EPA/600/R-17/451 j December 2017 | vwvw.epa.gov/isa
United States
Environmental Protection
Agency
Integrated Science Assessment
for Sulfur Oxides -
Health Criteria
¦ MB) J
V j ¦
Office of Research and Development
National Center for Environmental Assessment, RTP Division
-------
United States
Environmental Protection
tl #* Agency
EPA/600/R-17/451
December 2017
https://www.epa. sov/isa
Integrated Science Assessment
for Sulfur Oxides—Health Criteria
December 2017
National Center for Environmental Assessment—RTP Division
Office of Research and Development
U.S. Enviromnental Protection Agency
Research Triangle Park, NC
-------
DISCLAIMER
This document has been reviewed in accordance with the U.S. Environmental Protection Agency policy
and approved for publication. Mention of trade names or commercial products does not constitute
endorsement or recommendation for use.
ii
-------
CONTENTS
INTEGRATED SCIENCE ASSESSMENT TEAM FOR SULFUR OXIDES—HEALTH
CRITERIA XVII
AUTHORS, CONTRIBUTORS, AND REVIEWERS XX
CLEAN AIR SCIENTIFIC ADVISORY COMMITTEE SULFUR OXIDES NAAQS REVIEW
PANEL XXV
ACRONYMS AND ABBREVIATIONS XXVII
PREFACE XXXIV
Table I History of the primary National Ambient Air Quality Standards
for sulfur oxides since 1971 xxxvii
EXECUTIVE SUMMARY XLI
Table ES-1 Causal determinations for relationships between sulfur
dioxide exposure and health effects from the 2008 and
2017 Integrated Science Assessment for Sulfur Oxides xlviii
CHAPTER 1 INTEGRATIVE SYNTHESIS OF THE ISA 1-1
1.1 Purpose and Overview of the Integrated Science Assessment 1-1
1.2 Process for Developing The Integrated Science Assessment 1-3
1.3 Organization of the Integrated Science Assessment 1-6
1.4 From Emissions Sources to Exposure to Sulfur Dioxide 1-6
1.4.1 Emission Sources and Distribution of Ambient Concentrations 1-7
1.4.2 Assessment of Human Exposure 1-9
1.5 Dosimetry and Mode of Action of Sulfur Dioxide 1-12
1.5.1 Dosimetry of Inhaled Sulfur Dioxide 1-13
1.5.2 Mode of Action of Inhaled Sulfur Dioxide 1-14
1.6 Health Effects of Sulfur Dioxide 1-16
1.6.1 Respiratory Effects 1-17
1.6.1.1 Respiratory Effects Associated with Short-Term Exposure to Sulfur Dioxide 1-17
1.6.1.2 Respiratory Effects Associated with Long-Term Exposure to Sulfur Dioxide 1-19
1.6.2 Health Effects beyond the Respiratory System 1-19
1.6.2.1 Cardiovascular Effects Associated with Short-Term Exposure to Sulfur
Dioxide 1-19
1.6.2.2 Cardiovascular Effects Associated with Long-Term Exposure to Sulfur
Dioxide 1-20
1.6.2.3 Reproductive and Developmental Effects 1-20
1.6.2.4 Total Mortality Associated with Short-Term Exposure to Sulfur Dioxide 1-21
1.6.2.5 Total Mortality Associated with Long-Term Exposure to Sulfur Dioxide 1-21
1.6.2.6 Cancer 1-22
Table 1-1 Key evidence contributing to causal determinations for
sulfur dioxide exposure and health effects evaluated in
the current Integrated Science Assessment for Sulfur
Oxides 1-23
1.7 Policy-Relevant Considerations 1-27
1.7.1 Durations and Lag Structure of Sulfur Dioxide Exposure Associated with Health
Effects 1-27
iii
-------
CONTENTS (Continued)
1.7.2 Concentration-Response Relationships and Thresholds 1-27
1.7.3 Regional Heterogeneity in Effect Estimates 1-28
1.7.4 Public Health Significance 1-29
1.7.4.1 Characterizing Adversity of Health Effects 1-29
1.7.4.2 At-Risk Populations and Lifestages for Health Effects Related to Sulfur
Dioxide Exposure 1-30
1.7.4.3 Summary of Public Health Significance of Health Effects Related to Sulfur
Dioxide Exposure 1-31
1.8 Summary and Health Effects Conclusions 1-31
CHAPTER 2 ATMOSPHERIC CHEMISTRY AND AMBIENT CONCENTRATIONS OF
SULFUR DIOXIDE AND OTHER SULFUR OXIDES 2-1
2.1 Introduction 2-1
2.2 Anthropogenic and Natural Sources of Sulfur Dioxide 2-2
2.2.1 U.S. Anthropogenic Sources 2-2
Figure 2-1 Sulfur dioxide emissions by sector in tons, 2014 2-3
Figure 2-2 Distribution of electric power generating units (438),
scaled according to annual sulfur dioxide emissions
across the U.S. based on the 2014 National Emissions
Inventory 2-4
2.2.2 National Geographic Distribution of Large Sources 2-6
Figure 2-3 Geographic distribution of (top) continental U.S.
facilities (526) emitting more than 1,000 tpy SO2, with
(bottom) an enlargement of the midwestern states,
including the Ohio River Valley, where a large number
of these sources are concentrated 2-7
Figure 2-4 Sulfur dioxide sources (377) identified by state/local air
agencies under the U.S. Environmental Protection
Agency's Data Requirements Rule, as of July 18, 2016 2-8
2.2.3 U.S. Anthropogenic Emission Trends 2-9
Table 2-1 Summary of current U.S. Environmental Protection
Agency sulfur dioxide trends data by emissions sector.
Values shown in bold indicate increased emissions 2-10
Figure 2-5 National sulfur dioxide emissions trends by sector (103
tpy), 1970-2014 2-11
2.2.4 Natural Sources 2-11
2.2.4.1 The Global Sulfur Cycle 2-12
2.2.4.2 Volcanoes as a Natural Source of Sulfur Dioxide 2-12
Figure 2-6 Sulfur dioxide released during the July 12-20, 2008
eruption of the Okmok Volcano in Alaska's Aleutian
Islands (image derived from data collected by the
Atmospheric Infrared Sounder instrument aboard the
National Aeronautics and Space Administration Aqua
satellite) 2-13
Figure 2-7 Geographic location of volcanoes and other potentially
active volcanic areas within the continental U.S 2-14
Figure 2-8 National Aeronautics and Space Administration/Ozone
Monitoring Instrument image of the KTIauea sulfur
dioxide plume during its March 20-27, 2008 eruption 2-15
2.2.4.3 Wildfires as a Natural Source of Sulfur Dioxide 2-16
2.2.5 Reduced Sulfur Compounds as Indirect Sources of Sulfur Dioxide 2-16
Table 2-2 Largest global sulfide emissions sources, ranked
according to total sulfur emissions (103 Tpy) 2-17
iv
-------
CONTENTS (Continued)
2.3 Atmospheric Chemistry and Fate of Sulfur Dioxide and Other Sulfur Oxides 2-18
2.3.1 Photochemical Removal of Atmospheric Sulfur Dioxide 2-19
2.3.2 Heterogeneous Oxidation of Sulfur Dioxide 2-21
Figure 2-9 The effect of pH on the rates of aqueous-phase sulfur
(IV) oxidation by various oxidants 2-22
2.3.3 Secondary Gas-phase and Particle-phase Sulfur Oxides 2-23
2.4 Measurement Methods 2-24
2.4.1 Federal Reference and Equivalent Methods 2-24
2.4.1.1 Minimum Performance Specifications 2-26
Table 2-3 Minimum performance specifications for sulfur dioxide
established in 40 Code of Federal Regulations Part 53,
Subpart B 2-27
2.4.1.2 Positive and Negative Interferences 2-27
2.4.2 Alternative Sulfur Dioxide Measurements 2-29
2.4.3 Ambient Sampling Network Design 2-31
Figure 2-10 Routinely operating sulfur dioxide monitoring networks:
National Core and State and Local Air Monitoring Sites,
reporting 1 hour and 5 minute sulfur dioxide
concentration data. Maps for the Boston, MA, New York
City, and Pittsburgh, PA metropolitan areas are
provided as examples of the variation in monitor
placement in the US Northeast 2-32
2.5 Environmental Concentrations 2-34
2.5.1 Sulfur Dioxide Metrics and Averaging Time 2-34
Table 2-4 Summary of sulfur dioxide metrics and averaging times 2-35
2.5.2 Spatial Variability 2-35
Table 2-5 Summary of sulfur dioxide data sets originating from the
Air Quality System database 2-36
2.5.2.1 Nationwide Spatial Variability 2-36
Table 2-6 National statistics of sulfur dioxide concentrations (parts
per billion) from Air Quality System monitoring sites,
2013-2015 2-38
Figure 2-11 Map of 99th percentile of 1-hour daily max sulfur
dioxide concentration reported at Air Quality System
monitoring sites, 2013-2015 2-39
Figure 2-12 Map of 99th percentile of 24-hour avg sulfur dioxide
concentration reported at Air Quality System monitoring
sites, 2013-2015 2-40
2.5.2.2 Urban Spatial Variability 2-40
Figure 2-13 Map of the Cleveland, OH focus area showing
emissions from large sources and the 99th percentile
5-minute hourly max concentration at ambient monitors
during 2013-2015 2-42
Table 2-7 Largest SO2 emissions sources, Cleveland, OH (as
noted in Figure 2-13) 2-43
Figure 2-14 Map of the Pittsburgh, PA focus area showing
emissions from large sources and the 99th percentile
5-minute hourly max concentration at ambient monitors
during 2013-2015 2-43
Table 2-8 Largest SO2 emissions sources, Pittsburgh, PA (as
noted in Figure 2-14) 2-44
Figure 2-15 Map of the New York City, NY focus area showing
emissions from large sources and the 99th percentile
v
-------
CONTENTS (Continued)
5-minute hourly max concentration at ambient monitors
during 2013-2015 2-45
Table 2-9 Largest SO2 emissions source, New York, NY (as noted
in Figure 2-15) 2-46
Figure 2-16 Map of the St Louis, MO-IL focus area showing
emissions from large sources and the 99th percentile
5-minute hourly max concentration at ambient monitors
during 2013-2015 2-46
Table 2-10 Largest SO2 emissions sources, St. Louis, MO-IL (as
noted in Figure 2-16) 2-47
Figure 2-17 Map of the Houston, TX focus area showing emissions
from large sources and the 99th percentile 5-minute
hourly max concentration at ambient monitors during
2013-2015 2-48
Table 2-11 Largest SO2 emissions source, Houston, TX (as noted
in Figure 2-17) 2-48
Figure 2-18 Map of the Gila County, AZ focus area showing
emissions from large sources and the 99th percentile
5-minute hourly max concentration at ambient monitors
during 2013-2015 2-49
Table 2-12 Largest SO2 emissions sources, Gila County, AZ (as
noted in Figure 2-18) 2-50
Table 2-13 1-h daily max sulfur dioxide concentration distribution
by Air Quality System monitoring site in six focus areas,
2013-2015 2-50
Table 2-14 5-minute hourly max sulfur dioxide concentrations by
Air Quality System monitoring sites in select focus
areas, 2013-2015 2-53
Figure 2-19 Pairwise correlations of 24-hour avg sulfur dioxide
versus distance between monitoring site pairs in six
focus areas. 2013 2015 2-57
Figure 2-20 Pairwise correlations of 5-minute hourly max data
versus distance between monitoring sites in six focus
areas. 2013 2015 2-58
2.5.3 Temporal Variability 2-59
2.5.3.1 Long-Term Trends 2-59
Figure 2-21 National sulfur dioxide air quality trend, based on the
99th percentile of the 1-hour daily max concentration for
163 sites, 1980-2015. A 76% decrease in the national
average was observed from 1990 2015 2-60
2.5.3.2 Seasonal Trends 2-61
Figure 2-22 Sulfur dioxide month-to-month variability based on
1-hour daily max concentrations at Air Quality System
sites in each focus area. 2013 2015 2-62
2.5.3.3 Diel Variability 2-63
Figure 2-23 Diel variability based on 1-hour avg sulfur dioxide
concentrations in the six focus areas, 2013-2015 2-65
Figure 2-24 Diel trend based on 5-minute hourly max data in the six
focus areas. 2013 2015 2-66
Figure 2-25 Diel trend based on 5-minute hourly max data in the
Cleveland, OH and Gila County, AZ focus areas during
winter and summer. 2013 2015 2-68
2.5.4 Relationships between Hourly Mean and Peak Concentrations 2-69
vi
-------
CONTENTS (Continued)
Figure 2-26 Scatterplot of 5-minute hourly max versus 1-hour avg
sulfur dioxide concentrations, 2013-2015 2-70
Figure 2-27 Scatterplot of 5-minute hourly max versus 1-hour avg
sulfur dioxide concentrations by focus area,
2013-2015 2-71
Table 2-15 Pearson correlation coefficient comparing 1-hour avg
with 5-minute hourly max and peak-to-mean ratio for
maximum sulfur dioxide concentrations in the six focus
areas, 2013-2015 2-72
2.5.5 Background Concentrations 2-73
Figure 2-28 1-hour daily max sulfur dioxide concentrations
measured at (A) Hilo, HI and (B) Pahala, HI 2-74
Figure 2-29 Average 24-hour ambient sulfur dioxide concentrations
during low and high (volcanic gas) concentration study
periods (November 26, 2007 to June 6, 2008) for Ka'u
District, located downwind of KTIauea Volcano 2-75
2.6 Atmospheric Modeling 2-75
2.6.1 Dispersion Modeling 2-76
2.6.2 Chemical Transport Models 2-82
2.7 Summary 2-84
CHAPTER 3 EXPOSURE TO AMBIENT SULFUR DIOXIDE 3-1
3.1 Introduction 3-1
3.2 Conceptual Overview of Human Exposure 3-1
3.2.1 Exposure Terminology 3-1
3.2.2 Conceptual Model of Personal Exposure 3-3
3.2.3 Exposure Considerations Specific to Sulfur Dioxide 3-5
3.3 Methodological Considerations for Use of Exposure Data 3-6
3.3.1 Measurements 3-6
3.3.1.1 Fixed-Site Monitoring 3-6
3.3.1.2 Personal and Microenvironmental Monitoring Techniques 3-7
3.3.2 Modeling 3-8
Table 3-1 Comparison of models used for estimating exposure
concentration or exposure. Factors available in each
model are checked 3-9
3.3.2.1 Source Proximity Models 3-9
3.3.2.2 Land Use Regression Models 3-11
3.3.2.3 Inverse Distance Weighting 3-14
3.3.2.4 Dispersion Models 3-15
3.3.2.5 Chemical Transport Models 3-16
3.3.2.6 Microenvironmental Exposure Models 3-16
3.3.3 Choice of Exposure Surrogates in Epidemiologic Studies 3-18
Table 3-2 Summary of exposure assignment methods, their
typical uses in sulfurdioxide epidemiologic studies,
strengths, limitations, related errors, and uncertainties 3-19
3.4 Exposure Assessment, Error, and Implications for Epidemiologic Inference 3-25
3.4.1 Relationships between Personal Exposure and Ambient Concentration 3-26
3.4.1.1 Parameters Influencing Infiltration Factors 3-26
3.4.1.2 Indoor-Outdoor Relationships 3-29
3.4.1.3 Personal-Ambient Relationships 3-30
3.4.2 Factors Contributing to Error in Estimating Exposure to Ambient Sulfur Dioxide 3-30
3.4.2.1 Activity Patterns 3-31
vii
-------
CONTENTS (Continued)
Table 3-3 Mean fraction of time spent in outdoor locations by
various age groups in the National Human Activity
Pattern Survey study 3-32
3.4.2.2 Spatial Variability 3-35
Figure 3-1 Map of the Cleveland, OH core-based statistical area
including National Emissions Inventory facility locations,
urban sulfur dioxide monitor locations, and distance to
each facility with respect to core-based statistical area
block group population density estimates for 2011.
National Emissions Inventory facility emissions ranged
from 1,942 tons/year to 48,300 tons/year 3-37
Table 3-4 2011 American Community Survey population
estimates of people living within a specified distance of
an urban sulfur dioxide monitor in the Cleveland, OH
core-based statistical area. Population estimates are
based on census block group estimates 3-38
Figure 3-2 Map of the Pittsburgh, PA core-based statistical area
including National Emissions Inventory facility locations,
urban sulfur dioxide monitor locations, and distance to
each facility with respect to core-based statistical area
block group population density estimates for 2011.
