient sulfur oxides and suspended particles on respiratory health of preadolescent children. Am. Rev.
26               Respir. Dis. 133:834-842.
27       Watkinson, W. P.; Campen, M. J.; Costa, D. L. (1998) Cardiac arrhythmia induction after exposure  to residual oil
28               fly ash particles in a rodent model of pulmonary hypertension. Toxicol. Sci. 41: 209-216.
29       Watkinson, W. P.; Campen, M. J.; Dreher, K. L.; Su, W.-Y.; Kodavanti, U.P.; Highfill, J.  W.; Costa, D. L. (2000)
30               Thermoregulatory effects following exposure to particulate matter in healthy and
31               cardiopulmonary-compromised rats. J. Therm. Biol.  25: 131-137.
32       Weschler, C. J.; Shields, H. C. (1999) Indoor ozone/terpene reactions as a source of indoor particles. Atmos.
33               Environ. 33:2301-2312.
34       Whitby, K. T. (1978) The physical characteristics of sulfur aerosols. Atmos. Environ. 12:  135-159.
35       Whitby, K. T.; Sverdrup, G. M. (1980) California aerosols: their physical and chemical characteristics. In: Hidy,
36               G. M.; Mueller, P. K.; Grosjean, D.; Appel, B. R.; Wesolowski, J. J., eds. The character and origins of
37               smog aerosols: a digest of results from the California Aerosol Characterization Experiment (ACHEX).
38               New York, NY: John Wiley & Sons, Inc.; pp. 477-517. (Advances in environmental science and
39               technology: v. 9).
40       Wichmann, H.-E.; Spix, C.; Tuch, T.; Wolke, G.; Peters, A.; Heinrich, J.; Kreyling, W. G.; Heyder,  J. (2000) Daily
41               mortality and fine  and ultrafine particles in Erfurt, Germany. Part I: role of particle number and particle
42               mass. Cambridge,  MA: Health Effects Institute; Research Report no. 98.
43       Williams, R.;  Suggs, J.; Zweidinger, R.; Evans, G.; Creason, J.; Kwok, R.; Rodes, C.; Lawless,  P.; Sheldon, L.
44               (2000a) The 1998  Baltimore Particulate Matter Epidemiology-Exposure Study: part 1.  Comparison of
45               ambient, residential outdoor, indoor and apartment particulate matter monitoring. J. Exposure Anal.
46               Environ. Epidemiol. 10:518-532.
47       Williams, R.;  Suggs, J.; Creason, J.; Rodes, C.; Lawless, P.; Kwok, R.; Zweidinger, R.; Sheldon, L.  (2000b) The
48               1998 Baltimore Particulate Matter Epidemiology-Exposure Study: part 2. Personal exposure assessment
49               associated with an elderly study population. J. Exposure Anal. Environ. Epidemiol. 10: 533-543.
50       Williams, R. W.; Creason, J.; Zweidinger, R.; Watts, R. R.; Shy, C. (2000c) Indoor, outdoor, and personal exposure
51               monitoring of particulate air pollution: the Baltimore elderly epidemiology-exposure pilot study.
52               Atmos. Environ. 34: 4193-4204.
53       Wilson, W. E.; Suh, H. H. (1997) Fine particles and coarse particles:  concentration relationships relevant to
54               epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
         March 2001                                    9-114         DRAFT-DO NOT QUOTE OR CITE

-------
1       Woodruff, T. J.; Grille, J.; Schoendorf, K. C. (1997) The relationship between selected causes of postneonatal
2              infant mortality and particulate air pollution in the United States. Environ. Health Perspect. 105: 608-612.
3       Zanobetti, A.; Schwartz, J. (2000) Race, gender, and social status as modifiers of the effects of PM,0 on mortality.
4              J. Occup. Environ. Med. 42: 469-474.
5       Zhang, T.; Huang, C.; Johns, E. J. (1997) Neural regulation of kidney function by the somatosensory system in
6              normotensive and hypertensive rats. Am. J. Physiol. 273: R1749-R1757.
7       Zhang, H.; Triche, E.; Leaderer, B. (2000) Model for the analysis of binary time series of respiratory symptoms.
8              Am. J. Epidemiol. 151: 1206-1215.
9
      March 2001                                   9-115        DRAFT-DO NOT QUOTE OR CITE