National Emissions Inventory facility emissions ranged
from 1,279 tons/year to 46,467 tons/year 3-39
Table 3-5 2011 American Community Survey population
estimates of people living within a specified distance of
an urban sulfur dioxide monitor in the Pittsburgh, PA
core-based statistical area. Population estimates are
based on census block group estimates 3-40
3.4.2.3 Temporal Variability 3-41
Figure 3-3 Pearson correlations between 1-h avg and 5-minute
hourly max, 24-h avg and 1-h daily max, and 24-h avg
and 5-minute daily max sulfur dioxide concentrations 3-42
3.4.2.4 Method Detection Limit, Instrument Accuracy, and Instrument Precision 3-43
3.4.3 Copollutant Relationships 3-44
3.4.3.1 Temporal Relationships among Ambient Sulfur Dioxide and Copollutant
Exposures 3-46
Figure 3-4 Distribution of Pearson correlation coefficients for
comparison of 24-h avg sulfur dioxide concentration
from the year-round data set with collocated National
Ambient Air Quality Standards pollutants (and sulfur in
PM2.5) from Air Quality System during 2013-2015 3-47
Figure 3-5 Distribution of Pearson correlation coefficients for
comparison of daily 1-h max sulfur dioxide
concentration from the year-round data set with
collocated National Ambient Air Quality Standards
pollutants (and sulfur in PM2.5) from Air Quality System
during 2013-2015 3-48
Figure 3-6 Distribution of Pearson correlation coefficients for
comparison of daily 24-h avg sulfur dioxide ambient
concentration stratified by season with collocated
National Ambient Air Quality Standards pollutants (and
PM2.5) from Air Quality System during 2013-2015 3-49
Figure 3-7 Distribution of Pearson correlation coefficients for
comparison of daily 1-h max sulfur dioxide ambient
concentration stratified by season with collocated
National Ambient Air Quality Standards pollutants (and
PM2.5) from Air Quality System during 2013-2015 3-50
viii
-------
CONTENTS (Continued)
Figure 3-8 Summary of temporal sulfur dioxide-copollutant
correlation coefficients from measurements reported in
the literature, sorted by temporal averaging period 3-53
3.4.3.2 Spatial Relationships among Ambient Sulfur Dioxide and Copollutants 3-54
3.4.4 Implications for Epidemiologic Studies of Different Designs 3-54
3.4.4.1 Community Time-Series Studies 3-56
3.4.4.2 Long-Term Cohort Studies 3-60
3.4.4.3 Panel Studies 3-63
3.5 Summary and Conclusions 3-64
CHAPTER 4 DOSIMETRY AND MODE OF ACTION 4-1
4.1 Introduction 4-1
4.1.1 Structure and Function of the Respiratory Tract 4-1
Figure 4-1 Diagrammatic representation of respiratory tract regions
in humans 4-2
4.1.2 Breathing Rates and Breathing Habit 4-3
4.1.2.1 Breathing Rates 4-3
Table 4-1 Ventilation rates in humans as a function of activity 4-3
4.1.2.2 Breathing Habit 4-4
4.2 Dosimetry of Inhaled Sulfur Dioxide 4-7
4.2.1 Chemistry 4-8
4.2.2 Absorption 4-9
4.2.3 Distribution 4-11
4.2.4 Metabolism 4-13
4.2.5 Elimination 4-14
4.2.6 Sources and Levels of Exogenous and Endogenous Sulfite 4-15
4.3 Mode of Action of Inhaled Sulfur Dioxide 4-17
4.3.1 Activation of Sensory Nerves in the Respiratory Tract 4-18
4.3.2 Injury to Airway Mucosa 4-22
4.3.3 Modulation of Airway Responsiveness and Allergic Inflammation 4-23
4.3.4 Induction of Systemic Effects 4-28
4.3.5 Role of Endogenous Sulfur Dioxide/Sulfite 4-30
4.3.6 Mode of Action Framework 4-31
Figure 4-2 Summary of evidence for the proposed mode of action
linking short-term exposure to sulfur dioxide and
respiratory effects 4-31
Figure 4-3 Summary of evidence for the proposed mode of action
linking long-term exposure to sulfur dioxide and
respiratory effects 4-34
Figure 4-4 Summary of evidence for the proposed mode of action
linking exposure to sulfur dioxide and extrapulmonary
effects 4-36
CHAPTER 5 INTEGRATED HEALTH EFFECTS OF EXPOSURE TO SULFUR OXIDES 5-1
5.1 Introduction 5-1
5.1.1 Scope of the Chapter 5-1
5.1.2 Evidence Evaluation and Integration to Form Causal Determinations 5-2
5.1.2.1 Evaluation of Individual Studies 5-2
5.1.2.2 Integration of Scientific Evidence 5-3
5.1.3 Summary 5-4
5.2 Respiratory Effects 5-5
5.2.1 Short-Term Exposure 5-5
ix
-------
CONTENTS (Continued)
5.2.1.1 Introduction 5-5
5.2.1.2 Asthma Exacerbation 5-6
Table 5-1 Study-specific details from controlled human exposure
studies of individuals with asthma 5-9
Table 5-2 Percentage of adults with asthma in controlled human
exposure studies experiencing sulfur dioxide-induced
decrements in lung function and respiratory symptoms 5-16
Figure 5-1 Distribution of individual airway sensitivity to sulfur
dioxide. The cumulative percentage of subjects is
plotted as a function of provocative concentration,
which is the concentration of sulfur dioxide that
provoked a 100% increase in specific airway resistance
compared to clean air 5-19
Table 5-3 Percent change in post- versus pre-exposure measures
of forced expiratory volume in 1 second relative to clean
air control after 5-10-minute exposures to sulfur dioxide
during exercise 5-20
Table 5-4 Percent change in post- versus pre-exposure measures
of specific airway resistance relative to clean air control
after 5-10-minute exposures to sulfur dioxide during
exercise 5-21
Table 5-5 Epidemiologic studies of lung function in adults with
asthma published since the 2008 ISA for Sulfur Oxides 5-27
Table 5-6 Epidemiologic studies of lung function in children with
asthma published since the 2008 ISA for Sulfur Oxides 5-32
Table 5-7 Study-specific details from controlled human exposure
studies of respiratory symptoms 5-38
Table 5-8 Epidemiologic studies of respiratory symptoms in
populations with asthma published since the 2008 ISA
for Sulfur Oxides 5-43
Figure 5-2 Associations between short-term average ambient
sulfur dioxide concentrations and respiratory symptoms
and asthma medication use in children with asthma 5-47
Figure 5-3 Percent increase in asthma hospital admissions and
emergency department visits from U.S. and Canadian
studies evaluated in the 2008 SOx ISA and recent
studies in all-year and seasonal analyses for a 10-ppb
increase in 24-h avg or40-ppb increase in 1-h max
sulfur dioxide concentrations 5-51
Table 5-9 Study-specific details and mean and upper percentile
concentrations from asthma hospital admission and
emergency department visit studies conducted in the
U.S. and Canada and evaluated in the 2008 SOx ISA
and studies published since the 2008 SOx ISA 5-52
Figure 5-4 Concentration-response for associations between 3-day
average (lag 0-2) sulfur dioxide concentrations and
emergency department visits for pediatric asthma at the
5th to 95th percentile of sulfur dioxide concentrations in
the Atlanta. GA area 5-71
Figure 5-5 Rate ratio and 95% confidence intervals for
single-pollutant and joint effect models for each
pollutant combination in warm and cold season
analyses for an interquartile range increase in each
pollutant at lag 0-2 days 5-73
x
-------
CONTENTS (Continued)
Figure 5-6 Rate ratio and 95% confidence interval for association
between self-organizing map-based multipollutant day
type and pediatric asthma emergency department visits
at lag 1 5-75
Table 5-10 Epidemiologic studies of pulmonary inflammation and
oxidative stress in populations with asthma published
since the 2008 ISA for Sulfur Oxides 5-80
Table 5-11 Study-specific details from animal toxicological studies
of subclinical effects underlying asthma 5-82
5.2.1.3 Allergy Exacerbation 5-87
5.2.1.4 Chronic Obstructive Pulmonary Disease Exacerbation 5-89
Figure 5-7 Percent increase in chronic obstructive pulmonary
disease hospital admissions and emergency
department visits from U.S. and Canadian studies
evaluated in the 2008 SOx ISA and recent studies in
all-year analyses for a 10-ppb increase in 24-h avg or
40-ppb increase in 1-h max sulfur dioxide
concentrations 5-91
Table 5-12 Study-specific details and mean and upper percentile
concentrations from chronic obstructive pulmonary
disease hospital admission and emergency department
visit studies conducted in the U.S. and Canada and
evaluated in the 2008 SOx ISA and studies published
since the 2008 SOx ISA 5-92
5.2.1.5 Respiratory Infection 5-97
Table 5-13 Study-specific details and mean and upper percentile
concentrations from respiratory infection hospital
admission and emergency department visit studies
conducted in the U.S. and Canada and evaluated in the
2008 SOx ISA and studies published since the 2008
SOx ISA 5-99
Figure 5-8 Percent increase in respiratory infection hospital
admissions and emergency department visits from U.S.
and Canadian studies evaluated in the 2008 SOx ISA
and recent studies in all-year and seasonal analyses for
a 10-ppb increase in 24-h avg or 40-ppb increase in 1-h
max sulfur dioxide concentrations 5-103
5.2.1.6 Aggregated Respiratory Conditions 5-106
Figure 5-9 Percent increase in respiratory disease hospital
admissions and emergency department visits from U.S.
and Canadian studies evaluated in the 2008 SOx ISA
and recent studies in all-year and seasonal analyses for
a 10-ppb increase in 24-h avg or 40-ppb increase in 1-h
max sulfur dioxide concentrations 5-108
Table 5-14 Study-specific details and mean and upper percentile
concentrations from respiratory disease hospital
admission and emergency department visit studies
conducted in the U.S. and Canada and evaluated in the
2008 SOx ISA and studies published since the 2008
SOx ISA 5-109
5.2.1.7 Respiratory Effects in General Populations and Healthy Individuals 5-116
Table 5-15 Study-specific details from controlled human exposure
studies of lung function and respiratory symptoms in
healthy adults 5-119
XI
-------
CONTENTS (Continued)
Table 5-16 Epidemiologic studies of lung function in healthy adults
and adults in the general population published since the
2008 ISA for Sulfur Oxides 5-122
Table 5-17 Epidemiologic studies of lung function in healthy
children and children in the general population
published since the 2008 ISA for Sulfur Oxides 5-125
Table 5-18 Study-specific details from animal toxicological studies
of lung function 5-131
Table 5-19 Epidemiologic studies of respiratory symptoms in
healthy adults and children and groups in the general
population published since the 2008 ISA for Sulfur
Oxides 5-135
Table 5-20 Study-specific details from animal toxicological studies
of subclinical effects 5-141
5.2.1.8 Respiratory Mortality 5-144
Figure 5-10 Percent increase in chronic obstructive pulmonary
disease mortality associated with a 10 |jg/m3 (3.62 ppb)
increase in 24-h avg sulfur dioxide concentrations at
various single and multiday lags in four Chinese cities 5-146
Figure 5-11 City-specific concentration-response curves for
short-term sulfur dioxide exposures and daily chronic
obstructive pulmonary disease mortality in four Chinese
cities 5-148
5.2.1.9 Summary and Causal Determination 5-149
Table 5-21 Summary of evidence for a causal relationship between
short-term sulfur dioxide exposure and respiratory
effects 5-150
5.2.2 Long-Term Exposure 5-156
5.2.2.1 Development and Severity of Asthma 5-157
Table 5-22 Selected epidemiologic studies of long-term exposure
to SO2 and the development of asthma and intervention
studies/natural experiments 5-158
Table 5-23 Study-specific details from animal toxicological studies 5-167
5.2.2.2 Development of Allergy 5-169
5.2.2.3 Lung Function 5-170
5.2.2.4 Respiratory Infection 5-172
5.2.2.5 Development of Other Respiratory Diseases: Chronic Bronchitis, Chronic
Obstructive Pulmonary Disease, and Acute Respiratory Distress Syndrome 5-173
5.2.2.6 Respiratory Mortality 5-174
5.2.2.7 Summary and Causal Determination 5-174
Table 5-24 Summary of evidence for suggestive of, but not
sufficient to infer, a causal relationship between
long-term sulfur dioxide exposure and respiratory
effects 5-176
5.3 Cardiovascular Effects 5-183
5.3.1 Short-Term Exposure 5-183
5.3.1.1 Introduction 5-183
5.3.1.2 Myocardial Infarction and Ischemic Heart Disease 5-184
Figure 5-12 Results of studies of short-term sulfur dioxide exposure
and hospital admissions for ischemic heart disease 5-186
Table 5-25 Mean and upper percentile concentrations of sulfur
dioxide from ischemic heart disease hospital admission
and emergency department visit studies 5-187
xii
-------
CONTENTS (Continued)
5.3.1.3 Arrhythmias and Cardiac Arrest 5-190
Table 5-26 Epidemiologic studies of arrhythmia and cardiac arrest 5-192
5.3.1.4 Cerebrovascular Diseases and Stroke 5-194
Table 5-27 Mean and upper percentile concentrations of sulfur
dioxide from cerebrovascular disease and
stroke-related hospital admission and emergency
department visit studies 5-195
Figure 5-13 Results of studies of short-term sulfur dioxide exposure
and hospital admissions for cerebrovascular disease
and stroke 5-197
5.3.1.5 Blood Pressure and Hypertension 5-198
5.3.1.6 Venous Thromboembolism 5-200
5.3.1.7 Heart Failure 5-200
5.3.1.8 Aggregated Cardiovascular Disease 5-201
Table 5-28 Mean and upper percentile concentrations of sulfur
dioxide from cardiovascular-related hospital admission
and emergency department visit studies: U.S. and
Canadian studies from the 2008 ISA for Sulfur Oxides
and recent studies 5-202
Table 5-29 Mean and upper percentile concentrations of sulfur
dioxide from cardiovascular-related hospital admission
and emergency department visit studies 5-206
Figure 5-14 Studies of hospital admissions and emergency
department visits for all cardiovascular disease 5-209
5.3.1.9 Cardiovascular Mortality 5-210
Figure 5-15 Percent increase in stroke mortality associated with a
10 |jg/m3 (3.62 ppb) increase in sulfur dioxide
concentrations using different lag structures 5-213
Figure 5-16 Pooled concentration-response curves for sulfur dioxide
and daily stroke mortality in eight Chinese cities for a
10 |jg/m3 (3.62 ppb) increase in 24-hour avg
concentrations at lag 0-1 day 5-214
5.3.1.10 Subclinical Effects Underlying Cardiovascular Diseases 5-215
Table 5-30 Epidemiologic studies of biomarkers of cardiovascular
effects 5-220
5.3.1.11 Summary and Causal Determination 5-224
Table 5-31 Summary of evidence, which is inadequate to infer a
causal relationship between short-term sulfur dioxide
exposure and cardiovascular effects 5-226
5.3.2 Long-Term Exposure 5-231
5.3.2.1 Introduction 5-231
5.3.2.2 Myocardial Infarction and Ischemic Heart Disease 5-232
Table 5-32 Epidemiologic studies of the association of long-term
exposure to sulfur dioxide with cardiovascular disease 5-233
5.3.2.3 Cerebrovascular Diseases and Stroke 5-237
Table 5-33 Epidemiologic studies of the association of long-term
exposure to sulfur dioxide with stroke 5-238
5.3.2.4 Blood Pressure and Hypertension 5-240
Table 5-34 Epidemiologic studies of the association of long-term
exposure to sulfur dioxide with hypertension 5-241
5.3.2.5 Other Cardiovascular Effects 5-242
5.3.2.6 Cardiovascular Mortality 5-243
5.3.2.7 Subclinical Effects Underlying Cardiovascular Diseases 5-243
xiii
-------
CONTENTS (Continued)
5.3.2.8 Summary and Causal Determination 5-244
Table 5-35 Summary of evidence, which is inadequate to infer a
causal relationship between long-term sulfur dioxide
exposure and cardiovascular effects 5-245
5.4 Reproductive and Developmental Effects 5-246
5.4.1 Introduction 5-246
Table 5-36 Key reproductive and developmental epidemiologic
studies for sulfur dioxide 5-248
Table 5-37 Study specific details from animal toxicological studies
of the reproductive and developmental effects of sulfur
dioxide 5-253
5.4.2 Fertility, Reproduction, and Pregnancy 5-253
5.4.3 Birth Outcomes 5-256
5.4.3.1 Fetal Growth 5-256
5.4.3.2 Preterm Birth 5-257
5.4.3.3 Birth Weight 5-259
5.4.3.4 Birth Defects 5-261
5.4.3.5 Fetal Mortality 5-261
5.4.3.6 Infant Mortality 5-262
5.4.4 Developmental Outcomes 5-263
5.4.4.1 Respiratory Outcomes 5-263
5.4.4.2 Other Developmental Effects 5-263
5.4.5 Summary and Causal Determination 5-264
Table 5-38 Summary of evidence inadequate to infer a causal
relationship between sulfur dioxide exposure and
reproductive and developmental effects 5-265
5.5 Mortality 5-268
5.5.1 Short-Term Exposure 5-268
5.5.1.1 Introduction 5-268
5.5.1.2 Associations between Short-Term Sulfur Dioxide Exposure and Mortality in
All-Year Analyses 5-270
Table 5-39 Air quality characteristics of multicity studies and
meta-analyses evaluated in the 2008 SOx ISA and
recently published multicity studies and meta-analyses 5-271
Figure 5-17 Percent increase in total mortality from multicity studies
and meta-analyses evaluated in the 2008 ISA for Sulfur
Oxides (black circles) and recently published multicity
studies (red circles) for a 10-ppb increase in 24-hour
avg sulfur dioxide concentrations 5-274
Figure 5-18 Percent increase in total, cardiovascular, and
respiratory mortality from multicity studies evaluated in
the 2008 ISA for Sulfur Oxides (black) and recently
published multicity studies (red) for a 10-ppb increase in
24-h avg sulfur dioxide concentrations 5-276
5.5.1.3 Potential Confounding ofthe Sulfur Dioxide-Mortality Relationship 5-277
Table 5-40 Percent increase in total, cardiovascular, and
respiratory mortality for a 10-ppb increase in 24-h avg
sulfur dioxide concentrations at lag 0-1 in single and
copollutant models 5-279
Figure 5-19 Percent increase in total, cardiovascular, and
respiratory mortality associated with a 10 |jg/m3
(3.62 ppb) increase in 24-h avg sulfur dioxide
concentrations, lag 0-1, in single and copollutants
models in Public Health and Air Pollution in Asia cities 5-281
xiv
-------
CONTENTS (Continued)
Figure 5-20 Percent increase in daily mortality associated with a
10 |jg/m3 (3.62 ppb) increase in 24-h avg sulfur dioxide
concentrations at lag 0-1 days using various degrees of
freedom per year for time trend, China Air Pollution and
Health Effects Study cities, 1996-2008 5-283
Figure 5-21 Percent increase in total mortality associated with a
10 |jg/m3 (3.62 ppb) increase in 24- h avg sulfur dioxide
concentrations at lag 0-1 in Public Health and Air
Pollution in Asia cities, using different degrees of
freedom per year for time trend 5-284
5.5.1.4 Modification of the Sulfur Dioxide-Mortality Relationship 5-285
5.5.1.5 Sulfur Dioxide-Mortality Concentration-Response Relationship and Related
Issues 5-286
Figure 5-22 Percent increase in daily mortality associated with a
10 |jg/m3 (3.62 ppb) increase in 24-h avg sulfur dioxide
concentrations, using various lag structures for sulfur
dioxide in the China Air Pollution and Health Effects
Study cities, 1996-2008 5-287
Figure 5-23 Percent increase in total mortality associated with a
10 |jg/m3 (3.62 ppb) increase in 24-h avg sulfur dioxide
concentrations for different lag structures in individual
Public Health and Air Pollution in Asia cities and in
combined four city analyses 5-288
Figure 5-24 Flexible ambient concentration-response relationship
between short-term sulfur dioxide (ppb) exposure (24-h
avg concentrations) and total mortality at lag 1 5-290
Figure 5-25 Concentration-response curves for total mortality
(degrees of freedom = 3) for sulfur dioxide in each of
the four Public Health and Air Pollution in Asia cities 5-291
5.5.1.6 Summary and Causal Determination 5-292
Table 5-41 Summary of evidence, which is suggestive of, but not
sufficient to infer, a causal relationship between
short-term sulfur dioxide exposure and total mortality 5-293
5.5.2 Long-Term Exposure 5-297
Table 5-42 Summary of studies of long-term exposure and
mortality 5-298
5.5.2.1 U.S. Cohort Studies 5-303
5.5.2.2 European Cohort Studies 5-306
5.5.2.3 Asian Cohort Studies 5-308
5.5.2.4 Cross-Sectional Analysis Using Small Geographic Scale 5-308
5.5.2.5 Summary and Causal Determination 5-310
Figure 5-26 Relative risks (95% confidence interval) of sulfur
dioxide-associated total mortality 5-312
Figure 5-27 Relative risks (95% confidence interval) of sulfur
dioxide-associated cause-specific mortality 5-313
Table 5-43 Summary of evidence, which is inadequate to infer a
causal relationship between long-term sulfur dioxide
exposure and total mortality 5-314
5.6 Cancer 5-315
5.6.1 Introduction 5-315
5.6.2 Cancer Incidence and Mortality 5-316
5.6.2.1 Lung Cancer Incidence and Mortality 5-316
5.6.2.2 Bladder Cancer Incidence and Mortality 5-320
5.6.2.3 Incidence of Other Cancers 5-321
xv
-------
CONTENTS (Continued)
5.6.2.4 Summary of Cancer Incidence and Mortality 5-322
5.6.3 Genotoxicity and Mutagenicity 5-322
5.6.4 Summary and Causal Determination 5-323
Table 5-44 Summary of evidence, which is inadequate to infer a
causal relationship between long-term sulfur dioxide
exposure and cancer 5-324
Table A-1 Scientific considerations for evaluating the strength of
inference from studies on the health effects of sulfur
oxides 5-327
CHAPTER 6 POPULATIONS AND LIFESTAGES POTENTIALLY AT INCREASED
RISK FOR HEALTH EFFECTS RELATED TO SULFUR DIOXIDE
EXPOSURE 6-1
6.1 Introduction 6-1
6.2 Approach to Evaluating and Characterizing the Evidence for At-Risk Factors 6-2
Table 6-1 Characterization of evidence for factors potentially
increasing the risk for sulfur dioxide-related health
effects 6-4
6.3 Pre-existing Disease 6-4
Table 6-2 Prevalence of respiratory diseases among adults and
children by age and region in the U.S. in 2012 6-6
6.3.1 Asthma 6-6
Table 6-3 Controlled human exposure, epidemiology, and animal
toxicology studies evaluating pre-existing asthma and
sulfur dioxide exposure 6-8
6.4 Genetic Factors 6-10
6.5 Sociodemographic and Behavioral Factors 6-11
6.5.1 Lifestage 6-11
6.5.1.1 Children 6-12
Table 6-4 Epidemiologic studies evaluating childhood lifestage
and sulfur dioxide exposure 6-13
6.5.1.2 Older Adults 6-15
Table 6-5 Epidemiologic studies evaluating older adult lifestage
and sulfur dioxide exposure 6-16
6.5.2 Sex 6-18
Table 6-6 Epidemiologic studies evaluating effect modification by
sex and sulfur dioxide exposure 6-19
6.5.3 Socioeconomic Status 6-20
6.5.4 Smoking 6-21
6.6 Conclusions 6-22
Table 6-7 Summary of evidence for potential increased sulfur
dioxide exposure and increased risk of sulfur
dioxide-related health effects 6-23
REFERENCES R-1
xvi
-------
INTEGRATED SCIENCE ASSESSMENT TEAM FOR
SULFUR OXIDES —HEALTH CRITERIA
Executive Direction
Dr. John Vandenberg (Director, RTP Division)—National Center for Environmental
Assessment—Office of Research and Development, U.S. Environmental Protection
Agency, Research Triangle Park, NC
Ms. Debra Walsh (Deputy Director, RTP Division)—National Center for Environmental
Assessment, Office of Research and Development, U.S. Environmental Protection
Agency, Research Triangle Park, NC
Dr. Reeder Sams II (Acting Deputy Director, RTP Division)—National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Dr. Steven J. Dutton (Acting Deputy Director, RTP Division; Branch Chief)—National
Center for Environmental Assessment, Office of Research and Development, U.S.
Environmental Protection Agency, Research Triangle Park, NC
Dr. Mary Ross (Branch Chief)—National Center for Environmental Assessment, Office of
Research and Development, U.S. Environmental Protection Agency, Research Triangle
Park, NC
Dr. Ellen Kirrane (Acting Branch Chief)—National Center for Environmental Assessment,
Office of Research and Development, U.S. Environmental Protection Agency, Research
Triangle Park, NC
Dr. Jennifer Richmond-Bryant (Acting Branch Chief)—National Center for Environmental
Assessment, Office of Research and Development, U.S. Environmental Protection
Agency, Research Triangle Park, NC
Dr. Alan Vette (Acting Branch Chief)—National Center for Environmental Assessment,
Office of Research and Development, U.S. Environmental Protection Agency, Research
Triangle Park, NC
Scientific Staff
Dr. Tom Long (Team Leader, Integrated Science Assessment for Sulfur Oxides)—National
Center for Environmental Assessment, Office of Research and Development,
U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. James Brown—National Center for Environmental Assessment, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Barbara Buckley—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Steven J. Dutton—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
xvii
-------
Dr. Brooke L. Hemming—National Center for Environmental Assessment, Office of
Research and Development, U.S. Environmental Protection Agency, Research Triangle
Park, NC
Dr. Erin Hines—National Center for Environmental Assessment, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Ellen Kirrane—National Center for Environmental Assessment, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Dennis Kotchmar—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Thomas Luben—National Center for Environmental Assessment, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Ms. Connie Meacham—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Jennifer Nichols—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Elizabeth Oesterling Owens—National Center for Environmental Assessment, Office of
Research and Development, U.S. Environmental Protection Agency, Research Triangle
Park, NC
Dr. Molini M. Patel—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Joseph P. Pinto—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Jennifer Richmond-Bryant—National Center for Environmental Assessment, Office of
Research and Development, U.S. Environmental Protection Agency, Research Triangle
Park, NC
Mr. Jason Sacks—National Center for Environmental Assessment, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. David Svendsgaard—National Center for Environmental Assessment, Office of
Research and Development, U.S. Environmental Protection Agency, Research Triangle
Park, NC
Dr. Lisa Vinikoor-Imler—National Center for Environmental Assessment, Office of
Research and Development, U.S. Environmental Protection Agency, Research Triangle
Park, NC
Technical Support Staff
Ms. Marieka Boyd—National Center for Environmental Assessment, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Mr. Kenneth J. Breito—Senior Environmental Employment Program, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
xviii
-------
Ms. Eleanor Jamison—Senior Environmental Employment Program, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Mr. Ryan Jones—National Center for Environmental Assessment, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Ms. Olivia Philpott—Senior Environmental Employment Program, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Mr. Samuel S. Thacker—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Mr. Richard N. Wilson—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
xix
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
Authors
Dr. Tom Long (Team Leader, Integrated Science Assessment for Sulfur Oxides)—National
Center for Environmental Assessment, Office of Research and Development, U.S.
Environmental Protection Agency, Research Triangle Park, NC
Dr. Michael Breen—National Exposure Research Laboratory, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. James Brown—National Center for Environmental Assessment, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Barbara Buckley—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Mr. Evan Coffman—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Ms. Laura Datko-Williams—Oak Ridge Institute for Science and Education, National Center
for Environmental Assessment, Office of Research and Development, U.S.
Environmental Protection Agency, Research Triangle Park, NC
Dr. Steven J. Dutton—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Rachelle Duvall—National Exposure Research Laboratory, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Brooke L. Hemming—National Center for Environmental Assessment, Office of
Research and Development, U.S. Environmental Protection Agency, Research Triangle
Park, NC
Dr. Erin Hines—National Center for Environmental Assessment, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Ellen Kirrane—National Center for Environmental Assessment, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Dennis Kotchmar—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Thomas Luben—National Center for Environmental Assessment, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Jennifer Nichols—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Dr. Michelle Oakes—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
xx
-------
Dr. Elizabeth Oesterling Owens—National Center for Environmental Assessment, Office of
Research and Development, U.S. Environmental Protection Agency, Research Triangle
Park, NC
Dr. Molini M. Patel—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Steven Perry—National Exposure Research Laboratory, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Joseph P. Pinto—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Kristen Rappazzo—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Dr. Jennifer Richmond-Bryant—National Center for Environmental Assessment, Office of
Research and Development, U.S. Environmental Protection Agency, Research Triangle
Park, NC
Mr. Jason Sacks—National Center for Environmental Assessment, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Tina Stevens—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Dr. George Thurstonf—Department of Environmental Medicine, New York University
School of Medicine, Tuxedo, NY
Dr. Lisa Vinikoor-Imler—National Center for Environmental Assessment, Office of
Research and Development, U.S. Environmental Protection Agency, Research Triangle
Park, NC
Dr. Gregory Welleniusf—Department of Community Health (Epidemiology Section),
Brown University, Providence, RI
f Under subcontract through ICF International
Contributors
Ms. Breanna Alman—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Mr. Adam Benson—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Ms. Candis Edwards—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Ms. Charlene Finley—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
xxi
-------
Ms. Beth Gatling—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Mr. William Griffin—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Ms. Rachel Housego—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Mr. Satoru Ito—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Ms. Emily Lau—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Ms. Meagan Madden—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Ms. April Maxwell—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Mr. Ihab Mikati—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Ms. Danielle Moore—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Mr. Kyle Painter—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Dr. Adam Reff—Office of Air Quality Planning and Standards, Office of Air and Radiation,
U.S. Environmental Protection Agency, Research Triangle Park, NC
Ms. Alexandra Ross—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Ms. Kaylyn Siporin—Curriculum for the Environment and Ecology, University of North
Carolina, Chapel Hill, NC
Mr. Doug Solomon—Office of Air Quality Planning and Standards, Office of Air and
Radiation, U.S. Environmental Protection Agency, Research Triangle Park, NC
Ms. Adrien Wilkie—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
xxii
-------
Ms. Brianna Young—Oak Ridge Institute for Science and Education, National Center for
Environmental Assessment, Office of Research and Development, U.S. Environmental
Protection Agency, Research Triangle Park, NC
Reviewers
Dr. Sara Adar—School of Public Health, University of Michigan, Ann Arbor, MI
Mr. Ed Avol—Keck School of Medicine, University of Southern California,
Los Angeles, CA
Dr. Philip Bromberg—School of Medicine, University of North Carolina, Chapel Hill, NC
Dr. Jeffrey Brook—Environment Canada, Toronto, ON
Mr. Matthew Davis—Office of Children's Health Protection, U.S. Environmental Protection
Agency, Washington, DC
Dr. Russell Dickerson—Department of Atmospheric and Oceanic Science, University of
Maryland, College Park, MD
Dr. Tyler Fox—Office of Air Quality Planning and Standards, Office of Air and Radiation,
U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Nicole Hagan—Office of Air Quality Planning and Standards, Office of Air and
Radiation, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Douglas Johns—Division of Respiratory Disease Studies, National Institute for
Occupational Safety and Health, Morgantown, WV
Mr. William Keene—Department of Environmental Sciences, University of Virginia,
Charlottesville, VA
Dr. James Kelly—Office of Air Quality Planning and Standards, Office of Air and
Radiation, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Leila Lackey—Office of Science Policy, Office of Research and Development, U.S.
Environmental Protection Agency, Research Triangle Park, NC
Mr. John Langstaff—Office of Air Quality Planning and Standards, Office of Air and
Radiation, U.S. Environmental Protection Agency, Research Triangle Park, NC
Ms. Connie Meacham—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Qingyu Meng—School of Public Health, Rutgers University, Piscataway, NJ
Dr. Deirdre Murphy—Office of Air Quality Planning and Standards, Office of Air and
Radiation, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. Jennifer Peel—Colorado School of Public Health, Colorado State University,
Fort Collins, CO
Dr. Edward Schelegle—School of Veterinary Medicine, University of California—Davis,
Davis, CA
Dr. Michael Stewart—Office of Air Quality Planning and Standards, Office of Air and
Radiation, U.S. Environmental Protection Agency, Research Triangle Park, NC
xxiii
-------
Dr. James Thurman—Office of Air Quality Planning and Standards, Office of Air and
Radiation, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. John Vandenberg—National Center for Environmental Assessment, Office of Research
and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Dr. James Wagner—College of Veterinary Medicine, Michigan State University,
East Lansing, MI
Dr. Lewis Weinstock—Office of Air Quality Planning and Standards, Office of Air and
Radiation, U.S. Environmental Protection Agency, Research Triangle Park, NC
Ms. Debra Walsh—National Center for Environmental Assessment, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
Ms. Melina Williams—Air and Radiation Law Office, Office of General Counsel, U.S.
Environmental Protection Agency, Washington, DC
xxiv
-------
CLEAN AIR SCIENTIFIC ADVISORY COMMITTEE
SULFUR OXIDES NAAQS REVIEW PANEL
Chair of the Sulfur Oxides Review Panel
Dr. AnaDiez-Roux*, Drexel University, Philadelphia, PA
Sulfur Oxides Review Panel Members
Mr. George A. Allen**—Northeast States for Coordinated Air Use Management
(NESCAUM), Boston, MA
Dr. John R. Balmes—University of California, San Francisco, CA
Dr. James Boylan—Georgia Department of Natural Resources, Atlanta, GA
Dr. Judith Chow**—Desert Research Institute, Reno, NV
Dr. Aaron Cohen—Health Effects Institute, Boston, MA
Dr. Alison Cullen—University of Washington, Seattle, WA
Dr. Delbert Eatough—Brigham Young University, Provo, UT
Dr. Christopher Frey***—North Carolina State University, Raleigh, NC
Dr. William Griffith—University of Washington, Seattle, WA
Dr. Steven Hanna—Hanna Consultants, Kennebunkport, ME
Dr. Jack Harkema**—Michigan State University, East Lansing, MI
Dr. Daniel Jacob—Harvard University, Cambridge, MA
Dr. Farla Kaufman—California Environmental Protection Agency, Sacramento, CA
Dr. Donna Kenski**—Lake Michigan Air Directors Consortium, Rosemont, IL
Dr. David Peden—University of North Carolina at Chapel Hill, Chapel Hill, NC
Dr. Richard Schlesinger—Pace University, New York, NY
Dr. Elizabeth A. (Lianne) Sheppard**—University of Washington, Seattle, WA
Dr. Frank Speizer—Harvard Medical School, Boston, MA
Dr. James Ultman—Pennsylvania State University, University Park, PA
Dr. Ronald Wyzga**—Electric Power Research Institute, Palo Alto, CA
* Chair of the statutory Clean Air Scientific Advisory Committee (CASAC) appointed by the
U.S. EPA Administrator
**Members of the statutory CASAC appointed by the U.S. EPA Administrator
** immediate Past CASAC Chair
XXV
-------
Science Advisory Board Staff
Mr. Aaron Yeow (Designated Federal Officer)—U.S. Environmental Protection Agency,
Science Advisory Board (1400R), 1200 Pennsylvania Avenue, NW, Washington, DC
20460-0001, Phone: 202-564-2050, Fax: 202-565-2098, (yeow.aaron@epa.gov)
(FedEx: 1300 Pennsylvania Avenue, NW, Suite 31150, Washington, DC 20004)
xxvi
-------
ACRONYMS AND ABBREVIATIONS
Acronym/
Abbreviation Meaning
(NH4)2SC>4 ammonium sulfate
35SC>2 raiolabeled SO2
a alpha, exposure factor
a.m. ante meridiem (before noon)
AA adenine-adenine genotype
AB Alberta
ACS American Cancer Society
ADMS Advanced Dispersion Modeling
System
AER air exchange rate; Atmospheric and
Environmental Research
AERMOD American Meteorological
Society/U.S. EPA Regulatory Model
AG adenine-guanine genotype
AHR airway hyperresponsiveness
AIRES Aerosol Research Inhalation
Epidemiology Study
AIRS Aerometric Information Retrieval
System; Atmospheric Infrared
Sounder
AL Alabama
ALRI acute lower respiratory infection
APEX Air Pollution Exposure model
APHEA Air Pollution and Health:
A European Approach study
APIMS atmospheric pressure ionization mass
spectrometry
AQCD air quality criteria document
AQMEII Air Quality Model Evaluation
International Initiative
AQS air quality system
ARDS Acute Respiratory Distress
Syndrome
ARIES Aerosol Research Inhalation
Epidemiology Study
ARP Acid Rain Program
AT Atascadero
Acronym/
Abbreviation Meaning
ATD atmospheric transport and dispersion
ATS American Thoracic Society
avg average
AZ Arizona
P beta
B[a]P benzo[a]pyrene
BAL bronchoalveolar lavage
BALF bronchoalveolar lavage fluid
bax B-cell lymphoma 2-like protein 4
BC black carbon
Bcl-2 B-cell lymphoma 2
BK Bangkok
BMI body mass index
BP blood pressure
BS black smoke
BTEX benzene, toluene, ethylbenzene,
xylene
C degrees Celsius; the product of
microenvironmental concentration;
carbon
C1 sulfur dioxide + nitrogen dioxide
C2 sulfur dioxide + PM10
C3 sulfur dioxide + ozone
CA California
Ca central site ambient SO2
concentration
CAA Clean Air Act
CALPUFF California Puff Model
CAMP Childhood Asthma Management
Program
CAPES China Air Pollution and Health
Effects Study
CASAC Clean Air Scientific Advisory
Committee
xxvii
-------
Acronym/
Abbreviation Meaning
CASTNet Clean Air Status and Trends
Network
CBSA core-based statistical area
CDC Centers for Disease Control and
Prevention
CFR Code of Federal Regulations
cGMP cyclic guanosine monophosphate
CH3SH methyl mercaptan
CHAD Consolidated Human Activity
Database
CHD coronary heart disease
CHF congestive heart failure
CHIMERE regional chemistry transport model
CI(s) confidence interval(s)
CIMS chemical ionization mass
spectroscopy
cIMT carotid intima-media thickness
CI chlorine radical
CMAQ Community Multiscale Air Quality
CO carbon monoxide; Colorado
CO2 carbon dioxide
COH coefficient of haze
Cone concentration
Cong. congress
COPD chronic obstructive pulmonary
disease
COX-2 cyclooxygenase-2
C-R concentration-response (relationship)
CRDS cavity ring-down spectroscopy
CRP c-reactive protein
CS2 carbon disulfide
CSAPR Cross-State Air Pollution Rule
CT Connecticut
CTM chemical transport models
CVD cardiovascular disease
D.C. Cir District of Columbia Circuit
DBP diastolic blood pressure
Acronym/
Abbreviation Meaning
D.C. District of Columbia
DEcCBP diesel exhaust particle extract-coated
carbon black particles
DEN diethylnitrosamine
DEP diesel exhaust particles
df degrees of freedom
DMDS dimethyl disulfide
DMS dimethyl sulfide
DNA deoxyribonucleic acid
DO AS differential optical absorption
spectroscopy
DPB diastolic blood pressure
DVT deep vein thrombosis
e.g. exempli gratia (for example)
Ea exposure to SO2 of ambient origin
EBC exhaled breath condensate
EC elemental carbon
ECA Emissions Control Areas
ECG electrocardiographic
ED emergency department
EGF epidermal growth factor
EGFR epidermal growth factor receptor
EGU electric power generating unit
EIB exercise-induced bronchospasm
EKG electrocardiogram
ELF epithelial lining fluid
EMSA electrophoretic mobility shift assay
Ena exposure to SO2 of nonambient
origin
eNO exhaled nitric oxide
EP entire pregnancy
EPA U.S. Environmental Protection
Agency
ET total exposure over a time period of
interest
ET extrathoracic
EWPM emission-weighted proximity model
xxviii
-------
Acronym/
Abbreviation
Meaning
Acronym/
Abbreviation
Meaning
F
FB
FEF25-75%
female
fractional bias
forced expiratory flow at 25-75% of
exhaled volume
HERO
HF
HI
Health and Environmental Research
Online
high frequency component of HRV
Hawaii
FEF50%
FEF75%
forced expiratory flow at 50% of
forced vital capacity
forced expiratory flow at 75% of
forced vital capacity
HK
HO2
HPDM
Hong Kong
hydroperoxyl radical
Hybrid Plume Dispersion Model
FEFmax
maximum forced expiratory flow
HR
hazard ratio(s); heart rate
FEM
federal equivalent method
HRV
heart rate variability
FEV
forced expiratory volume
HS
hemorrhagic stroke
FEVi
forced expiratory volume in 1 second
HSC
Harvard Six Cities
FOXp3
forkhead box P3
HSOs
bisulfate radical
FPD
flame photometric detection
HSCV
bisulfite
FR
Federal Register
HSC
Harvard Six Cities
FRC
functional residual capacity
Hz
hertz
FRM
federal reference method
i.e.
id est (that is)
FVC
forced vital capacity
i.p.
intraperitoneal
g
gram
ICAM-1
intercellular adhesion molecule 1
GA
Georgia
ICC
intraclass correlation coefficient
GALA II
GG
GIS
GP
GPS
GSTM1
GSTP1
h
H+
Genes-Environments and Admixture
in Latino Americans
guanine-guanine genotype
geographic information system
general practice
global positioning system
glutathione S-transferase Mu 1
glutathione S-transferase Pi 1
hour(s)
hydrogen ion
ICD
ICS
IDW
IFN-y
IgE
IgG
IHD
IKKP
International Classification of
Diseases; implantable cardioverter
defibrillators
inhaled corticosteroid
inverse distance weighting
interferon gamma
immunoglobulin E
immunoglobulin G
ischemic heart disease
inhibitor of nuclear factor kappa-B
kinase subunit beta
H2O
water
IL
Illinois
H2O2
hydrogen peroxide
IL-4
interleukin-4
H2S
hydrogen sulfide
IL-5
interleukin-5
H2SO3
sulfurous acid
IL-6
interleukin-6
H2SO4
sulfuric acid
IN
Indiana
HC
hydrocarbon
IQR
xxix
interquartile range
-------
Acronym/
Abbreviation Meaning
IRP Integrated Review Plan
ISA Integrated Science Assessment
ISAAC International Study of Asthma and
Allergies in Children
IUGR intrauterine growth restriction
IkBoi nuclear factor of kappa light
polypeptide gene enhancer in B-cells
inhibitor, alpha
j microenvironment
JE joint model estimate
k reaction rate; decay constant derived
from empirical data; rate of SO2 loss
in the microenvironment
kg kilogram(s)
km kilometer(s)
L liter(s)
LBW low birth weight
LDL lower detection limit
LF low-frequency component of HRV
LIF laser induced fluorescence
LOESS locally weighted scatterplot
smoothing
Lp-PLA2 lipoprotein-associated phospholipase
A2
LRS lower respiratory symptoms
LUR land use regression
LX lung adenoma-susceptible mouse
strain
m meter
M male
Ml Month 1
M2 Month 2
M3 Month 3
MA Massachusetts
MACC Modeling Atmospheric Composition
and Climate
max maximum
MAX-DOAS multiaxis differential optical
absorption spectroscopy
Acronym/
Abbreviation Meaning
MCh methacholine
MD Maryland
MDL method detection limit
ME Maine
MG geometric mean
mg milligram
MI myocardial infarction ("heart
attack"); Michigan
min minimum; minute
MINAP Myocardial Ischaemia National
Audit Project
MISA Meta-analysis of the Italian studies
on short-term effects of air pollution
mL milliliters)
mm millimeters
MMEF maximum midexpiratory flow
MMFR maximal midexpiratory flow rate
mmHg millimeters of mercury
mmol millimole
MN micronuclei formation, Minnesota
MNPCE polychromatophilic erythroblasts of
the bone marrow
mo month(s)
MO Missouri
MOA mode(s) of action
MODIS Moderate Resolution Imaging
Spectroradiometer
mRNA messenger ribonucleic acid
ms millisecond
MUC5AC mucin 5AC glycoprotein
n sample size; total number of
microenvironments that the
individual has encountered
N population number
N2 molecular nitrogen
NAAQS National Ambient Air Quality
Standards
NaCl sodium chloride
XXX
-------
Acronym/
Abbreviation
NALF
NASA
NBP
NCore
NEI
NFkB
NH
NH3
nh4+
NHAPS
NHLBI
NJ
nm
NMMAPS
NO
NO2
NOs
non-HS
NOx
NR
NY
O3
obs
OC
OCD
OCS
OH
OHCA
OMI
ON
OR
Meaning
nasal lavage fluid
National Aeronautics and Space
Administration
NOx Budget Program
National Core network
National Emissions Inventory
nuclear factor kappa-light-chain-
enhancer of activated B cells
New Hampshire
ammonia
ammonium ion
National Human Activity Pattern
Survey
National Heart, Lung, and Blood
Institute
New Jersey
nanometer
The National Morbidity Mortality
Air Pollution Study
nitric oxide
nitrogen dioxide
nitrate radical
non-hemorrhagic stroke
the sum of NO and NO2
not reported
New York
ozone
observations
organic carbon
Off-shore and Coastal Dispersion
model
carbonyl sulfide
hydroxide; Ohio
out-of-hospital cardiac arrests
Ozone Monitoring Instrument
Ontario
odds ratio(s)
Acronym/
Abbreviation
OVA
P
P
p.m.