-------

-------
                          APPENDIX 9A

  Key Quantitative Estimates of Relative Risk for Particulate Matter-Related
   Health Effects Based on Epidemiologic Studies of North American Cities
   Assessed in the 1996 Particulate Matter Air Quality Criteria Document
March 2001                         9A-1       DRAFT-DO NOT QUOTE OR CITE

-------
         TABLE 9A-1. EFFECT ESTIMATES PER 50-^g/m3 INCREASE
   IN 24-HOUR PM,,, CONCENTRATIONS FROM U.S. AND CANADIAN STUDIES
Study Location
RR (±CI)
Only PM
in Model

RR (±CI) Reported
Other Pollutants PM10 Levels
in Model Mean (Min/Max)f
Increased Total Acute Mortality
Six Cities2
Portage, WI
Boston, MA
Topeka, KS
St. Louis, MO
Kingston/Knoxville, TN
Steubenville, OH
St. Louis, MOC
Kingston, TNC
Chicago, ILh
Chicago, 1L8
Utah Valley, UTb
Birmingham, ALd
Los Angeles, CAr
Increased Hospital Admissions (for
Respiratory Disease
Toronto, Canada'
Tacoma, WA'
New Haven, CTJ
Cleveland, OHk
Spokane, WA1
COPD
Minneapolis, MNn
Birmingham, ALm
Spokane, WA1
Detroit, MI°

1.04(0.98, 1
1.06(1.04,1
0.98 (0.90, 1
1.03(1.00, 1
1.05(1.00, 1
1.05(1.00, 1
1.08(1.01, 1
1.09(0.94, 1
1.04(1.00, 1
1.03(1.02, 1
1.08(1.05,1
1.05(1.01, 1
1.03(1.00, 1
Elderly > 65 years)

1.23(1.02, 1
1.10(1.03,1
1.06(1.00, 1
1.06(1.00, 1
1.08(1.04,1

1.25(1.10, 1
1.13(1.04, 1
1.17(1.08,1
1.10(1.02, 1

.09)
.09)
.05)
.05)
.09)
.08)
.12)
.25)
.08)
.04)
.11)
.10)
.055)


.43){
.17)
.13)
.11)
.14)

.44)
.22)
.27)
.17)
—
— 18 (±11.7)
— 24 (±12.8)
— 27 (±16.1)
— 31 (±16.2)
— 32 (±14.5)
— 46 (±32.3)
1.06(0.98,1.15) 28(1/97)
1.09(0.94,1.26 30(4/67)
— 37 (4/365)
1.02(1.01,1.04) 38(NR/128)
1.19(0.96,1.47) 47(11/297)
— 48(21,80)
1.02(0.99,1.036) 58(15/177)


1.12(0.88, 1.36): 30-39*
1.11(1.02,1.20) 37(14,67)
1.07(1.01,1.14) 41(19,67)
— 43(19,72)
— 46(16,83)

— 36(18,58)
— 45(19,77)
— 46(16,83)
— 48 (22, 82)
March 2001
9A-2
DRAFT-DO NOT QUOTE OR CITE

-------
          TABLE 9A-1 (cont'd). EFFECT ESTIMATES PER 50-^g/m3 INCREASE
     IN 24-HOUR PM1(1 CONCENTRATIONS FROM U.S. AND CANADIAN STUDIES
Study Location
Pneumonia
Minneapolis, MN"
Birmingham, ALm
Spokane, WA'
Detroit, MP
Ischemic HP
Detroit, MP
RR (±C1)
Only PM
in Model
1.08(1.01, 1.15)
1.09(1.03,1.15)
1.06(0.98,1.13)
—

1.02(1.01,1.03)
RR (±CI) Reported
Other Pollutants PM10 Levels
in Model Mean (Min/Max)*
— 36(18,58)
— 45 (19, 77)
— 46(16,83)
1.06(1.02,1.10) 48(22,82)