P53
PA
PAH(s)
PAPA
Pb
PC
PC(S02)
PE
PEF
PEFR
Penh
PM
PM10
Meaning
ovalbumin
probability
Pearson correlation
post meridiem (after noon)
tumor protein 53
Pennsylvania
polycyclic aromatic hydrocarbon(s)
Public Health and Air Pollution in
Asia
lead
provocative concentration
provocative concentration of SO2
pulmonary embolism
peak expiratory flow
peak expiratory flow rate
enhanced pause
particulate matter
In general terms, particulate matter
with a nominal aerodynamic
diameter less than or equal to 10 |xm;
a measurement of thoracic particles
(i.e., that subset of inhalable particles
thought small enough to penetrate
beyond the larynx into the thoracic
region of the respiratory tract). In
regulatory terms, particles with an
upper 50% cutpoint of 10 ± 0.5 |xm
aerodynamic diameter (the 50%
cutpoint diameter is the diameter at
which the sampler collects 50% of
the particles and rejects 50% of the
particles) and a penetration curve as
measured by a reference method
based on Appendix J of 40 CFR
Part 50 and designated in accordance
with 40 CFR Part 53 or by an
equivalent method designated in
accordance with 40 CFR Part 53.
xxxi
-------
Acronym/
Abbreviation
PMlO-2.5
PM2.5
PMR
PNC
PPb
ppm
PWEI
Qi
Q2
Q3
Q4
Meaning
In general terms, particulate matter
with a nominal aerodynamic
diameter less than or equal to 10 |xm
and greater than a nominal 2.5 |xm; a
measurement of thoracic coarse
particulate matter or the coarse
fraction of PM10. In regulatory terms,
particles with an upper 50% cutpoint
of 10 |xm aerodynamic diameter and
a lower 50% cutpoint of 2.5 |xm
aerodynamic diameter (the 50%
cutpoint diameter is the diameter at
which the sampler collects 50% of
the particles and rejects 50%) of the
particles) as measured by a reference
method based on Appendix O of
40 CFR Part 50 and designated in
accordance with 40 CFR Part 53 or
by an equivalent method designated
in accordance with 40 CFR Part 53.
In general terms, particulate matter
with a nominal aerodynamic
diameter less than or equal to
2.5 |xm; a measurement of fine
particles. In regulatory terms,
particles with an upper 50% cutpoint
of 2.5 |im aerodynamic diameter (the
50%o cutpoint diameter is the
diameter at which the sampler
collects 50%o of the particles and
rejects 50% of the particles) and a
penetration curve as measured by a
reference method based on
Appendix L of 40 CFR Part 50 and
designated in accordance with
40 CFR Part 53, by an equivalent
method designated in accordance
with 40 CFR Part 53, or by an
approved regional method
designated in accordance with
Appendix C of 40 CFR Part 58.
peak-to-mean ratio
particle number concentration
parts per billion
parts per million
Population Weighted Emissions
Index
1 st quartile or quintile
2nd quartile or quintile
3rd quartile or quintile
4th quartile or quintile
Acronym/
Abbreviation
Q5
QT interval
r
R2
Raw
REA
redox
RH
RHC
RIOPA
RMSE
rMSSD
RR
RSP
RSV
RT
S
S. Rep
S2O
SBP
sCI
SD
SDCCE
SDNN
SEARCH
sec
SES
Sess.
SGA
sGAW
SH
Meaning
5th quintile
time between start of Q wave and
end of T wave in ECG
correlation coefficient
square of the correlation coefficient
airway resistance
Risk and Exposure Assessment
reduction-oxidation
relative humidity
robust highest concentration
Relationship Among Indoor,
Outdoor, and Personal Air
root mean squared error
root-mean-square of successive
differences
risk ratio(s), relative risk
respirable suspended particles
respiratory syncytial virus
total respiratory resistance
sulfur
Senate Report
disulfur monoxide
systolic blood pressure
stabilized Criegee intermediate
standard deviation
simulated downwind coal
combustion emissions
standard deviation of all normal-to-
normal intervals
Southeast Aerosol Research
Characterization
second(s)
socioeconomic status
session
small for gestational age
specific airway conductance
Shanghai
xxxii
-------
Acronym/
Abbreviation Meaning
SHEDS Stochastic Human Exposure and
Dose Simulation
SHEEP Stockholm Heart Epidemiology
Programme
SLAMS state and local air monitoring
stations
SO sulfur monoxide
502 sulfur dioxide
S032~ sulfite
503 sulfur trioxide
504 sulfur tetroxide
S042~ sulfate
SOM self-organizing map
SOx sulfur oxides
SPE single-pollutant model estimate
SPM source proximity model; suspended
particulate matter
sRaw specific airway resistance
ST segment segment of the electrocardiograph
between the end of the S wave and
beginning of the T wave
STN Speciation Trends Network
SYP synaptophysin
t fraction of time spent in a
microenvironment across an
individual's microenvironmental
exposures, time
T1 first trimester
T2 second trimester
T3 third trimester
TB tracheobronchial
TBARS thiobarbituric acid reactive
substances (species)
TC total hydrocarbo
Thl T-helper 1
Th2 T-helper 2
TIA transient ischemic attack
TNF-a tumor necrosis factor alpha
tpy tons per year
Acronym/
Abbreviation Meaning
TSP total suspended solids
TX Texas
U.K. United Kingdom
U.S. United States of America
U.S.C. U.S. Code
|x mu; micro
UFP ultrafme particulate matter
|xg/m3 micrograms per cubic meter
URS upper respiratory symptoms
UT Utah
UV ultraviolet
UVF ultraviolet fluorescence
VE minute volume
Vmax maximal flow of expired vital
capacity
Vmax25 maximal expiratory flow rate at 25%
Vmax5o maximal expiratory flow rate at 50%
Vmax75 maximal expiratory flow rate at 75%
VOC volatile organic compound
vs versus
VSGA very small for gestational age
VTE venous thromboembolism
WBC white blood cell
WH Wuhan
WHO World Health Organization
wk week
WHI Women's Health Initiative
WI Wisconsin
WRF Weather Research and Forecasting
yr year(s)
|xg microgram
xxxiii
-------
PREFACE
Legislative Requirements for the Review of the National Ambient
Air Quality Standards
Two sections of the Clean Air Act (CAA) govern the establishment, review, and revision
of the National Ambient Air Quality Standards (NAAQS). Section 108 [42 U.S. Code
(U.S.C.) 7408] directs the Administrator to identify and list certain air pollutants and then
to issue air quality criteria for those pollutants. The Administrator is to list those air
pollutants that in his or her "judgment, cause or contribute to air pollution which may
reasonably be anticipated to endanger public health or welfare," "the presence of which
in the ambient air results from numerous or diverse mobile or stationary sources," and
"for which ... [the Administrator] plans to issue air quality criteria ..." [42 U.S.C.
7408(a)(1); (CAA. 1990a) I. Air quality criteria are intended to "accurately reflect the
latest scientific knowledge useful in indicating the kind and extent of all identifiable
effects on public health or welfare, which may be expected from the presence of [a]
pollutant in the ambient air ..." [42 U.S.C. 7408(a)(2)], Section 109 [42 U.S.C. 7409;
(CAA. 1990bVI directs the Administrator to propose and promulgate "primary" and
"secondary" NAAQS for pollutants for which air quality criteria are issued.
Section 109(b)(1) defines a primary standard as one "the attainment and maintenance of
which in the judgment of the Administrator, based on such criteria and allowing an
adequate margin of safety, are requisite to protect the public health."1 A secondary
standard, as defined in Section 109(b)(2), must "specify a level of air quality the
attainment and maintenance of which, in the judgment of the Administrator, based on
such criteria, is requisite to protect the public welfare from any known or anticipated
adverse effects associated with the presence of [the] air pollutant in the ambient air."2
The requirement that primary standards provide an adequate margin of safety was
intended to address uncertainties associated with inconclusive scientific and technical
information available at the time of standard setting. It was also intended to provide a
1 The legislative history of Section 109 indicates that a primary standard is to be set at"... the maximum permissible
ambient air level... which will protect the health of any [sensitive] group of the population," and that for this
purpose "reference should be made to a representative sample of persons comprising the sensitive group rather than
to a single person in such a group" S. Rep. No. 91:1196, 91st Cong., 2d Sess. 10 (1970).
2 Section 302(h) of the Act [42 U.S.C. 7602(h)] provides that all language referring to effects on welfare includes,
but is not limited to, "effects on soils, water, crops, vegetation, man-made materials, animals, wildlife, weather,
visibility and climate, damage to and deterioration of property, and hazards to transportation, as well as effects on
economic values and on personal comfort and well-being ..." (CAA. 20051.
xxxiv
-------
reasonable degree of protection against hazards that research has not yet identified.1 Both
kinds of uncertainty are components of the risk associated with pollution at levels below
those at which human health effects can be said to occur with reasonable scientific
certainty. Thus, in selecting primary standards that provide an adequate margin of safety,
the Administrator is seeking not only to prevent pollution levels that have been
demonstrated to be harmful but also to prevent lower pollutant levels that may pose an
unacceptable risk of harm, even if the risk is not precisely identified as to nature or
degree. The CAA does not require the Administrator to establish a primary NAAQS at a
zero-risk level or at background concentration levels, but rather at a level that reduces
risk sufficiently so as to protect public health with an adequate margin of safety.2 In so
doing, protection is provided for both the population as a whole and those groups and
lifestages potentially at increased risk for health effects from exposure to the air pollutant
for which each NAAQS is set.
In addressing the requirement for an adequate margin of safety, the U.S. Environmental
Protection Agency (U.S. EPA) considers such factors as the nature and severity of the
health effects involved, the size of the sensitive group(s), and the kind and degree of the
uncertainties. The selection of any particular approach to providing an adequate margin
of safety is a policy choice left specifically to the Administrator's judgment.3
In setting standards that are "requisite" to protect public health and welfare as provided in
Section 109(b), the U.S. EPA's task is to establish standards that are neither more nor less
stringent than necessary for these purposes. In so doing, the U.S. EPA may not consider
the costs of implementing the standards.4 Likewise, "Attainability and technological
feasibility are not relevant considerations in the promulgation of national ambient air
quality standards."5
Section 109(d)(1) requires that "not later than December 31, 1980, and at 5-year intervals
thereafter, the Administrator shall complete a thorough review of the criteria published
under Section 108 and the national ambient air quality standards ... and shall make such
revisions in such criteria and standards and promulgate such new standards as may be
1 See Lead Industries Association v. EPA, 647 F.2d 1130, 1154 [District of Columbia Circuit (D.C. Cir.) 1980];
American Petroleum Institute v. Costle, 665 F.2d 1176, 1186 (D.C. Cir. 1981); American Farm Bureau Federation
v. EPA, 559 F. 3d 512, 533 (D.C. Cir. 2009^Association of Battery Recyclers v. EPA, 604 F. 3d 613, 617-18 (D.C.
Cir. 2010).
2 See Lead Industries v. EPA, 647 F.2d at 1156 n.51; Mississippi v. EPA, 744 F. 3d 1334, 1339, 1351, 1353 (D.C.
Cir. 2013).
3 See Lead Industries Association v. EPA, 647 F.2d at 1161-62; Mississippi v. EPA, 744 F. 3d at 1353.
4 See generally, Whitman v. American Trucking Associations, 531 U.S. 457, 465-472, 475-476 (2001).
5 See American Petroleum Institute v. Costle, 665 F. 2d at 1185.
XXXV
-------
appropriate ...Consistent with this provision, this final ISA contains the air quality
criteria addressing the human health effects of SOx for the current review and reflects the
U.S. EPA's periodic review of those criteria. Section 109(d)(2) requires that an
independent scientific review committee "shall complete a review of the criteria ... and
the national primary and secondary ambient air quality standards ... and shall
recommend to the Administrator any new ... standards and revisions of existing criteria
and standards as may be appropriate ...Since the early 1980s, this independent review
function has been performed by the Clean Air Scientific Advisory Committee
(CAS AC).1
Overview and History of the Reviews of the Primary National
Ambient Air Quality Standard for Sulfur Oxides
NAAQS are defined by four basic elements: indicator, averaging time, level, and form.
The indicator defines the pollutant to be measured in the ambient air for the purpose of
determining compliance with the standard. The averaging time defines the time period
over which air quality measurements are to be obtained and averaged or cumulated,
considering evidence of effects associated with various time periods of exposure. The
level of a standard defines the air quality concentration used (i.e., an ambient air
concentration of the indicator pollutant) in determining whether the standard is achieved.
The form of the standard defines the air quality statistic, the value of which is compared
to the level of the standard in determining whether an area attains the standard. For
example, the form of the current primary 1-h sulfur oxides standard is the 3-yr avg of the
99th percentile of the annual distribution of 1-h daily maximum sulfur dioxide (SO2)
concentrations. The Administrator considers these four elements collectively in
evaluating the protection to public health provided by the primary NAAQS.
The U.S. EPA considers the term sulfur oxides to refer to multiple gaseous oxidized
sulfur species such as sulfur dioxide (SO2) and sulfur trioxide (SO3). In setting the current
standard in 2010, SO2 was chosen as the indicator for sulfur oxides because as in
previous reviews, it was recognized as the most abundant sulfur oxide species in the
atmosphere, and there is a large body of health effects evidence associated with SO2. The
atmospheric chemistry, exposure, and health effects associated with sulfur compounds
present in particulate matter (PM) were most recently considered in the U.S. EPA's 2012
review of the NAAQS for PM. The ecological effects of sulfur oxides are being
considered in a separate Integrated Science Assessment (ISA) for Oxides of Nitrogen,
Oxides of Sulfur, and Particulate Matter—Ecological Criteria (U.S. EPA. 2017a). while
1 Lists of CASAC members and of members of the CASAC Augmented for Sulfur Oxides Panel are available at:
http://vosemite.epa.gov/sab/sabproduct.nsfAVebCASAC/CommitteesandMembership7QpenDocument.
xxxvi
-------
the visibility, climate, and materials damage-related welfare effects of particulate sulfur
compounds are being evaluated in the ISA for particulate matter, as described in the
Integrated Review Plan for the National Ambient Air Quality Standards for Particulate
Matter (U.S. EPA. 2016c).
The U.S. EPA issued the air quality criteria for sulfur oxides in 1969 [34 Federal Register
(FR) 1988; (HEW. 1969)1. Based on these criteria, the U.S. EPA promulgated NAAQS
for sulfur oxides in 1971, establishing the indicator as SO2 [36 FR 8186; (U.S. EPA.
1971)1. The 1971 primary standards were set at 365 (ig/m3 [equal to 0.14 parts per million
(ppm)] averaged over a 24-h period, not to be exceeded more than once per year, and at
80 |ig/m3 (equal to 0.03 ppm) annual arithmetic mean. Since then, the Agency has
completed multiple reviews of the air quality criteria and standards, as summarized in
Table I.
Table I History of the primary National Ambient Air Quality Standards for
sulfur oxides since 1971.
Final Rule/
Decisions
Indicator Averaging Time Level Form
1971
SO2 24 h 140 ppba One allowable exceedance
36 FR 8186
1 yr 30 ppba Annual arithmetic average
Apr 30, 1971
1996
Both the 24-h and annual average standards retained without revision.
61 FR 25566
May 22, 1996
2010
SO2 1 h 75 ppb 3-yr average of the 99th percentile of the
75 FR 35520
annual distribution of daily maximum 1-h
June 22, 2010
concentrations
24-h and annual SO2 standards revoked.
FR = Federal Register; S02 = sulfur dioxide.
aThe level of the 24-h S02 standard was 365 |jg/m3 or 0.14 parts per million (ppm) [equivalent to 140 parts per billion (ppb)]. The
level of the annual S02 standard was 80 |jg/m3 or 0.03 ppm (30 ppb) (36 FR 8186). The levels are presented in ppb for ease of
comparison with the 1-h standard issued in 2010 (75 FR 35520).
In 1982, the U.S. EPA published the Air Quality Criteria for Particulate Matter and
Sulfur Oxides (U.S. EPA. 1982a) along with an addendum of newly published controlled
human exposure studies, which updated the scientific criteria upon which the initial
standards were based (U.S. EPA. 1982b). In 1986, a second addendum was published
presenting newly available evidence from epidemiologic and controlled human exposure
xxxvii
-------
studies (U.S. EPA. 1986). In 1988, the U.S. EPA published a proposed decision not to
revise the existing standards (53 FR 14926). However, the U.S. EPA specifically
requested public comment on the alternative of revising the current standards and adding
a new 1-h primary standard of 0.4 ppm SO2 to protect against short-term peak exposures.
As a result of public comments on the 1988 proposal and other post-proposal
developments, the U.S. EPA published a second proposal on November 15, 1994 (59 FR
58958). The 1994 re-proposal was based in part on a supplement to the second addendum
of the criteria document, which evaluated new findings on the respiratory effects of
short-term SO2 exposures in individuals with asthma (U.S. EPA. 1994). As in the 1988
proposal, the U.S. EPA proposed to retain the existing 24-h and annual standards. The
U.S. EPA also solicited comment on three regulatory alternatives to further reduce the
health risk posed by exposure to high 5-minute peaks of SO2 if additional protection were
judged to be necessary. The three alternatives were: (1) revising the existing primary
NAAQS for sulfur oxides by adding a new 5-minute standard of 0.60 ppm SO2;
(2) establishing a new regulatory program under Section 303 of the Act to supplement
protection provided by the existing NAAQS, with a trigger level of 0.60 ppm SO2 with
one expected exceedance; and (3) augmenting implementation of existing standards by
focusing on those sources or source types likely to produce high 5-minute concentrations
of S02.
Based upon an exposure analysis conducted by the U.S. EPA, the Administrator
concluded that short-term (e.g., 5-minute) exposure of individuals with asthma to SO2 at
levels that can reliably elicit adverse health effects was likely to be a rare event when
viewed in the context of the entire population of individuals with asthma. Thus, the
Administrator judged that high 5-minute SO2 concentrations did not pose a broad public
health problem when viewed from a national perspective, and a 5-minute standard was
not promulgated. In addition, no other regulatory alternative was finalized, and the 24-h
and annual average primary SOx standards were retained in 1996 (61 FR 25566).
The American Lung Association and the Environmental Defense Fund challenged the
U.S. EPA's decision not to establish a 5-minute standard. On January 30, 1998, the Court
of Appeals for the District of Columbia ("D.C. Circuit") found that the U.S. EPA had
failed to adequately explain its determination that no revision to the SO2 NAAQS was
appropriate and remanded the decision back to the U.S. EPA for further explanation.1
Specifically, the court found that the U.S. EPA had not provided adequate rationale to
support the Agency judgment that exposures to high 5-minute concentrations of SO2 do
not pose a public health problem from a national perspective even though these peaks
will likely cause adverse health impacts in a subset of individuals with asthma. Following
1 See American Lung Ass'n v. EPA, 134 F. 3d 388 (D.C. Cir. 1998).
xxxviii
-------
the remand, the U.S. EPA requested that states voluntarily submit 5-minute SO2
monitoring data to be used to conduct air quality analyses in order to gain a better
understanding of the magnitude and frequency of high, 5-minute peak SO2
concentrations. The data submitted by states and the analyses based on this data helped
inform the last review of the SOx NAAQS, which ultimately addressed the issues raised
in the 1998 remand.
The last review of the health-related air quality criteria for sulfur oxides and the primary
SOx standard was initiated in May 2006 (71 FR 28023).1,2 The Agency's plans for
conducting the review were presented in the Integrated Review Plan (IRP) for the
Primary National Ambient Air Quality Standards for Sulfur Oxides (U.S. EPA. 2007a').
which included consideration of comments received during a CASAC consultation as
well as public comment on a draft IRP. The science assessment for the review was
described in the 2008 ISA for Sulfur Oxides—Health Criteria (U.S. EPA. 2008d').
multiple drafts of which received review by CASAC and the public. The U.S. EPA also
conducted quantitative human risk and exposure assessments after having consulted with
CASAC and receiving public comment on a draft analysis plan (U.S. EPA. 2007b). These
technical analyses were presented in the Risk and Exposure Assessment (REA) to
Support the Review of the SO2 Primary National Ambient Air Quality Standards (U.S.
EPA. 2009c'). multiple drafts of which were reviewed by CASAC and the public.
On June 22, 2010, the U.S. EPA revised the primary SO2 NAAQS to provide requisite
protection of public health with an adequate margin of safety (75 FR 35520).
Specifically, after concluding that the then-existing 24-h and annual standards were
inadequate to protect public health with an adequate margin of safety, the U.S. EPA
established a new 1-h SO2 standard at a level of 75 parts per billion (ppb), based on the
3-yr avg of the annual 99th percentile of 1-h daily maximum concentrations. This
standard was promulgated to provide protection against S02-related health effects
associated with short-term exposures ranging from 5 min to 24 h. More specifically, the
U.S. EPA concluded that a 1-h SO2 standard at 75 ppb would substantially limit
exposures associated with the adverse respiratory effects (e.g., decrements in lung
function and/or respiratory symptoms) reported in exercising asthmatics following
5-10 min exposures in controlled human exposure studies, as well as the more serious
health associations (e.g., respiratory-related emergency department visits and
hospitalizations) reported in epidemiologic studies that mostly used daily metrics
1 Documents related to reviews completed in 2010 and 1996 are available at: https://www.epa.gov/naaas/sulfur-
dioxide-so2-primarv-air-aualitv-standards.
2 The U.S. EPA conducted a separate review of the secondary SO2 NAAQS jointly with a review of the secondary
NAAQS for oxides of nitrogen. The Agency retained those secondary standards, without revision, to address the
direct effects on vegetation of exposure to oxides of nitrogen and sulfur (77 FR 202 f 8).
xxxix
-------
(1-h daily max and 24-h avg). In the last review, the U.S. EPA also revoked the
then-existing 24-h and annual primary standards based largely on the recognition that the
new 1-h standard at 75 ppb would generally maintain 24-h and annual SO2 concentrations
well below the NAAQS, as well as the lack of evidence indicating the need for such
longer-term standards (75 FR 35549-50). The decision to set a 1-h standard at 75 ppb—in
part to substantially limit exposure to 5-min concentrations of SO2 resulting in adverse
respiratory effects in exercising asthmatic individuals—also addressed the issues raised in
the remand by the D.C. Circuit in 1998.
As mentioned above, the U.S. EPA's last review placed considerable weight on
substantially limiting health effects associated with high 5-min SO2 concentrations. Thus,
as part of the final rulemaking, the U.S. EPA for the first time required that state and
local agencies operating continuous SO2 analyzers report either the highest 5-min
concentration for each hour of the day, or all twelve 5-min concentrations for each hour
of the day. The rationale for this requirement was that such data were recognized as
critical in the 2010 review and additional monitoring data were anticipated to be valuable
for informing future health studies and NAAQS reviews (75 FR 35522).
After publication of the final rule, a number of industry groups and states filed petitions
for review arguing that the U.S. EPA failed to follow notice-and-comment rulemaking
procedures, and that the decision to establish the 1-h SO2 NAAQS at 75 ppb was arbitrary
and capricious because it was lower than statutorily authorized. The D.C. Circuit rejected
these challenges, thereby upholding the standard in its entirety.1
1 See National Environmental Development Association's Clean Air Project v. EPA, 686 F. 3d 803 (D.C. Cir. 2012),
cert, denied Asarco LLC v. EPA, 133 S. Ct. 983 (Jan. 22, 2013).
xl
-------
EXECUTIVE SUMMARY
Purpose and Scope of the Integrated Science Assessment
This Integrated Science Assessment (ISA) is a comprehensive evaluation and synthesis of
policy-relevant science aimed at characterizing (1) exposures to sulfur oxides (SOX) in
ambient air, for which sulfur dioxide (SO2) is currently the primary atmospheric
indicator, and (2) the health effects associated with these exposures.1 Thus, this ISA
serves as the scientific foundation for the review of the primary (health-based) National
Ambient Air Quality Standard (NAAQS) for SOx. The indicator2 for the current standard
is SO2 because at the time the standard was set it was identified as the most prevalent
species of SOx in the atmosphere and the one for which there is a large body of scientific
evidence on health effects. The health effects of sulfate and other particulate sulfur
compounds are considered as part of the review of the NAAQS for particulate matter
[e.g., in the 2009 Integrated Science Assessment for Particulate Matter (U.S. EPA.
2009a)] ,3 The ecological effects of sulfur oxides are being considered in a separate ISA
for Oxides of Nitrogen, Oxides of Sulfur, and Particulate Matter—Ecological Criteria
(U.S. EPA. 2017a). while the visibility, climate, and materials damage-related welfare
effects of particulate sulfur compounds are being evaluated in the ISA for particulate
matter (U.S. EPA. 2016c).
In 2010, the U.S. Environmental Protection Agency (U.S. EPA) established a new 1-hour
SO2 primary standard of 75 ppb as a 3-year avg of the 99th percentile of each year's
1-hour daily max concentrations (75 FR 35520).4 The 1-hour standard was established to
protect against an array of respiratory effects associated with short-term exposures in
at-risk populations, such as people with asthma. This standard was based on direct
evidence of S02-related effects in controlled human exposure studies of exercising
individuals with asthma, as well as epidemiologic evidence of associations between SO2
concentrations in ambient air and respiratory-related emergency department visits and
hospitalizations. The U.S. EPA also revoked the existing 24-hour and annual primary
SO2 standards of 140 and 30 ppb, respectively, based largely on the recognition that the
new 1-hour standard would generally maintain 24-hour and annual SO2 concentrations
1 The general process for developing an ISA, including the framework for evaluating weight of evidence and
drawing scientific conclusions and causal judgments, is described in a companion document, Preamble to the
Integrated Science Assessments (U.S. EPA. 2015b). https://www.epa.gov/isa.
2 The four components to a NAAQS are: (1) indicator (e.g., SO2), (2) level (e.g., 75 ppb), (3) averaging time
(e.g., 1 hour), and (4) form (e.g., 3 year avg of the 99th percentile of each year's daily 1-hour max concentrations).
3 In this ISA, the blue electronic links can be used to navigate to cited chapters, sections, tables, figures, and studies.
4 The legislative requirements and history of the SO2 NAAQS are described in detail in the Preface to this ISA.
xli
-------
well below the then-existing standards and on the lack of evidence indicating the need for
such longer-term standards (75 FR 35549-50). The U.S. EPA also began requiring that
state and local agencies operating continuous SO2 analyzers to report either the highest
5-minute avg SO2 concentrations for each hour of the day or all twelve 5-minute avg SO2
concentrations for each hour of the day.
This ISA updates the 2008 ISA for Sulfur Oxides lYU.S. EPA. 2008d'). hereafter referred
to as the 2008 SOx ISA] with studies and reports published from January 2008 through
August 2016. The U.S. EPA conducted in-depth searches to identify peer-reviewed
literature on relevant topics such as health effects, atmospheric chemistry, ambient air
concentrations, and exposure. Information was also solicited from subject-matter experts
and the public during a kick-off workshop held at the U.S. EPA in June 2013.
Additionally, drafts of this ISA were reviewed by the CASAC at public meetings held in
January 2016 and March 2017. Members of the public also had an opportunity to
comment on drafts of the ISA. To fully describe the state of available science, The U.S.
EPA also included in this ISA the most relevant studies from previous assessments.
As in the 2008 SOx ISA, this ISA determines the causal nature of relationships with
health effects only for SO2 (Chapter 5). It does not make causality determinations for
health effects of other SOx species because SO2 is the most abundant SOx species in the
atmosphere (Chapter 2). transformation products of SOx such as sulfate are considered in
the ISA for Particulate Matter (U.S. EPA. 2009a'). and the health literature is focused on
SO2. Key to interpreting the health effects evidence is understanding the sources,
chemistry, and distribution of SO2 in the ambient air (Chapter 2) that influence exposure
(Chapter 3). the uptake of inhaled SO2 in the respiratory tract, and what biological
mechanisms may subsequently be affected (Chapter 4). Further, the ISA aims to
characterize the independent effect of SO2 on health (Chapter 5). The ISA also informs
policy-relevant issues (Chapter 1 and Chapter 6). such as (1) exposure durations and
patterns associated with health effects; (2) concentration-response relationship(s),
including evidence of potential thresholds for effects; and (3) populations or lifestages at
increased risk for health effects related to SO2 exposure (Section 1.7.4 and Chapter 6).
Sources and Human Exposure to Sulfur Dioxide
As explained above, this ISA characterizes health effects related to ambient air SO2
exposure. This requires understanding the factors that affect both the exposure to ambient
SO2 and the uncertainty in estimating exposure. These factors include spatial variability
in SO2 concentrations, exposure to copollutants, and uncharacterized time-activity
patterns.
xlii
-------
Emissions of SO2 have decreased by approximately 79% from 1990 to 2014 subsequent
to several federal air quality regulatory programs. Coal-fired electricity generation units
are the dominant sources, emitting 3.2 million tons of SO2 in 2014, nearly 5 times more
than the next largest source (coal-fired boilers for industrial fuel combustion;
Section 2.2). Preliminary data suggest power plant emissions have continued to decline
through 2015-2016. In addition to emission rate, important factors that affect SO2
concentrations at downwind locations include source characteristics (e.g., height of
emissions, temperature, emission rate), local meteorology (e.g., wind, atmospheric
stability, humidity, and cloud/fog cover), and chemistry in the plume (Section 2.3).
The national avg daily 1-hour max SO2 concentration reported during 2013-2015 was
5.3 ppb with a 99th percentile concentration of 64 ppb (Section 2.5). However, 1-hour
daily max SO2 concentrations were 75 ppb or higher during this 3-year period at some
monitoring sites located near point sources, such as power plants or metals processing
facilities, or natural sources, such as volcanoes (which can produce hourly concentrations
in excess of 2,000 ppb). The national 99th percentile 5-minute hourly max concentrations
during 2013-2015 was 24 ppb, suggesting concentrations above 100 ppb are rare,
although monitoring sites near large sources had high concentrations (100 ppb or more),
including four monitoring sites near smelters in Gila County, AZ with 99th percentile
5-minute hourly max concentrations ranging from 116 to 252 ppb. Hourly 5-minute max
concentrations tracked closely with their corresponding 1-h avg concentrations, with 75%
of sites having a correlation above 0.9, indicating that fluctuations in 5-minute hourly
max concentrations are well represented by changes in 1-h avg concentrations. The ratio
of 5-minute hourly max concentrations to their corresponding 1-h avg concentrations was
generally in the range of 1-3, although higher ratios were also observed during some
hours. Background SO2 concentrations due to natural sources and man-made sources
located outside the U.S. are very low across most of the country (annual avg less than
0.03 ppb) except in areas influenced by local cross-border sources and areas affected by
volcanoes, such as Hawaii and parts of the West Coast.
Air quality models are used to estimate SO2 concentrations over various averaging times
in locations without ambient SO2 monitors (Section 2.6). As part of the implementation
program for the 2010 primary NAAQS for SOx, air quality modeling may be used to
characterize air quality for determining compliance with the standard where existing SO2
monitors may not capture peak 1-hour concentrations (75 FR 35520). The widely used
dispersion model American Meteorological Society/U.S. EPA Regulatory Model
(AERMOD) is based on Gaussian dispersion models with enhancements to improve
modeling of SO2 plumes. Model evaluations conducted over averaging times from 1 hour
to 1 year indicate that AERMOD is relatively unbiased in estimating upper-percentile
1-hour concentration values. Lagrangian puff dispersion models, such as CALPUFF
xliii
-------
(California Puff Model), have been developed as an alternative to Gaussian dispersion
models. Uncertainties in model predictions are influenced by uncertainties in model
inputs, particularly emissions data and meteorological conditions.
Correlations between ambient air concentrations of SO2 and other criteria pollutants are
generally low (<0.4), although they vary across location, study, and SO2 averaging time
and are greater than 0.7 at some monitoring sites (Section 3.4.3). Median correlations of
1-hour daily max and 24-h avg SO2 concentrations with particulate matter, nitrogen
dioxide (NO2), and carbon monoxide (CO) during 2013-2015 ranged from 0.2-0.4,
while for ozone (O3) the median daily copollutant correlation with SO2 was less than 0.1
(Figure 3-5).
Estimating exposure concentrations of ambient SO2 for use in epidemiologic studies can
be done in multiple ways. Air quality monitoring data from a limited number of fixed-site
monitors, which are assumed to represent population exposure, are frequently used, but
these monitors may not capture the spatial variation in SO2 concentrations across an
urban area, which can be relatively high in areas affected by large point sources.
Modeling approaches combining air quality data with geographic information or
time-activity patterns, or both, can provide estimates of local ambient concentration or
exposure concentration, although more complex approaches need more detailed inputs
and have the potential for uncertainty related to missing sources, overly smooth
concentration gradients, and other factors.
"Exposure error," which refers to the bias and uncertainty associated with using exposure
metrics to represent the actual exposure of an individual or population, can contribute to
error in health effect estimates in epidemiologic studies (Section 3.4.4). Several
exposure-related factors (including uncharacterized time-activity patterns, spatial and
temporal variability of SO2 concentrations, and distance of individuals and populations
from air quality monitors used in the statistical analyses) contribute to error in estimating
exposure to SO2. Variation in activity patterns across individuals and overtime results in
corresponding variations in exposure concentration. Uncharacterized spatial variability in
SO2 concentrations can contribute to exposure error that tends to add uncertainty and
reduce the magnitude of effect estimates in daily time-series epidemiologic studies. For
long-term (e.g., annual) studies, the effect estimate may be increased or reduced by using
fixed-site monitoring data, depending on the relative locations of sources, monitors, and
exposed people. The exposure error associated with using fixed-site monitors is generally
expected to widen confidence intervals so that the nominal coverage is below 95% for
exposure effect estimates.
xliv
-------
Dosimetry and Mode of Action of Inhaled Sulfur Dioxide
Understanding the absorption and fate of SO2 in the body (dosimetry) and the biological
pathways that potentially underlie health effects (mode of action) is crucial to
characterizing the biological plausibility of SO2 exposure as the cause of observed health
effects.
Inhaled SO2 is readily absorbed in the nasal passages of resting humans and laboratory
animals (Section 4.2). As physical activity increases, there is an increase in breathing rate
and a shift to breathing through the mouth, resulting in greater SO2 penetration into the
lower airways. Relative to healthy adults, children, and individuals with asthma or
allergic rhinitis have an increased amount of oral breathing, and thus, may be expected to
have greater SO2 penetration into the lungs.
The distribution and clearance of inhaled SO2 from the respiratory tract involves several
chemical transformations, particularly the formation of sulfite and S-sulfonates. Sulfite is
metabolized into sulfate, which is rapidly excreted through the urine, while S-sulfonates
are cleared more slowly from the circulation over a period of days. Although SCh-derived
products have been found in the blood and urine within minutes of an inhalation
exposure, a substantial portion of these products appear to be retained within the upper
airways, particularly during nasal breathing, with only slow absorption into the blood.
Although inhaled SO2 produces sulfite that may be distributed through the circulation,
overall sulfite levels are heavily influenced by production within the body (endogenous
production) and by eating food with sulfur-containing amino acids or sulfite itself
(Section 4.2.6). For both adults and children, metabolism of sulfur-containing amino
acids produces much more sulfite than is ingested as food additives. Sulfite produced
endogenously generates levels two or more orders of magnitude higher than
inhalation-derived sulfite levels for both children and adults, even for full-day exposures
to 75 ppb SO2 (i.e., the level of the 1-hour NAAQS). Sulfite ingestion from food
additives varies widely, but is generally expected to exceed sulfite intake from inhalation
in both adults and children, even for full-day exposures to 75 ppb SO2. However, an
important distinction is that inhalation-derived SO2 products can accumulate in the
respiratory tract, whereas sulfite from ingestion or endogenous production does not.
SO2 inhalation produces bronchoconstriction in both healthy adults and those with
asthma (Section 4.3). but the underlying processes are somewhat different. The response
to SO2 in healthy adults occurs primarily from activation of sensory nerves in the
respiratory tract resulting in neural reflex responses through the vagus nerve and occurs
at higher concentrations than the response in people with asthma. In adults with asthma,
the response is only partly due to this neural reflex response, with inflammatory
xlv
-------
mediators also being involved. Inhalation of SO2 increases allergic inflammation in adults
with asthma and in animals with allergic airways disease, which shares many features
with asthma. Furthermore, SO2 inhalation increases allergic sensitization in animals not
already allergic, and once allergic, these animals respond to an allergen challenge with
greater allergic inflammation and airway obstruction (likely due to bronchoconstriction)
compared to animals who were not exposed to SO2. These findings suggest that allergic
inflammation and increased airway responsiveness due to short-term SO2 exposure
(minutes up to 1 month) may be linked to asthma exacerbation seen in epidemiologic
studies.
For long-term SO2 exposure (more than 1 month to years), animal studies provide
additional evidence of airway inflammation, airway remodeling, airway
hyperresponsiveness (AHR), and allergic sensitization. In animals that are not allergic,
SO2 inhalation leads to airway inflammation and allergic sensitization. In animals with
allergic airway disease, SO2 exposure increases airway responsiveness and airway
remodeling. Thus, inhalation of SO2 may lead to the development and worsening of
allergic airway disease. The development of AHR may link long-term exposure to SO2 to
the epidemiologic outcome of physician-diagnosed asthma (new onset asthma).
While there is some evidence for extrapulmonary effects of inhaled SO2, the potential
modes of action underlying these responses are uncertain. Controlled human exposure
studies provide evidence suggesting activation of sensory nerves in the respiratory tract
resulting in a neural reflex response by SO2 exposure, which could lead to changes in
heart rate or heart rate variability. Additionally, the transport of sulfite into the circulation
could result in redox stress, but this is likely to only occur at elevated or prolonged
exposures due to the body's efficient metabolism of sulfite to sulfate.
Health Effects of Sulfur Dioxide Exposure
This ISA integrates information on SO2 exposure and health effects from controlled
human exposure, epidemiologic, and toxicological studies to form conclusions about the
causal nature of relationships between SO2 exposure and health effects. For most health
effect categories, with the exception of reproductive and developmental effects, effects
are evaluated separately for short-term exposures and long-term exposures. Health effects
are considered in relation to the full range of SO2 concentrations relevant to ambient
conditions. Based on upper-percentile ambient air concentrations (Section 2.5) and the
emphasis in the ISA on ambient-relevant exposures within one to two orders of
magnitude of current conditions [Preamble to the ISAs (U.S. EPA. 2015b). Section 5c],
xlvi
-------
SO2 concentrations up to 2,000 ppb1 are defined to be ambient-relevant. A consistent and
transparent framework [Preamble to the ISAs (U.S. EPA. 2015b). Table II] is applied to
classify the health effects evidence according to a five-level hierarchy:
1. Causal relationship
2. Likely to be a causal relationship
3. Suggestive of, but not sufficient to infer, a causal relationship
4. Inadequate to infer a causal relationship
5. Not likely to be a causal relationship
The causal determinations presented in Table ES-1 are informed by recent findings and
whether these recent findings, integrated with information from the 2008 SOx ISA,
support a change in causal conclusions. Important considerations include: (1) determining
whether laboratory studies of humans and animals demonstrate an independent health
effect of SO2 exposure and what the potential underlying biological mechanisms are;
(2) determining whether there is consistency in epidemiologic evidence across various
methods used to estimate SO2 exposure; (3) examining epidemiologic studies of the
potential influence of factors that could bias associations observed with SO2 exposure;
(4) determining the coherence of findings integrated across controlled human exposure,
epidemiologic, and toxicological studies; and (5) making judgments regarding error and
uncertainty in the collective body of available studies.
1 The 2,000-ppb upper limit applies mostly to animal toxicological studies and also a few controlled human
exposure studies. Experimental studies examining SO2 exposures greater than 2,000 ppb were included if they
provided information on the uptake of SO2 in the respiratory tract or on potential biological mechanisms.
xlvii
-------
Table ES-1 Causal determinations for relationships between sulfur dioxide
exposure and health effects from the 2008 and 2017 Integrated
Science Assessment for Sulfur Oxides.
Health Effect Category3 and
Exposure Duration13
Causal Determination
2008 SOx ISAC
2017 SOx ISA
Respiratory effects—short-term exposure
Section 5.2.1, Table 5-21
Causal relationship
Causal relationship
Respiratory effects—long-term exposure
Section 5.2.2. Table 5-24
Inadequate to infer a causal
relationship
Suggestive of, but not sufficient
to infer a causal relationship
Cardiovascular effects—short-term
exposure
Section 5.3.1. Table 5-34
Inadequate to infer a causal
relationship
Inadequate to infer a causal
relationship
Cardiovascular effects—long-term
exposure
Section 5.3.2. Table 5-35
Not included
Inadequate to infer a causal
relationship
Reproductive and developmental effectsd
Section 5.4. Table 5-38
Inadequate to infer a causal
relationship
Inadequate to infer a causal
relationship
Total mortality—short-term exposure
Section 5.5.1. Table 5-41
Suggestive of, but not sufficient to
infer, a causal relationship
Suggestive of, but not sufficient
to infer a causal relationship
Total mortality—long-term exposure
Section 5.5.2. Table 5-43
Inadequate to infer a causal
relationship
Inadequate to infer a causal
relationship
Cancer—long-term exposure
Section 5.6, Table 5-44
Inadequate to infer a causal
relationship
Inadequate to infer a causal
relationship
ISA = Integrated Science Assessment; SOx = sulfur oxides.
aAn array of outcomes is evaluated as part of a broad health effect category: physiological measures (e.g., airway
responsiveness), clinical outcomes (e.g., hospital admissions), and cause-specific mortality. Total mortality includes all
nonaccidental causes of mortality and is informed by findings for the spectrum of morbidity effects (e.g., respiratory,
cardiovascular) that can lead to mortality. The sections and tables referenced include a detailed discussion of the evidence that
supports the causal determinations and the S02 concentrations with which health effects have been associated.
bShort-term exposure refers to time periods of minutes up to 1 mo, while long-term exposures are more than 1 mo to yr.