1.02(1.00,1.03) 48(22,82)
Increased Respiratory Symptoms
Lower Respiratory
Six Cities"
Utah Valley, UT

Utah Valley, UTS
Cough
Denver, CO"
Six Cities"
Utah Valley, UTS
Decrease in Lung Function
Utah Valley, UTr
Utah Valley, UTS
Utah Valley, UT™

2.03(1.36,3.04)
1.28(1.06, 1.56)1
1.01 (0.81, 1.27)"
1.27(1.08, 1.49)

1.09(0.57,2.10)
1.51 (1.12,2.05)
1.29(1.12, 1.48)

55 (24, 86)**
30(10,50)**
29(7,51)*"

Similar RR 30(13,53)
— 46(11/195)

— 76(7/251)

— 22 (0.5/73)
Similar RR 30(13,53)
— 76(7/251)

— 46(11/195)
— 76(7/251)
— 55(1,181)
References
 "Schwartz et al (I996a).
 "Popeetal (1992, 1994)/O,
 cDockeryetal. (1992)/O3.
 'Schwartz (1993).
 "Ito and Thurston (1996)/O,.
 'Kinney et al. (1995)/O3, CO.
 hStyeretal. (1995).
 'Thurston et al. (1994)/O3.
 'Schwartz (1995)/SO2.
 "Schwartz et al (1996b)
'Schwartz (1996)
"Schwartz (1994a)
"Schwartz (1994b)
"Schwartz (1994c).
"Schwartz and Moms (1995)/O3, CO, SO,.
'Schwartz etal (1994)
'Popeetal. (1991).
'Pope and Dockery (1992)
'Schwartz (1994d).
"Pope and Kanner (1993).
 "Ostroetal. (1991).
 1Min/Max 24-h PMIO in parentheses unless noted
  otherwise as standard deviation (±SD), 10 and
  90percentile(lO, 90). NR = not reported
 TChildren
 "Asthmatic children and adults
 "Means of several cities
 "PEFR decrease in mL/s.
 "*FEV, decrease.
 !RR refers to total population, not just >65 years
March 2001
                     9A-3
DRAFT-DO NOT QUOTE OR CITE

-------
  TABLE 9A-2.  EFFECT ESTIMATES PER VARIABLE INCREMENTS IN 24-HOUR
       CONCENTRATIONS OF FINE PARTICLE INDICATORS (PM2 5, SO^, H+)
                        FROM U.S. AND CANADIAN STUDIES
Acute Mortality
Six City3
Portage, WI
Topeka, KS
Boston, MA
St. Louis, MO
Kingston/Knoxville, TN
Steubenville, OH
Indicator

PM25
PM25
PM25
PM25
PM25
PM2,
RR (±CI) per 25 //g/m3
PM Increase

1.030(0.993, 1.071)
1.020(0.951,1.092)
1.056(1.038, 1.0711)
1.028(1.010,1.043)
1.035(1.005,1.066)
1.025(0.998, 1.053)
Reported PM
Levels Mean
(Min/Max)*