°Previous causal determinations taken from the 2008 SOx ISA (U.S. EPA. 2008d').
Reproductive and developmental effects studies consider a wide range of exposure durations.
xlviii
-------
Sulfur Dioxide Exposure and Respiratory Effects
As in the 2008 SOx ISA (U.S. EPA. 2008d). the current ISA concludes that there is a
causal relationship between short-term SO2 exposure and respiratory effects, particularly
in individuals with asthma (Section 5.2.1). This determination is based on consistent,
coherent, and biologically plausible evidence for asthma exacerbation due to SO2
exposure. The clearest evidence for this conclusion comes from controlled human
exposure studies available at the time of the 2008 SOx ISA showing lung function
decrements and respiratory symptoms in individuals with asthma exposed to SO2 for
5-10 minutes at elevated breathing rates. The effects observed in these studies are
consistent with the processes leading to asthma exacerbation described in the mode of
action section (Section 4.3). Epidemiologic evidence, including recent studies not
available at the time of the 2008 SOx ISA, also supports a causal relationship, primarily
due to studies reporting positive associations for asthma hospital admissions and
emergency department visits with short-term SO2 exposures, specifically for children.
This is coherent with studies showing that children have increased airway responsiveness
to a trigger and have a greater fraction of oral breathing relative to adults, suggesting they
will have a greater response to SO2 exposure than adults. Hospital admissions and
emergency department visit studies that examined potential copollutant confounding
reported associations were generally unchanged in copollutant models. Additional
support comes from studies reporting positive associations between short-term SO2
exposures and respiratory symptoms in children with asthma, although the evidence from
respiratory symptoms studies in adults with asthma is less consistent. Finally,
epidemiologic studies that report consistent positive associations between short-term SO2
concentrations and respiratory mortality indicate a potential continuum of effects.
For long-term SO2 exposure and respiratory effects the evidence is suggestive of, but not
sufficient to infer, a causal relationship (Section 5.2.2). This conclusion is based on
coherence among findings of a limited number of new epidemiologic studies showing
associations between long-term SO2 exposure and increases in asthma incidence among
children and results of animal toxicological studies that provide a pathophysiologic basis
for the development of asthma. However, uncertainty remains regarding the influence of
other pollutants or mixtures of pollutants on the observed associations with SO2 because
these new epidemiologic studies have not examined the potential for copollutant
confounding. Some epidemiologic evidence regarding respiratory symptoms and/or
respiratory allergies among children also provides limited support for a possible
relationship between long-term SO2 exposure and the development of asthma. This
represents a change in the causal determination made in the 2008 SOx ISA from
inadequate to suggestive, based on a limited body of new evidence.
xlix
-------
Sulfur Dioxide Exposure and Other Health Effects
There is more uncertainty regarding relationships between SO2 exposure and health
effects outside of the respiratory system. SO2 itself is unlikely to enter the bloodstream;
however, its reaction products, such as sulfite, may do so. The amount of circulating
sulfite due to inhalation of SO2 at concentrations relevant to ambient air is far less than
the contribution from metabolism of sulfur-containing amino acids.
For short-term SO2 exposure and total mortality, the current ISA reaches the same
conclusion as the 2008 SOx ISA (U.S. EPA. 2008d): that the evidence is suggestive of,
but not sufficient to infer, a causal relationship (Section 5.5. IV This conclusion is based
on previous and recent multicity epidemiologic studies providing consistent evidence of
positive associations. While recent multicity studies have analyzed some key
uncertainties and data gaps identified in the 2008 SOx ISA, questions remain regarding
the potential for SO2 to have an independent effect on mortality, considering issues such
as the limited number of studies that examined copollutant confounding, evidence for a
decrease in the size of S02-mortality associations in copollutant models with NO2 and
PM10, and the lack of a potential biological mechanism for mortality following short-term
exposures to SO2.
For the remaining health effect categories (short-term and long-term SO2 exposure and
cardiovascular effects, long-term exposure and total mortality, reproductive and
developmental effects, and long-term exposure and cancer), the evidence is inadequate to
infer a causal relationship, mainly due to inconsistent evidence across specific outcomes
and uncertainties regarding exposure measurement error, copollutant confounding, and
potential modes of action. These conclusions are consistent with those made in the 2008
SOx ISA, as illustrated in Table ES-1.
Policy-Relevant Considerations for Health Effects Associated
with Sulfur Dioxide Exposure
This section describes issues relevant for considering the potential importance of impacts
of ambient SO2 exposure on public health, including exposure durations observed to
cause health effects, the shape of the concentration-response relationship, regional
differences, and at-risk populations and lifestages.
Evidence from controlled human exposure studies of respiratory effects after exposures
of 5-10 minutes indicates a rapid onset of S02-related effects (Section 5.2.1).
Epidemiologic studies of asthma hospital admissions and emergency department visits
using daily exposure metrics (24-h avg and 1-h daily max) show positive associations
that are generally unchanged in copollutant models, although these associations could be
1
-------
due to very short duration exposures (5-10 minutes) experienced during the day. The
rapid onset of effects is also coherent with the limited number of epidemiologic studies
that examined lag structures and reported associations within the first few days of
exposure. These aspects of the evidence with regard to exposure duration were cited in
establishing the 1-hour averaging time for the current primary NAAQS for SOx.
Substantial inter-individual variability was observed in controlled human exposure
studies of SO2 and respiratory effects, but there was a clear increase in the magnitude of
respiratory effects with increasing exposure concentrations between 200 and 1,000 ppb
during 5-10 minute SO2 exposures (Section 5.2.1.2). Both the number of affected
individuals with asthma and the severity of the response increased as SO2 concentrations
increased. Epidemiologic studies evaluating the shape of the ambient air
concentration-response function have found no evidence for a population-level threshold
or nonlinearity, although the evidence is limited.
SO2 concentrations in ambient air are highly spatially heterogeneous, with SO2
concentrations at some monitors possibly not highly correlated with the community
average concentration (Section 3.4.2.2V The predominance of point sources results in an
uneven distribution of SO2 concentrations across an urban area. This spatial and temporal
variability in SO2 concentrations can contribute to exposure error in epidemiologic
studies, whether the studies rely on fixed-site monitor data or concentration modeling for
exposure assessment.
Consistent with the findings of the 2008 SOx ISA (U.S. EPA. 2008d). this ISA concludes
there is adequate evidence that people with asthma, particularly children, are at increased
risk for SC>2-related health effects compared with those without asthma (Chapter 6). This
conclusion is based on the evidence for short-term SO2 exposure and respiratory effects
(specifically lung function decrements), for which a causal relationship has been
determined. The ISA concludes there is suggestive evidence that children are at increased
risk for SCh-related health effects, based on their increased ventilation rates relative to
body mass and increased oral breathing, together with some epidemiologic evidence of
increased associations between SO2 and respiratory effects relative to adults, even though
recent epidemiologic evidence is less consistent. There is also evidence from
epidemiologic studies of respiratory hospitalizations, particularly among adults older than
75 years, suggestive of increased risk of SC>2-related health effects for older adults
relative to other lifestages.
li
-------
CHAPTER 1 INTEGRATIVE SYNTHESIS OF
THE ISA
1.1 Purpose and Overview of the Integrated Science Assessment
The Integrated Science Assessment (ISA) is a comprehensive evaluation and synthesis of
the policy-relevant science "useful in indicating the kind and extent of all identifiable
effects on public health or welfare which may be expected from the presence of [a]
pollutant in the ambient air," as described in Section 108 of the Clean Air Act (CAA,
1990a).1 This ISA communicates critical science judgments of the health-related air
quality criteria for the broad category of sulfur oxides (SOx). As such, this ISA serves as
the scientific foundation for the review of the current primary (health-based) National
Ambient Air Quality Standard (NAAQS) for SOx. Consistent with Section 109(d)(1) of
the Clean Air Act, this final ISA contains the air quality criteria addressing the human
health effects of SOx for the current review and reflects the EPA's periodic review of
those criteria. SOx include several related gaseous compounds such as sulfur dioxide
(SO2) and sulfur trioxide (SO3) (Section 2.3). SO2 was chosen as the indicator2 for the
current NAAQS because as in previous reviews, it was identified as the most abundant
sulfur oxide species in the atmosphere (U.S. EPA. 1996b; HEW. 1969).3 and the one for
which there is a large body of evidence on health effects following exposure to SO2 (75
FR 35536). In addition, the 2010 Final Rule concluded that "measures leading to
reductions in population exposures to SO2 can generally be expected to lead to reductions
in population exposures to SOx." (75 FR 35536). Health effects of particulate
sulfur-containing species (e.g., sulfate) are being considered in the current review of the
NAAQS for particulate matter (PM) and were previously evaluated in the 2009 ISA for
PM (U.S. EPA. 2009a). The ecological effects of sulfur oxides are being considered in a
separate ISA for Oxides of Nitrogen, Oxides of Sulfur, and Particulate Matter-Ecological
Criteria (U.S. EPA, 2017a). while the visibility, climate, and materials-related welfare
effects of particulate sulfur compounds are evaluated in the ISA for particulate matter
(U.S. EPA. 2016c).
1 The general process for developing an ISA, including the framework for evaluating weight of evidence and
drawing scientific conclusions and causal judgments, is described in a companion document, Preamble to the
Integrated Science Assessments (U.S. EPA. 2015b').
2 The four components to a NAAQS are (1) indicator (e.g., SO2); (2) level (e.g., 75 ppb); (3) averaging time
(e.g., 1 h), and (4) form (e.g., 3-year avg of the 99th percentile of each year's 1-h daily max concentrations).
3 In this ISA, the blue electronic links can be used to navigate to cited chapters, sections, tables, figures, and studies.
1-1
-------
This ISA evaluates relevant scientific literature published since the 2008 ISA for Sulfur
Oxides rOJ.S. EPA. 2008d'). or 2008 SOx ISA], integrating key information and
judgments contained in the 2008 SOx ISA and the 1982 Air Quality Criteria Document
(AQCD) for Particulate Matter and Sulfur Oxides (U.S. EPA. 1982a) and its Addenda
(U.S. EPA. 1994. 1986. 1982b). Thus, this ISA updates the state of the science that was
available for the 2008 SOx ISA, which informed decisions on the primary SO2 NAAQS
in the review completed in 2010. In 2010, the U.S. Environmental Protection Agency
(U.S. EPA) established a new 1-hour standard of 75 parts per billion (ppb) SO2 as a
3-year avg of the 99th percentile of each year's 1-hour daily max concentrations.1 The
1-hour standard was established to protect against an array of respiratory effects
associated with short-term exposures in potential at-risk populations such as people with
asthma. This standard was based on direct evidence of S02-related effects in controlled
human exposure studies of exercising individuals with asthma, as well as epidemiologic
evidence of associations between SO2 concentrations in ambient air and
respiratory-related emergency department (ED) visits and hospitalizations. The U.S. EPA
also revoked the existing 24-hour and annual primary SO2 standards of 140 and 30 ppb,
respectively. This decision was largely based on the recognition that the new 1-hour
standard at 75 ppb would generally maintain 24-hour and annual SO2 concentrations well
below the then-existing standards and on the lack of evidence indicating the need for such
longer-term standards (75 FR 35549-50).
This new review of the primary NAAQS for SOx is guided by several policy-relevant
questions that are identified in The Integrated Review Plan for the Primary National
Ambient Air Quality Standard for Sulfur Dioxide (U.S. EPA. 2014b). To address these
questions and update the scientific judgments in the 2008 ISA for Sulfur Oxides (U.S.
EPA. 2008d). this ISA aims to:
• Characterize the evidence for health effects associated with short-term (minutes
up to 1 month) and long-term (more than 1 month to years) exposure to SOx by
integrating findings across scientific disciplines and across related health
outcomes (e.g., respiratory effects) and by considering important uncertainties
identified in interpreting the scientific evidence, including the role of SO2 within
the broader mixture of pollutants in the ambient air.
• Inform policy-relevant issues related to quantifying health risks, such as exposure
concentrations, durations, and patterns associated with health effects;
concentration-response (C-R) relationships and existence of thresholds below
which effects do not occur; and populations and lifestages potentially with
increased risk of health effects related to exposure to SOx.
Sulfur dioxide is the most abundant species of SOx in the atmosphere, with other SOx
species being much less prevalent (Section 2.1). Nearly all studies on the health effects of
1 The legislative requirements and history of the SO2 NAAQS are described in detail in the Preface to this ISA.
1-2
-------
SOx focus on SO2. In evaluating the health evidence, this ISA considers possible
influences of other atmospheric pollutants, including interactions of SO2 with
co-occurring pollutants, such as PM (including particulate sulfur compounds), nitrogen
oxides (NOx), carbon monoxide (CO), and ozone (O3).
In addressing policy-relevant questions, this ISA aims to characterize the independent
health effects of SO2. As described in this ISA, recent evidence continues to support a
causal relationship between short-term SO2 exposure and respiratory effects based on the
consistency of findings; coherence among evidence from controlled human exposure,
epidemiologic, and toxicological studies; and biological plausibility for effects
specifically related to asthma exacerbation. The information summarized in this ISA will
serve as the scientific foundation for the review of the current primary 1-hour SO2
NAAQS.
1.2 Process for Developing The Integrated Science Assessment
The U.S. EPA uses a structured and transparent process to evaluate scientific information
and determine the causal nature of relationships between air pollution exposures and
health effects [details provided in the Preamble to the Integrated Science Assessments
(U.S. EPA. 2015b)l. The ISA development process describes approaches for literature
searches, criteria for selecting and evaluating relevant studies, and a framework for
evaluating the weight of evidence and forming causal determinations. As part of this
process, the ISA is reviewed by the Clean Air Scientific Advisory Committee (CASAC),
which is a formal independent panel of scientific experts, and by the public. As this ISA
informs the review of the primary SO2 NAAQS, it integrates and synthesizes information
characterizing exposure to SO2 and potential relationships with health effects. Relevant
studies include those examining atmospheric chemistry, spatial and temporal trends, and
exposure assessment, as well as U.S. EPA analyses of air quality and emissions data.
Relevant health research includes epidemiologic, controlled human exposure, and
toxicological studies on health effects, as well as studies on dosimetry and modes of
action.
The U.S. EPA initiated the current review of the primary NAAQS for SOx in August
2013 with a call for information from the public (U.S. EPA. 2013c). Thereafter, the U.S.
EPA routinely conducted literature searches to identify relevant peer-reviewed studies
published since the previous ISA (i.e., from January 2008 through August 2016).
Multiple search methods were used [Preamble to the IS As (U.S. EPA. 2015b). Section 2],
including searches in the PubMed and Web of Science databases. Subject-area experts
and the public were also able to recommend studies and reports during a science/policy
1-3
-------
issue "kick-off' workshop held by the U.S. EPA in June 2013. The U.S. EPA identified
additional studies considered to be definitive on particular topics from previous
assessments to include in this ISA. Studies that did not address a topic described in the
preceding paragraph based on title were excluded. Studies that were judged to be
potentially relevant based on review of the abstract or full text and "considered" for
inclusion in the ISA are documented and can be found at the Health and Environmental
Research Online (HERO) website. The HERO project page for this ISA
(https://hero.epa.gov/hero/sulfur-oxides') contains the references that are cited in the ISA,
the references that were considered for inclusion but not cited, and electronic links to
bibliographic information and abstracts.
Categories of health effects were considered for evaluation in this ISA if they were
examined in previous U.S. EPA assessments for SOx or in multiple recent studies. For
other categories of health effects, literature searches were conducted to determine the
extent of available health evidence. These searches identified a few recently published
epidemiologic studies on outcomes such as migraine/headache, depression, suicide, eye
irritation/conjunctivitis, rheumatic disease, and gastrointestinal disorders [Supplemental
Table 5S-1 (U.S. EPA. 2016f)l. Literature searches have also identified a few recently
published toxicological studies on hematological effects, mRNA and protein expression
in the brain, sensory symptoms, and effects in other organs (e.g., liver, spleen)
[Supplemental Table 5S-2 (U.S. EPA. 2015cYI. These health effects are not evaluated in
the current ISA because relationship is lacking between the toxicological and
epidemiological health effects examined in these studies and because the results have a
large potential for publication bias (i.e., a greater likelihood of publication for studies
showing effects compared with those showing no effect).
The Preamble to the ISAs (U.S. EPA. 2015b) describes the general framework for
evaluating scientific information, including criteria for assessing study quality and
developing scientific conclusions. Aspects specific to evaluating studies of SOx are
described in the Annex to Chapter 5, Table A-l. For epidemiologic studies, emphasis is
placed on studies that (1) characterize quantitative relationships between SO2 and health
effects, (2) examine exposure metrics that well represent the variability in concentrations
in the study area, (3) consider the potential influence of other air pollutants and factors
correlated with SO2, (4) examine potential at-risk populations and lifestages, or
(5) combine information across multiple cities. With respect to the evaluation of
controlled human exposure and animal toxicological studies, emphasis is placed on
studies that examine effects relevant to humans and SO2 concentrations relevant to
ambient exposures (i.e., exposures to SO2 in ambient air). Based on peak ambient air
concentrations (Section 2.5) and the ISA's emphasis on ambient-relevant exposures
within one to two orders of magnitude of current ambient concentrations, SO2
1-4
-------
concentrations of 2,000 ppb1 or less are defined to be ambient-relevant. Experimental
studies with higher exposure concentrations were included if they contributed to an
understanding of dosimetry or potential modes of action. For the evaluation of human
exposure to ambient SO2, emphasis is placed on studies that examine the quality of data
sources used to assess exposures, such as fixed-site monitors, personal exposure
monitors, and dispersion models. The ISA also emphasizes studies that examine factors
that influence exposure, such as time-activity patterns and building ventilation
characteristics.
The ISA draws conclusions about relationships between SO2 exposure and health effects
by integrating information across scientific disciplines and related health outcomes and
synthesizing evidence from previous and recent studies. Determinations are made about
causation, not just association, and are based on judgments of aspects such as the
consistency, coherence, and biological plausibility of observed effects (i.e., evidence for
effects on key events in the mode of action) as well as related uncertainties. The ISA uses
a formal causal framework [Table II of the Preamble to the ISAs (U.S. EPA. 2015b)! to
classify the weight of evidence according to the five-level hierarchy summarized below.
• Causal relationship: the consistency and coherence of evidence integrated
across scientific disciplines and related health outcomes are sufficient to rule out
chance, confounding, and other biases with reasonable confidence.
• Likely to be a causal relationship: there are studies in which results are not
explained by chance, confounding, or other biases, but uncertainties remain in the
evidence overall. For example, the influence of other pollutants is difficult to
address, or evidence across scientific disciplines may be limited or inconsistent.
• Suggestive of, but not sufficient to infer, a causal relationship: evidence is
generally supportive but not entirely consistent or is limited overall. Chance,
confounding, and other biases cannot be ruled out.
• Inadequate to infer a causal relationship: there is insufficient quantity, quality,
consistency, or statistical power of results from studies.
• Not likely to be a causal relationship: several adequate studies, examining the
full range of anticipated human exposure concentrations and potential at-risk
populations and lifestages, consistently show no effect.
1 The 2,000-ppb upper limit applies largely to animal toxicological studies but also a few controlled human
exposure studies.
1-5
-------
1.3
Organization of the Integrated Science Assessment
This ISA comprises the Preface (legislative requirements of the NAAQS and history of
the primary NAAQS for SOx), Executive Summary, and six chapters. This chapter
(Chapter 1) synthesizes the scientific evidence that best informs policy-relevant questions
that frame this review of the primary NAAQS for SOx. Chapter 2 characterizes the
sources, atmospheric processes involving SOx, and trends in ambient air concentrations.
Chapter 3 describes methods to estimate human exposure to SOx and the impact of error
in estimating exposure on relationships with health effects. Chapter 4 describes the
dosimetry and modes of action for SO2. Chapter 5 evaluates and integrates
epidemiologic, controlled human exposure, and toxicological evidence for health effects
related to short-term and long-term exposure to SOx. Chapter 6 evaluates information on
potential at-risk populations and lifestages. In addition, the Preamble to the ISAs (U.S.
EPA. 2015b) describes the general process for developing an ISA.