11. 2 (±7.8)
12.2 (±7.4)
15. 7 (±9.2)
18.7 (±10.5)
20.8 (±9.6)
29.6 (±21. 9)
Increased Hospitalization
Ontario, Canada6
Ontario, Canadac
NYC/Buffalo, NYd
so;
so;
03
so:
Torontod H+ (Nmol/m3)
so:
PM2,
1.03(1.02, 1.04)
1.03(1.02, 1.04)
1.03(1.02, 1.05)
1.05(1.01, 1.10)
1.16(1.03, 1.30)*
1.12(1.00, 1.24)
1.15(1.02, 1.78)
R = 3.1-8.2
R = 2.0-7.7
NR
28.8 (NR/391)
7.6 (NR, 48.7)
1 8.6 (NR, 66.0)
Increased Respiratory Symptoms
Southern Californiaf
Six Cities8
(Cough)
Six Cities8
(Lower Resp. Symp.)
so:
PM25
PM2 5 Sulfur
H+
PM25
PM2 5 Sulfur
H+
1.48(1.14, 1.91)
1.19(1.01, 1.42)**
1.23(0.95, 1.59)**
1.06(0.87, 1.29)"
1.44(1.15-1.82)**
1.82(1.28-2.59)**
1.05(0.25-1.30)**
R = 2-37
18.0(7.2,37)***
2.5(3.1,61)***
18.1 (0.8,5.9)***
18.0(7.2,37)***
2.5 (0.8, 5.9)***
18.1 (3.1,61)***
Decreased Lung Function
Uniontown, PAC
PM25
PEFR 23.1 (-0.3, 36.9) (per 25 //g/m3)
25/88 (NR/88)
References:
'Schwartz etal.(1996a).
bBurnettetal. (1994).
'Burnett et al. (1995) O3.
"Thurston et al. (1992, 1994).
"Neasetal. (1995).
fOstroetal.(1993).
gSchwartz et al. (1994).
'Min/Max 24-h PM indicator level shown in parentheses unless
otherwise noted as (±SD), 10 and 90 percentile (10,90) or
R = range of values from min-max, no mean value reported.
'Change per 100 nmoles/m3
"Change per 20 //g/m3 for PM2 5; per 5 /ug/m3 for PM2 5 sulfur;
 per 25 nmoles/m3 for H+.
"*50th percentile value (10,90 percentile).
March 2001
   9A-4
DRAFT-DO NOT QUOTE OR CITE

-------
              TABLE 9A-3. EFFECT ESTIMATES PER INCREMENTS3 IN
         ANNUAL MEAN LEVELS OF FINE PARTICLE INDICATORS FROM
                             U.S. AND CANADIAN STUDIES
Type of Health
Effect and Location
Indicator
Increased Total Chronic Mortality in Adults
Six City6


ACS Study'
(151U.S. SMSA)

Increased Bronchitis
Six Cityd
Six Cityc
24 Cityf
24 Cityf
24 Cityf
24 Cityf
Southern California8
PM15/IO
PM25
so:
PM25
so:
in Children
PM15/10
TSP
H+
so:
PM2I
PM10
so:
Change in Health Indicator per
Increment in PMa
Relative Risk (95% CI)
1.42(1.16-2.01)
1.31 (1.11-1.68)
1.46(1.16-2.16)
1.17(1.09-1.26)
1.10(1.06-1.16)
Odds Ratio (95% CI)
3.26(1.13, 10.28)
2.80(1.17,7.03)
2.65(1.22,5.74)
3.02(1.28,7.03)
1.97(0.85,4.51)
3.29(0.81, 13.62)
1.39(0.99, 1.92)
Range of City
PM Levels
Means (,ug/m3)

18-47
11-30
5-13
9-34
4-24

20-59
39-114
6.2-41.0
18.1-67.3
9.1-17.3
22.0-28.6
—
Decreased Lung Function in Children
Six City"
Six Cityc
24 CityIJ
24 City'
24 City1
24 City'
PMI5,10
TSP
H+ (52 nmoles/m3)
PM2,(15^g/m3)
SO: (7 /"g/m3)
PM10(17^g/m3)
NS Changes
NS Changes
-3.45% (-4.87, -2.01) FVC
-3.21% (-4.98, -1.41) FVC
-3. 06% (-4.50, -1.60) FVC
-2. 42% (-4.30, -.0.51) FVC
20-59
39-114
—
—
—
—
 aEstimates calculated annual-average PM increments assume: a 100-/^g/m3 increase for TSP; a 50-^g/m3
  increase for PM]0 and PM]5; a 25-jUg/m3 increase for PM2 5; and a 15-yUg/m3 increase for SO:, except where
  noted otherwise; a 100-nmole/m3 increase for H+.
 "Dockery et al. (1993).
 Tope etal. (1995).
 dDockeryetal.(1989).
 'Wareetal. (1986).
 TJockery etal. (1996).
 gAbbeyetal. (1995).
 hNS Changes = No significant changes.
 'Raizenne etal. (1996).
 ^Pollutant data same as for Dockery et al. (1996).
March 2001
9A-5
DRAFT-DO NOT QUOTE OR CITE

-------
REFERENCES

Abbey, D. E.; Ostro, B. E.; Petersen, F.; Burchette, R. J. (1995) Chronic respiratory symptoms associated with
      estimated long-term ambient concentrations of fine participates less than 2.5 microns in aerodynamic
      diameter (PM25) and other air pollutants. J. Exposure Anal. Environ. Epidemiol. 5: 137-159.
Burnett, R. T.; Dales, R. E.; Raizenne, M. E.; Krewski, D.;  Summers, P. W.; Roberts, G. R.; Raad-Young, M;
      Dann, T.; Brook, J. (1994) Effects of low ambient levels of ozone and sulfates on the frequency of
      respiratory admissions to Ontario hospitals. Environ. Res. 65: 172-194.
Burnett, R. T.; Dales, R.; Krewski, D.; Vincent, R.; Dann, T.; Brook, J.  R. (1995) Associations between ambient
      particulate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. Am. J. Epidemiol.
      142: 15-22.
Dockery, D. W.; Speizer, F. E.; Stram, D. O.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr.  (1989) Effects of
      inhalable particles on respiratory health of children. Am. Rev. Respir. Dis. 139: 587-594.
Dockery, D. W.; Schwartz, J.; Spengler, J. D. (1992) Air pollution and daily mortality: associations with particulates
      and acid aerosols. Environ. Res. 59: 362-373.
Dockery, D. W.; Pope, C.  A., Ill; Xu, X.; Spengler, J. D.; Ware, J. H.; Fay, M. E.; Ferris, B. G., Jr.; Speizer, F. E.
      (1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med.
      329: 1753-1759.
Dockery, D. W.; Cunningham, J.; Damokosh, A. I.; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
      Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
      symptoms. Environ. Health Perspect. 104: 500-505.
Ito, K.; Thurston, G. D. (1996) Daily PM,(/mortality associations: an investigation of at-risk subpopulations. J.
      Exposure Anal. Environ. Epidemiol. 6: 79-95.
Kinney, P. L.; Ito, K.; Thurston, G. D. (1995) A sensitivity analysis of mortality/PM10 associations in Los Angeles.
      In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on particulate air pollution and human
      mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol. 7: 59-69.
Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Tollerud, D. J.; Speizer, F.  E. (1995) The association of ambient air
      pollution with twice daily peak expiratory flow rate measurements in children. Am. J. Epidemiol.
      141: 111-122.
Ostro, B. D.; Lipsett, M. J.; Wiener, M. B.; Seiner, J. C. (1991) Asthmatic responses to airborne acid aerosols. Am.
      J. Public Health 81: 694-702.
Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Krupnick, A.; Harrington, W. (1993) Air pollution and respiratory
      morbidity among adults in Southern California. Am.  J. Epidemiol. 137: 691-700.
Pope, C. A., Ill; Dockery, D. W. (1992)  Acute health effects of PM10 pollution on symptomatic and asymptomatic
      children. Am. Rev.  Respir. Dis. 145: 1123-1128.
Pope, C. A., Ill; Kanner, R. E. (1993) Acute effects of PMIO pollution on pulmonary function of smokers with mild
      to moderate chronic obstructive pulmonary disease. Am. Rev. Respir. Dis.  147: 1336-1340.
Pope, C. A., Ill; Dockery, D. W.; Spengler, J. D.; Raizenne, M. E. (1991) Respiratory health and PM,0 pollution:
      a daily time series analysis. Am. Rev. Respir. Dis. 144: 668-674.
Pope, C. A., Ill; Schwartz, J.; Ransom, M. R. (1992) Daily mortality and PM10 pollution in Utah valley.
      Arch. Environ. Health 47: 211-217.
Pope, C. A., Ill; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
      (1995) Particulate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am. J.
      Respir. Crit. Care Med. 151: 669-674.
Raizenne, M.; Neas, L. M.; Damokosh, A. I.; Dockery, D. W.;  Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer,
      F. E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ.
      Health Perspect. 104: 506-514.
Schwartz, J. (1993) Air pollution and daily mortality in Birmingham, Alabama.  Am. J. Epidemiol.  137:  1136-1147.
Schwartz, J. (1994a) Air pollution and hospital admissions for  the elderly in Birmingham, Alabama. Am. J.
      Epidemiol. 139:589-598.
Schwartz, J. (1994b) PM,0, ozone, and hospital admissions for the elderly in Minneapolis, MN. Arch.  Environ.
      Health 49: 366-374.
Schwartz, J. (1994c) Air pollution and hospital admissions for  the elderly in Detroit, Michigan. Am. J. Respir. Crit.
      Care Med. 150:648-655.