The purpose of this chapter is not to summarize each of the chapters but to synthesize the
key findings for each topic that helped characterize SO2 exposure and relationships with
health effects. This chapter also integrates information across the ISA to inform
policy-relevant issues such as SO2 exposure metrics associated with health effects,
concentration-response relationships, and the public health impact of S02-related health
effects (Section 1.7). A key consideration in the health effects assessment is the extent to
which evidence indicates that SO2 exposure independently causes health effects. To that
end, this chapter draws upon information about the sources, distribution, and exposure to
ambient SO2 and identifies pollutants and other factors related to the distribution of or
exposure to SO2 in the ambient air that can potentially influence epidemiologic
associations observed between health effects and SO2 exposure (Section 1.4). The chapter
also summarizes information on the dosimetry and mode of action of inhaled SO2 that
can provide biological plausibility for observed health effects (Section 1.5). The
discussions of the health effects evidence and causal determinations (Section 1.6)
describe the extent to which epidemiologic studies accounted for factors that may
influence epidemiologic study results and the extent to which findings from controlled
human exposure and animal toxicological studies support independent relationships
between SO2 exposure and health effects.
1.4 From Emissions Sources to Exposure to Sulfur Dioxide
Characterizing human exposure is key to understanding the relationships between
ambient SO2 exposure and health effects. The sources of SOx and the transformations
that occur in ambient air influence the spatial and temporal pattern of SO2 concentrations.
1-6
-------
These patterns have implications for variation in exposure in the population, the
adequacy of methods used to estimate exposure, and in turn, the strength of inferences
that can be drawn about health effects related to SO2 exposure.
1.4.1 Emission Sources and Distribution of Ambient Concentrations
Emissions of SO2 have declined by approximately 79% for all sources from 1990 to 2014
subsequent to several U.S. air quality regulatory programs. Coal-fired electricity
generation units (EGUs) remain the dominant sources by nearly fivefold above the next
highest source (industrial fuel combustion), emitting 3.2 million tons of SO2 annually,
according to the 2014 National Emissions Inventory (NEI; Section 2.2). Preliminary
estimates through 2016 suggest further declines in emissions, particularly for EGUs.
In addition to source characteristics, such as emission rate, stack height, and plume
temperature, the two important variables that determine the concentration of SO2
downwind of the source are the photochemical and other removal processes
(e.g., formation of particle-phase reduced sulfur compounds) occurring in the emissions
plume and the local meteorology, including wind, atmospheric stability, humidity, and
cloud/fog cover (Section 2.3). The primary gas-phase photochemical SO2 oxidation
mechanism requires the hydroxyl radical (OH). Another oxidation mechanism involves a
Criegee intermediate biradical that participates in converting SO2 to SO3, which rapidly
reacts with water vapor to form sulfuric acid (H2SO4). The Criegee-based SO2 oxidation
mechanism may amplify the rate of SO2 removal and formation of organosulfur
compounds in areas with high concentrations of Criegee precursors
(i.e., low-molecular-weight organic gases, such as biogenic compounds and unsaturated
hydrocarbons present downwind of industrial sites and refineries). Aqueous-phase
oxidation of SO2 is also an important removal mechanism. Clouds and fog can reduce
local SO2 concentrations by converting it to H2SO4 in the droplet phase.
Changes were undertaken to the existing U.S. EPA monitoring network as a result of
decisions in the 2010 NAAQS review (Section 2.4). First, the automated pulsed
ultraviolet fluorescence (UVF) method, which is most commonly used by state and local
monitoring agencies for NAAQS compliance, was designated as a federal reference
method (FRM). Second, new SO2 monitoring guidelines require state and local agencies
operating continuous SO2 analyzers to report either the highest 5-minute concentration
for each hour of the day or all twelve 5-minute concentrations for each hour of the day.
Analysis of environmental concentrations of SO2 data reported in Section 2.5 reflect the
monitoring network changes, particularly the analysis of the recent 5-minute data.
1-7
-------
On a nationwide basis, the average 1-h daily max SO2 reported during 2013-2015 was
5.3 ppb with a 99th percentile concentration of 64 ppb (Section 2.5). However, 99th
percentile 1-h daily max SO2 concentrations were greater than 75 ppb at some monitoring
sites located near large anthropogenic sources (e.g., power plants or smelters). Volcanoes,
a large natural source of SO2, can produce nearby hourly concentrations over 2,000 ppb.
Nationally, the 99th percentile 5-minute hourly max concentration for 2013-2015 was
24.0 ppb, suggesting that concentrations of 100 ppb or more are relatively rare
nationwide, although monitoring sites near large sources had 99th percentile 5-min
hourly max concentrations above 100 ppb. For example, the four monitors in Gila
County, AZ located near smelters had 99th percentile 5-min hourly max concentrations
ranging from 116 to 252 ppb. Correlations between hourly 5-minute max SO2
concentrations and their corresponding 1-h avg concentrations were high, with
approximately 75% of sites having correlations greater than 0.9. Peak-to-mean ratios
(PMRs) between the two metrics were generally less than 3, although higher PMRs were
observed during some hours (Section 2.5.4). Background concentrations of SO2 from
natural sources and sources outside the U.S. are very low across most of the country
(annual average less than 0.03 ppb), accounting for less than 1% of ambient air
concentrations except in areas influenced by local cross-border sources and areas where
volcanic emissions are important, such as Hawaii and parts of the West Coast
(Section 2.5.5).
SO2 concentrations are highly variable across urban spatial scales, exhibiting moderate to
poor correlations between SO2 measured at different monitoring sites across a
metropolitan area. This high degree of urban spatial variability may not be fully captured
by fixed-site monitors used in epidemiologic studies, and thus, has implications for the
interpretation of human exposure and health effects data (Sections 2.5.2.2 and 3.4.4).
Air quality models, especially dispersion models, can be used to estimate SO2
concentrations over various averaging times in locations where monitoring is not
practical or sufficient (Section 2.6). Because existing monitors may not be sited in
locations to capture peak 1-hour concentrations, the implementation program for the 2010
primary NAAQS for SOx allows for air quality modeling to be used to characterize air
quality for informing designation decisions (75 FR 35520). In addition, modeling is
critical for assessing the impact of future sources or proposed modifications where
monitoring cannot inform and for designing and implementing mitigation techniques.
Dispersion models have also been used to estimate SO2 exposure concentrations in
epidemiologic studies, particularly in long-term studies (Section 3.3.2.4. Chapter 5). The
widely used dispersion model, the American Meteorological Society/U.S. EPA
Regulatory Model (AERMOD), is based on Gaussian dispersion models but includes
advancements such as the ability to incorporate boundary layer scaling formulations,
1-8
-------
surface and elevated emission points, interactions of plumes with buildings and terrain,
and source geometry. Several evaluations of the performance of AERMOD against field
study data over averaging times from 1 hour to 1 year found the model was relatively
unbiased in estimating upper percentage 1-hour concentration values. Lagrangian puff
dispersion models, such as CALPUFF, have been developed as an alternative to Gaussian
dispersion models. CALPUFF models SO2 as a tracer and then uses a Lagrangian step
algorithm to model non-steady-state dynamics, using time-varying winds specified by
meteorological models. CALPUFF simulations were found to improve in accuracy with
increasing integration times. In some cases, CALPUFF predictions are closer to measured
SO2 concentrations compared with AERMOD, but other evaluations have found larger
bias with CALPUFF than AERMOD. Uncertainties in model predictions are influenced
by uncertainties in model input data, particularly emissions and meteorological
conditions (e.g., wind).
1.4.2 Assessment of Human Exposure
Multiple techniques can be used to assign exposure for epidemiologic studies. Evaluation
of data from fixed-site or personal SO2 monitors is commonly used to derive an estimate
of exposure. Various modeling approaches may also be used (Section 3.3). Each has
strengths and limitations, as summarized in Table 3-1. Fixed-site monitors may be
intended to represent population exposure, although some monitors are located near
sources to capture high concentrations locally and are not typically used as the primary
data source in urban-scale epidemiologic studies. Fixed-site monitors may provide a
continuous record of SO2 concentrations over many years, but due to limited spatial
coverage, they may not fully capture the relatively high spatial variability in SO2
concentration across an urban area. Personal SO2 monitors can capture the study
participants' activity-related exposure across different microenvironments, but low
ambient SO2 concentrations often result in a substantial fraction of the samples being
below the limit of detection for averaging times of 24 hours or less. The time and expense
involved to deploy personal monitors make them more suitable for panel epidemiologic
studies than for large-scale time-series or cohort studies. Models can be used to estimate
exposure for individuals and large populations when personal exposure measurements are
unavailable. In general, more complex approaches provide more detailed exposure
estimates but may require additional input data, assumptions, and/or computational
resources. Depending on the model type, there is the potential for bias and reduced
precision due to model misspecification, missing sources, smoothing of concentration
gradients, and complex topography. Evaluation of model results helps demonstrate the
suitability of that approach for particular applications.
1-9
-------
New studies of the relationship between indoor and outdoor SO2 concentrations have
focused on publicly owned buildings rather than residences (Section 3.4.1.2). The results
of these studies are consistent with results of previous studies showing that
indoor: outdoor ratios and indoor-outdoor regression slopes cover an extremely wide
range, from near zero to near one. Differences in results among studies are due to
building characteristics (e.g., forced ventilation, building age, and building type),
personal activities such as opening windows and doors, and SO2 measurement
limitations. Due to indoor deposition and a relative lack of indoor sources of SO2, indoor
concentrations are often much lower than outdoor SO2 concentrations. These low indoor
concentrations also contribute to low personal exposure concentrations due to time spent
indoors. When reported, correlations between indoor and outdoor concentrations were
relatively high (>0.75), suggesting that variations in outdoor concentration drive indoor
concentrations, particularly considering the lack of indoor SO2 sources. These high
correlations were observed across seasons and geographic locations. The bulk of the
evidence for personal-ambient SO2 relationships was available at the time of the 2008
ISA for Sulfur Oxides (U.S. EPA. 2008d') and shows a wide range of correlations
between ambient concentration and personal exposure, in part due to a large fraction of
samples below the method detection limit (MDL) in several studies (Section 3.4.1.3).
When nearly all of the personal samples are below the MDL, no correlation can be
observed. However, when the bulk of the personal samples are above the MDL, personal
exposure is moderately correlated (r = 0.66) with ambient concentration.
"Exposure error" refers to the bias and uncertainty associated with using concentration
metrics to represent the actual exposure of an individual or population KLipfcrt and
Wvzga. 1996) Section 3.21. Exposure error has two components: (1) exposure
measurement error derived from uncertainty in the metric being used to represent
exposure and (2) use of a surrogate target parameter of interest in the epidemiologic study
in lieu of the true exposure, which may be unobservable (Section 3.2.1). Factors that
could contribute to error in estimating exposure to ambient SO2 include
time-location-activity patterns, spatial and temporal variability in SO2 concentrations, and
proximity of populations to monitoring sites and sources (Section 3.4.2V Activity patterns
vary both among and within individuals, resulting in corresponding variations in
exposure across a population and overtime. Variation in SO2 concentrations among
different microenvironments means that the amount of time spent in each location, as
well as exertion level, will influence an individual's exposure to ambient SO2. Time spent
in different locations has also been found to vary by age, with younger and older age
groups spending a greater percentage of time outdoors than adults of typical working age
(18-64 years). These variations in activity pattern contribute to differences in exposure
and, if uncharacterized, introduce error into population-averaged exposure estimates.
1-10
-------
Uncharacterized spatial and temporal variability in SO2 concentrations can contribute to
exposure error in epidemiologic studies. SO2 has low to moderate spatial correlations
among ambient air monitoring sites across urban geographic scales; thus, using fixed-site
monitor data for epidemiologic exposure assessment introduces exposure error into the
resulting health effect estimate. Spatial variability in the magnitude of concentrations
may affect cross-sectional and large-scale cohort studies by assigning exposures from one
or a small number of sites that do not capture all of the spatial variability within a city.
Some models are designed to address this situation by improving the characterization of
spatial and temporal variability. This issue may be less important for time-series studies,
which rely on day-to-day temporal variability in concentrations to evaluate health effects.
Proximity of populations to ambient air monitoring sites may influence how well human
exposure is represented by measurements at the monitors, although factors other than
distance play an important role as well. Many ambient SO2 monitoring sites are located
near dense population centers, but other near-source sites may not be near population
centers. Use of monitoring sites in epidemiologic studies introduces exposure error into
health effect estimates. The literature has shown that exposure error and related bias in
the health effect estimate is reduced by using averaging schemes in lieu of a single fixed-
site monitor (Section 3.4.2.2).
Exposure to copollutants, such as other criteria pollutants, may result in confounding of
health effect estimates. For SO2, daily concentrations generally exhibit low correlations
(median <0.4) with other daily NAAQS pollutant concentrations at collocated monitors
(Figure 3-5. Section 3.4.3). However, a wide range of copollutant correlations has been
observed across different monitoring sites, from moderately negative to moderately
positive. Data showing high daily SO2 correlations with nitrogen dioxide (NO2) and CO
may have been collected before the rule to reduce sulfur content in diesel fuel (66 FR
5002) took effect in 2006 and 2007. The minority of sites with stronger correlations may
introduce a greater degree of confounding into epidemiologic results. A similar impact is
expected for epidemiologic studies of long-term SO2 exposure, which also report a wide
range of copollutant correlations.
Exposure error can influence epidemiologic study results by biasing effect estimates
either toward or away from the null and widening confidence intervals beyond the
nominal coverage that would be produced if the true exposure had been used
(Section 3.4.4). The exposure error varies according to the study design, especially
regarding the study's spatial and temporal aspects. For example, in time-series and panel
studies, low personal-ambient correlations tend to bias the effect estimate toward the null,
while spatial variation in personal-ambient correlations across an urban area contributes
to widening of the confidence interval around the effect estimate so that the nominal
1-11
-------
coverage of the confidence intervals is below 95% for exposure effect estimates. For
long-term studies, bias of the health effect estimate may occur in either direction
depending on whether the monitor is over- or underestimating exposure for the
population of interest. In all study types, use of fixed-site monitors is expected to
decrease precision of the health effect estimate because spatial variation in
personal-ambient correlations across an urban area contributes to widening of the
confidence interval around the effect estimate so that the nominal coverage is below 95%
for exposure effect estimates.
Choice of exposure estimation method also influences the impact of exposure error on
epidemiologic study results. Fixed-site monitors offer a convenient source of time-series
data, but fixed-site measurements do not account for the effects of spatial variation in
SO2 concentration, differences between indoor and outdoor exposure to ambient SO2, and
varying activity patterns on personal exposure to SO2. Personal exposure measurements,
such as those made in panel epidemiologic studies, provide accurate and specific
exposure estimates, but sample size is often small, and only a limited set of health
outcomes can be studied. Modeled concentrations or exposures offer alternatives to
measurements, with the advantage of estimating exposures over a wide range of scales,
populations, and scenarios, particularly for locations lacking monitoring data. However,
depending on the model type, there is the potential for bias and reduced precision due to
model misspecification, missing sources, smoothing of concentration gradients, and
complex topography. Model estimates for unmonitored locations or time periods are most
informative when the model output has been compared to an independent set of measured
concentrations or exposures. The various sources of exposure error and their potential
impact are considered in evaluating the epidemiologic study results in Chapter 5 of this
ISA.
1.5 Dosimetry and Mode of Action of Sulfur Dioxide
Dosimetry of inhaled SO2, including the processes of absorption, distribution,
metabolism, and elimination, as well as information on the mode of action of inhaled
SO2, covering the processes by which inhaled SO2 initiates a cascade of molecular and
cellular responses and subsequent organ-level responses, is covered in Chapter 4.
Together, these sections provide the foundation for understanding how exposure to
inhaled SO2 can lead to health effects. This understanding enables our characterization of
the biological plausibility of SO2 exposure as the cause of health effects that may be
observed in epidemiologic studies.
1-12
-------
1.5.1
Dosimetry of Inhaled Sulfur Dioxide
Dosimetry of inhaled SO2 refers to the measurement or estimation of the amount of SO2
and its reaction products reaching and/or persisting at specific sites within the respiratory
tract and systemically after exposure. Factors affecting the transport and fate of SO2 in
the respiratory tract include respiratory tract morphology, respiratory functional
parameters, and the physicochemical properties of SO2 and epithelial lining fluid (ELF).
Health effects may be due to inhaled SO2 or its chemical reaction products, including
sulfite and S-sulfonates. Few studies have investigated SO2 dosimetry since the 2008 ISA
for Sulfur Oxides (U.S. EPA. 2008d). with most studies conducted prior to the 1982
AQCD (U.S. EPA. 1982a) and the 1986 Second Addendum (U.S. EPA. 1986V
Because SO2 is highly soluble in water, it is readily absorbed in the nasal passages of
both humans and laboratory animals under resting conditions (Section 4.2.2). During
nasal breathing, the majority of available data suggests 95% or greater SO2 absorption
occurs in the nasal passages, even under ventilation levels comparable to that during
exercise. With increasing physical activity, there is an increase in ventilatory rate and a
shift from nasal to oronasal breathing, resulting in greater SO2 penetration into the lower
respiratory tract. Even at rest, differences have been observed by age, sex, disease status,
and body mass index in the fraction of oral versus nasal breathing (Section 4.1.2).
Children inhale a larger fraction of air through their mouth than adults, and males tend to
inhale a larger fraction of air through their mouth than females (across all ages).
Individuals with allergies or upper respiratory infections experience increased nasal
resistance, and thus, increased fraction of oral breathing. Obesity, especially in boys, may
also contribute to increased nasal resistance and an increased oral fraction of breathing
relative to normal weight children. Due to their increased amount of oral breathing, these
individuals may be expected to have greater SO2 penetration into the lower respiratory
tract than healthy, normal weight adults. Children may also be expected to have a greater
intake dose of SO2 per body mass than adults.
Following absorption in the respiratory tract, SO2 rapidly forms a mixture of bisulfite and
sulfite, with the latter predominating. As much as 15-18% of the absorbed SO2 may be
desorbed and exhaled following cessation of exposure. Although some SO2 products
rapidly move from the respiratory tract into the blood and are distributed about the body,
experiments using radiolabeled 35S indicate that the majority of sulfur in S02-derived
products in the body at any given time following exposure is found in the respiratory tract
and may be detected there for up to a week following inhalation (Section 4.2.3). The
distribution and clearance of inhaled SO2 from the respiratory tract may involve several
intermediate chemical reactions and transformations, particularly the formation of sulfite
and S-sulfonates. Sulfite is metabolized into sulfate, primarily in the liver, which has
1-13
-------
higher sulfite oxidase levels than the lung or other body tissues (Section 4.2.4). Sulfite
oxidase activity is highly variable between species, with liver sulfite oxidase activity in
rats being 10-20 times greater than in humans. Urinary excretion of sulfate is rapid and
proportional to the concentration of SO2 products in the blood (Section 4.2.5).
S-sulfonates are cleared more slowly from the circulation with a clearance half-time of
days.
Sulfite levels in the body are predominately influenced by endogenous production and
ingestion of sulfite in food (Section 4.2.6). The primary endogenous contribution of
sulfite is from the catabolism of sulfur-containing amino acids (namely, cysteine and
methionine). Endogenous sulfite from ingested sulfur-containing amino acids far exceeds
exogenous sulfite from ingestion of food additives [by 140 and 180 times in adult
(19-50 years) females and males, respectively, and by 500 times or more in young
children (1-3 years)]. Endogenous sulfite production is two or more orders of magnitude
higher than inhalation-derived sulfite levels for both children and adults, even for full day
exposures to 75 ppb SO2 (the level of the 1-hour NAAQS). Ingestion rates of sulfite
added to foods vary widely; however, in general, sulfite ingestion is expected to exceed
sulfite intake from inhalation in adults and children even for full day exposures to 75 ppb
SO2. However, sulfite and sulfate from ingestion or endogenous production do not
accumulate primarily in respiratory tract tissues, as is the case for inhalation-derived SO2
products.
1.5.2 Mode of Action of Inhaled Sulfur Dioxide
Mode of action refers to a sequence of key events, endpoints, and outcomes that result in
a given toxic effect. The mode of action discussion in Section 4.3 of this ISA updates the
basic concepts derived from the SO2 literature presented in the 1982 AQCD (U.S. EPA.
1982a) and the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d) and introduces the recent
relevant literature. The main effects of SO2 inhalation are seen at the sites of absorption
(i.e., the respiratory tract) and include (1) activation of sensory nerves in the respiratory
tract resulting in a neural reflex response, (2) injury to airway mucosa, and (3) increased
airway responsiveness and allergic inflammation. Effects outside the respiratory tract
may occur at very high concentrations of inhaled SO2.
Reactive products formed as a result of SO2 inhalation are responsible for a variety of
downstream key events, which may include activation of sensory nerves in the
respiratory tract, release of inflammatory mediators, and modulation of allergic
inflammation or sensitization. These key events may collectively lead to several
endpoints, including bronchoconstriction and increased airway responsiveness. A
1-14
-------
characteristic feature of individuals with asthma is an increased propensity of their
airways to narrow in response to bronchoconstrictive stimuli relative to nonatopic
individuals without asthma. Thus, bronchoconstriction is characteristic of an asthma
attack. However, individuals without asthma may also experience bronchoconstriction in
response to SO2 inhalation; generally this occurs at higher concentrations (>1,000 ppb)
than in an individual with asthma. Additionally, SO2 exposure may increase airway
responsiveness to subsequent exposures of other stimuli such as allergens or
methacholine. These pathways may play a role in the epidemiologic study outcome of
asthma exacerbation.