March 2001                                    9A-6        DRAFT-DO NOT QUOTE OR CITE

-------
Schwartz, J. (1994d) Nonparametric smoothing in the analysis of air pollution and respiratory illness. Can. J. Stat.
      22: 1-17.
Schwartz, J. (1995) Short term fluctuations in air pollution and hospital admissions of the elderly for respiratory
      disease. Thorax 50: 531-538.
Schwartz, J. (1996) Air pollution and hospital admissions for respiratory disease. Epidemiology 7: 20-28.
Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit,
      Michigan. Am. J. Epidemiol. 142: 23-35.
Schwartz, J.; Dockery, D. W.; Neas, L. M.; Wypij, D.; Ware, J. H.; Spengler, J. D.; Koutrakis, P.; Speizer, F. E.;
      Ferris, B. G., Jr. (1994) Acute effects of summer air pollution on respiratory symptom reporting in children.
      Am. J. Respir. Crit. Care Med. 150: 1234-1242.
Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996a) Is daily mortality associated specifically with fine particles?
      J. Air Waste Manage. Assoc. 46: 927-939.
Schwartz, J.; Spix, C.; Touloumi, G.; Bacharova, L.; Barumamdzadeh, T.; le Tertre, A.; Piekarksi, T.;
      Ponce de Leon, A.; Ponka, A.; Rossi, G.; Saez, M.; Schouten, J. P. (1996b) Methodological issues in  studies
      of air pollution and daily counts of deaths or hospital admissions. In: St Leger, S., ed. The APHEA project.
      Short term effects of air pollution on health: a European approach using epidemiological time series data. J.
      Epidemiol. Community Health 50(suppl. 1): S3-S11.
Styer, P.; McMillan, N.; Gao, F.; Davis, J.; Sacks, J. (1995) Effect of outdoor airborne particulate matter on daily
      death counts. Environ. Health Perspect.  103: 490-497.
Thurston, G. D.; Ito, K.; Kinney, P. L.; Lippmann, M. (1992) A multi-year study of air pollution and respiratory
      hospital admissions in three New York State metropolitan areas: results for 1988 and 1989 summers.
      J. Exposure Anal. Environ. Epidemiol. 2: 429-450.
Thurston, G. D.; Ito, K.; Hayes, C. G.; Bates, D. V.; Lippmann, M. (1994)  Respiratory hospital admissions and
      summertime haze air pollution in Toronto, Ontario: consideration of the role of acid aerosols. Environ. Res.
      65:271-290.
Ware, J. H.; Ferris, B. G., Jr.; Dockery, D. W.; Spengler, J. D.; Stram, D. O.; Speizer, F. E. (1986) Effects of
      ambient sulfur oxides and suspended particles on respiratory health of preadolescent children. Am. Rev.
      Respir. Dis. 133:834-842.
March 2001                                    9A-7         DRAFT-DO NOT QUOTE OR CITE

-------

-------
 i                          EXECUTIVE SUMMARY
 2
 3                Air Quality Criteria for Participate Matter
 4                                  (March 2000)
 5
 6
 7         The Executive Summary of this Second External Review Draft of the Air Quality Criteria
 8     for Particulate Matter is under preparation and will be completed following public comment and
 9     CAS AC review of the earlier, more detailed chapters of this Second External Review Draft.
10     It will be included in subsequent drafts of the document.
      March 2001                           E-1        DRAFT-DO NOT QUOTE OR CITE

-------