The strongest evidence for the mode of action for respiratory effects following short-term
exposure comes from controlled human exposure studies. SO2 exposure resulted in
increased airway resistance due to bronchoconstriction in adults, both with and without
asthma. In adults without asthma, this response occurred primarily as a result of
activation of sensory nerves in the respiratory tract resulting in neural reflex responses
(Section 4.3.1). This is mediated by cholinergic parasympathetic pathways involving the
vagus nerve. However, in adults with asthma, evidence indicates that the response is only
partially due to vagal pathways and that inflammatory mediators such as histamine and
leukotrienes also play an important role. Studies in experimental animals also
demonstrate that SO2 exposure activates reflexes that are mediated by cholinergic
parasympathetic pathways involving the vagus nerve. However, noncholinergic
mechanisms (i.e., neurogenic inflammation) may also be involved.
Evidence demonstrates that SO2 exposure modulates allergic inflammatory responses
(Section 4.3.2). Enhancement of allergic inflammation (i.e., leukotriene-mediated
increases in numbers of sputum eosinophils) has been observed in adults with asthma
who were exposed for 10 minutes to 750 ppb SO2. In an animal model of allergic airway
disease, repeated exposure to 2,000 ppb SO2 led to an enhanced inflammatory response,
including allergic inflammation. In naive animals, repeated exposure to SO2 (as low as
100 ppb) over several days promoted allergic sensitization, inflammation, and AHR when
animals were subsequently sensitized and challenged with an allergen. Thus, allergic
inflammation and increased airway responsiveness may also link short-term SO2
exposure to asthma exacerbation.
Evidence for the mode of action for respiratory effects due to long-term SO2 exposure
comes from studies in both naive and allergic experimental animals, which demonstrate
allergic sensitization, allergic inflammation, AHR, and morphologic changes suggestive
of airway remodeling following exposure to SO2 (i.e., 2,000 ppb) over several weeks
(Section 4.3.3). These changes, however, are mild compared to histopathological
changes, such as mucous cell metaplasia and intramural fibrosis, which are generally
1-15
-------
observed following chronic exposure of naive animals to extremely high SO2
concentrations (10,000 ppb and higher). However, in allergic animals, exposure to SO2
over several weeks leads to morphologic responses indicative of airway remodeling and
to AHR (Section 4.3.6). Thus, repeated exposure to SO2 may lead to the development of
allergic airway disease, which shares many features with asthma, and to the worsening of
the allergic airway disease. The development of AHR may link long-term exposure to
SO2 to the epidemiologic outcome of new onset asthma.
Although some evidence suggests that SO2 inhalation results in extrapulmonary effects,
there is uncertainty regarding the mode of action underlying these responses
(Section 4.3.4). Evidence from controlled human exposure studies points to SO2
exposure-induced activation/sensitization of neural reflexes, possibly leading to altered
heart rate (HR) or heart rate variability (HRV). Evidence also points to transport of sulfite
into the circulation. Sulfite is highly reactive and may be responsible for redox stress
(possibly through autoxidation or peroxidase-mediated reactions to produce free radicals)
in the circulation and extrapulmonary tissues. However, this stress is likely to occur only
at very high SO2 concentrations or during prolonged exposures because circulating sulfite
is efficiently metabolized to sulfate in a reaction catalyzed by hepatic sulfite oxidase.
1.6 Health Effects of Sulfur Dioxide
This ISA evaluates relationships between an array of health effects and short-term and
long-term exposures to SO2 as examined in epidemiologic, controlled human exposure,
and animal toxicological studies. Short-term exposures are defined as those with
durations of minutes up to 1 month, with most studies examining effects related to
exposures in the range of 1 hour to 1 week. Long-term exposures are defined as those
with durations of more than 1 month to years. Drawing from the health effects evidence
described in detail in Chapter 5. information on dosimetry and modes of action presented
in Chapter 4. as well as issues regarding exposure assessment and potential confounding
described in Chapter 3. the subsequent sections and Table 1-1 present the key evidence
that informed the causal determinations for relationships between SO2 exposure and
health effects.
1-16
-------
1.6.1 Respiratory Effects
1.6.1.1 Respiratory Effects Associated with Short-Term Exposure to Sulfur Dioxide
Strong scientific evidence indicates that there is a causal relationship between short-term
SO2 exposure and respiratory morbidity, particularly in individuals with asthma, which is
consistent with the conclusions of the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d).
This determination is based on the consistency of findings within disciplines, coherence
among evidence from controlled human exposure, epidemiologic, and toxicological
studies, and biological plausibility for effects specifically related to asthma exacerbation
(Table 5-21).
This conclusion is primarily based on controlled human exposure studies included in the
2008 SOx ISA (U.S. EPA. 2008d) that showed lung function decrements and respiratory
symptoms in individuals with asthma exposed to SO2 for 5-10 minutes under increased
ventilation conditions. No new controlled human exposure studies have since been
conducted to evaluate the effect of SO2 on respiratory morbidity among individuals with
asthma. The existing studies consistently demonstrate that individuals with asthma
experience a moderate or greater decrement in lung function, defined as a >100%
increase in specific airway resistance (sRaw) or >15% decrease in forced expiratory
volume in 1 sec (FEVi), frequently accompanied by respiratory symptoms, following
peak exposures of 5-10 minutes with elevated ventilation rates at concentrations of
400-600 ppb (Section 5.2.1.2). A fraction of individuals with asthma (-5-30%) was
observed in these studies to have moderate decrements in lung function (i.e., >100%
increase in sRaw or >15% decrease in FEVi) at lower SO2 concentrations (200-300 ppb;
Table 5-2). Lung function decrements at these lower concentrations are less likely to be
accompanied by respiratory symptoms. Some studies have evaluated the influence of
asthma severity on response to SO2, but the most severe asthmatics have not been tested,
so their response is unknown. Adults with moderate to severe asthma demonstrated larger
absolute changes in lung function during exercise in response to SO2 than adults with
mild asthma, although this difference was attributed to a larger response to the exercise
component of the protocol rather than to SO2 itself. While adults with moderate to severe
asthma may have similar responses to SO2 as those with mild asthma, they have less
reserve capacity to deal with an insult compared with individuals with mild asthma;
therefore, the impact of SC>2-induced decrements in lung function is likely greater in
individuals with severe asthma than those with mild asthma. Although there are no
laboratory studies of children younger than 12 exposed to SO2, a number of studies have
evaluated airway responsiveness of children and adults to a bronchoconstrictive stimulus.
1-17
-------
These studies indicate that school-aged children, particularly boys, are expected to have
greater responses (i.e., greater lung function decrements) following exposure to SO2 than
adolescents and adults.
These findings are consistent with the current understanding of dosimetry and modes of
action (Section 1.5). Due to their increased fraction of oral breathing, individuals with
asthma may be expected to have greater SO2 penetration into the lower respiratory tract
than healthy adults. Reactive products formed as a result of SO2 inhalation, particularly
sulfites and S-sulfonates, are responsible for a variety of downstream key events, which
may include activation of sensory nerves in the respiratory tract resulting in a neural
reflex response, release of inflammatory mediators, and modulation of allergic
inflammation. These key events may lead to several endpoints including
bronchoconstriction and increased airway responsiveness, resulting in the outcome of
asthma exacerbation.
Epidemiologic evidence also provides support for a causal relationship, including
additional studies that add to the evidence provided by the 2008 ISA for Sulfur Oxides
(U.S. EPA. 2008dV Studies of asthma hospital admissions and ED visits report positive
associations with short-term SO2 exposures, particularly for children (i.e., <18 years of
age), with evidence from a limited number of studies that examine potential copollutant
confounding indicating that associations remain positive, though are in some instances
attenuated in magnitude, in copollutant models involving PM and other criteria pollutants
(Section 5.2.1.2V There is also some supporting evidence for positive associations
between short-term SO2 exposures and respiratory symptoms among children with
asthma (Section 5.2.1.2). Epidemiologic evidence of associations between short-term SO2
exposures and lung function or respiratory symptoms among adults with asthma is less
consistent (Section 5.2.1.2V Epidemiologic studies of cause-specific mortality that report
consistent positive associations between short-term SO2 exposures and respiratory
mortality provide support for a potential continuum of effects (Section 5.2.1.8V
There is some support for other SC>2-related respiratory effects including exacerbation of
chronic obstructive pulmonary disease (COPD) in individuals with COPD and other
respiratory effects including respiratory infection, aggregated respiratory conditions, and
respiratory mortality in the general population (Sections 5.2.1.5. 5.2.1.6. 5.2.1.7. and
5.2.1.8V The limited and inconsistent evidence for these nonasthma-related respiratory
effects does not contribute substantially to the causal determination.
1-18
-------
1.6.1.2
Respiratory Effects Associated with Long-Term Exposure to Sulfur Dioxide
Overall, the evidence is suggestive of, but not sufficient to infer, a causal relationship
between long-term SO2 exposure and respiratory effects, mainly the development of
asthma in children (Section 5.2.2). This represents a change from the conclusion in the
2008 ISA for Sulfur Oxides (U.S. EPA. 2008d') that the evidence was "inadequate to infer
a causal association." There is a limited number of recent longitudinal epidemiologic
studies that evaluate associations between asthma incidence among children and
long-term SO2 exposures, with the overall body of evidence lacking consistency. The
evidence from longitudinal studies showing increases in asthma incidence is coherent
with findings from animal toxicological studies that provide a pathophysiologic basis for
the development of asthma. In naive newborn animals, repeated SO2 exposure over
several weeks resulted in immune responses and airway inflammation, key steps in
allergic sensitization. In allergic newborn animals, studies with several days or several
weeks of repeated SO2 exposure found enhanced airway inflammation and some evidence
of airway remodeling and AHR. The combined epidemiologic and animal toxicological
evidence provides support for an independent effect of long-term exposure to SO2 on the
development of asthma in children, but key uncertainties remain, including exposure
measurement error and the potential for copollutant confounding. Some evidence of a
link between long-term exposure to SO2 and respiratory symptoms or respiratory
allergies among children further supports a possible relationship between long-term SO2
exposure and the development of asthma. Details of the causal determination are
provided in Table 5-24.
1.6.2 Health Effects beyond the Respiratory System
1.6.2.1 Cardiovascular Effects Associated with Short-Term Exposure to Sulfur
Dioxide
Overall, the available evidence is inadequate to infer a causal relationship between
short-term exposure to SO2 and cardiovascular health effects (Table 5-31. Section 5.3.1).
This conclusion is consistent with that of the 2008 ISA for Sulfur Oxides (U.S. EPA.
2008d), which concluded "the evidence as a whole is inadequate to infer a causal
relationship." Although multiple epidemiologic studies report positive associations
between short-term exposure to SO2 and a variety of cardiovascular outcomes, the results
are inconsistent across the specific cardiovascular outcomes, and the associations are
generally attenuated after copollutant adjustment. There is some experimental evidence in
1-19
-------
humans and animals for SCh-induced effects on the autonomic nervous system and
inflammation and other effects in tissues distal to the absorption site. However, the
limited and inconsistent evidence from the available experimental studies does not
demonstrate potentially biologically plausible mechanisms for, and is not coherent with,
cardiovascular effects such as triggering a myocardial infarction. Evidence for other
cardiovascular and related metabolic effects is inconclusive.
1.6.2.2 Cardiovascular Effects Associated with Long-Term Exposure to Sulfur
Dioxide
Overall, the evidence is inadequate to infer a causal relationship between long-term
exposure to SO2 and cardiovascular health effects (Table 5-35. Section 5.3.2). The
relationship between long-term SO2 exposure and cardiovascular outcomes was not
evaluated in the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d). Despite a number of
epidemiologic studies that report positive associations of long-term exposure to SO2
concentrations with cardiovascular disease and stroke, the evidence for any one endpoint
is limited and inconsistent. Exposure measurement error and the potential for copollutant
confounding are uncertainties in the interpretation of the evidence. Additionally, the
experimental evidence is insufficient to provide coherence or biological plausibility for
an independent effect of long-term exposure to SO2 on cardiovascular health.
1.6.2.3 Reproductive and Developmental Effects
Overall the evidence is inadequate to infer a causal relationship between exposure to SO2
and reproductive and developmental outcomes (Table 5-38. Section 5.4). consistent with
the conclusion reached in the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d).
There are several recent well-designed, well-conducted studies that indicate an
association between SO2 and reproductive and developmental health outcomes, including
fetal growth metrics, preterm birth, birth weight, and fetal and infant mortality. However,
a number of uncertainties are associated with the observed relationship between exposure
to SO2 and birth outcomes, such as timing of exposure windows, exposure error, and
spatial and temporal heterogeneity. Few studies have examined other health outcomes,
such as fertility, effects on pregnancy (e.g., pre-eclampsia, gestational diabetes), and
developmental effects, and there is little coherence or consistency among epidemiologic
and toxicological studies for these outcomes. There is limited toxicological evidence at
relevant dose ranges of SO2, making it difficult to evaluate the potential modes of action
for reproductive and developmental effects of SO2. Studies published since the 2008 SOx
ISA (U.S. EPA. 2008d) have not substantially reduced any of the uncertainties identified
1-20
-------
in the previous ISA, including exposure measurement error and the potential for
copollutant confounding; therefore, the evidence is inadequate to infer a causal
relationship between exposure to SO2 and reproductive and developmental outcomes.
1.6.2.4 Total Mortality Associated with Short-Term Exposure to Sulfur Dioxide
Multicity studies evaluated since the completion of the 2008 ISA for Sulfur Oxides
continue to provide consistent evidence of positive associations between short-term SO2
exposures and total mortality (Section 5.5. IV Although the body of evidence is larger
than at the time of the last review, key uncertainties and data gaps still remain, which
contribute to the conclusion that the evidence for short-term SO2 exposures and total
mortality is suggestive of, but not sufficient to infer, a causal relationship (Table 5-41).
This conclusion is consistent with that reached in the 2008 SOx ISA (U.S. EPA. 2008d').
Overall, recent multicity studies have further informed key uncertainties and data gaps in
the S02-mortality relationship identified in the 2008 SOx ISA including confounding,
modification of the S02-mortality relationship, potential seasonal differences in
S02-mortality associations, and the shape of the S02-mortality C-R relationship.
However, questions remain regarding whether SO2 has an independent effect on
mortality, and these lingering questions can be attributed to the limited number of studies
that examined potential copollutant confounding, the relative lack of copollutant analyses
with PM2 5, and the evidence indicating attenuation of S02-mortality associations in
copollutant models with NO2 and PM10. Additionally, a biological mechanism has not
been characterized to date that could lead to mortality as a result of short-term SO2
exposures.
1.6.2.5 Total Mortality Associated with Long-Term Exposure to Sulfur Dioxide
The overall evidence is inadequate to infer a causal relationship between long-term
exposure to SO2 and total mortality among adults (Table 5-43, Section 5.5.2), consistent
with the conclusion reached in the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d').
Recent evidence is generally consistent with the evidence included in the 2008 SOx ISA,
although some recent cohort epidemiologic studies provide evidence for improved
consistency in the association between long-term exposure to SO2 and both respiratory
and total mortality. However, none of these recent studies help to resolve the
uncertainties identified in the 2008 SOx ISA related to exposure measurement error,
copollutant confounding, or the geographic scale of the analysis.
1-21
-------
1.6.2.6 Cancer
The overall evidence for long-term SO2 exposure and cancer is inadequate to infer a
causal relationship (Table 5-44. Section 5.6). the same conclusion reached in the 2008
ISA for Sulfur Oxides (U.S. EPA. 2008d). Recent studies include evidence on lung
cancer as well as other cancer types. Although some studies of SO2 concentrations and
lung cancer mortality have reported null results, other studies that included various
cofounders and copollutants reported positive associations. Positive associations were
also observed in a study of SO2 concentrations and bladder cancer mortality but not in
ecological studies of bladder cancer incidence. Limited supportive evidence for mode of
action is available from genotoxicity and mutagenicity studies, but animal toxicological
studies provide no coherence with epidemiologic findings.
1-22
-------
Table 1-1 Key evidence contributing to causal determinations for sulfur dioxide exposure and health effects
evaluated in the current Integrated Science Assessment for Sulfur Oxides.
Health Effect Category,3 Exposure Duration6, and Causal Determination
SO2 Concentrations
Associated with Effects
Respiratory Effects and Short-Term Exposure (Section 5.2.1): Causal relationship
No change in causal determination from the 2008 SOx ISA (U.S. EPA. 2008d); new evidence is consistent with previous determination.
Key evidence
(Table 5-21)
Strongest evidence is for effects on asthma exacerbation. There is consistent evidence from multiple
high-quality controlled human exposure studies ruling out chance, confounding, and other biases.
These studies show decreased lung function and increased respiratory symptoms following peak
exposures of 5-10 min in exercising individuals with asthma. Additional consistent evidence from
multiple high-quality epidemiologic studies at relevant SO2 concentrations shows an increase in asthma
hospital admissions and ED visits in single- and multicity studies and in studies examining individuals of
all ages, including children and older adults. These associations are generally unchanged in copollutant
models involving PM and other criteria pollutants. Additionally, there is some supporting epidemiologic
evidence of associations with respiratory symptoms among children with asthma. Evidence is available
for activation of sensory nerves in the respiratory tract resulting in a neural reflex and/or inflammation
leading to bronchoconstriction and allergic inflammation leading to increased airway responsiveness.
Enhanced allergic sensitization, allergic inflammation, and airway responsiveness were observed in
guinea pigs exposed to SO2 repeatedly over several days and subsequently sensitized and challenged
with an allergen. This evidence represents key events orendpoints in the proposed mode of action
linking short-term SO2 exposure and asthma exacerbation.
Overall study means:
Controlled human exposure
studies of decreased lung
function: 200-600 ppb, with
a subset analysis of
responders showing
statistically significant
responses at 300 ppb
Controlled human exposure
studies of increased
respiratory symptoms:
400-1,000 ppb
Epidemiologic studies:
1-h max: 9.6-11 ppb
24-h avg: 1.0-37 ppb
Animal studies:
100 ppb
Respiratory Effects and Long-Term Exposure (Section 5.2.2): Suggestive of. but not sufficient to infer, a causal relationship
Change in causal determination from the 2008 SOx ISA (U.S. EPA. 2008d) (inadequate to infer a causal relationship) due to new, but limited, evidence, although
uncertainty remains regarding the potential for copollutant confounding.
Key evidence0
(Table 5-24)
Evidence from epidemiologic studies is generally supportive but not entirely consistent for increases in
asthma incidence and prevalence related to SO2 exposure. Uncertainty remains regarding potential
copollutant confounding, so chance, confounding, and other biases cannot be ruled out. The limited
animal toxicological evidence provides biological plausibility and coherence across lines of evidence.
There is some evidence for a mode of action involving inflammation and allergic sensitization.
Overall epidemiologic study
means:
2-4 ppb
Animal toxicological studies:
2,000 ppb
1-23
-------
Table 1-1 (Continued): Key evidence contributing to causal determinations for sulfur dioxide exposure and health
effects evaluated in the current Integrated Science Assessment for Sulfur Oxides.
Health Effect Category,3 Exposure Duration6, and Causal Determination
SO2 Concentrations
Associated with Effects
Cardiovascular Effects and Short-Term Exposure (Section 5.3.1) Inadequate to infer a causal relationship
No change in causal determination from the 2008 SOx ISA (U.S. EPA. 2008d); new evidence is consistent with previous determination.
Key evidence0
(Table 5-311
There is some evidence of increased hospital admissions and ED visits among adults for IHD, Ml, and
all CVD; coherence with ST-segment depression in adults with pre-existing coronary heart disease; and
increased risk of cardiovascular mortality. However, results are inconsistent across outcomes, and the
associations are generally attenuated after copollutant adjustment. Recent studies have not reduced
uncertainties identified in the previous ISA, including exposure measurement error and copollutant
confounding. There is insufficient evidence from epidemiologic panel studies and experimental studies
for clinical cardiovascular effects and to identify key events in a mode of action linking short-term SO2
exposure and cardiovascular effects.
Overall epidemiologic study
24-h avg means:
1.2-30 ppb
Cardiovascular Effects and Long-Term Exposure (Section 5.3.2) Inadequate to infer a causal relationship
Not included in the 2008 SOx ISA (U.S. EPA. 2008d).
Key evidence0
(Table 5-35)
Results of epidemiologic studies of long-term SO2 concentrations and Ml, CVD, and stroke events are
limited and inconsistent. There is limited coherence with evidence for cardiovascular mortality and weak
evidence to identify key events in a mode of action linking long-term SO2 exposure and cardiovascular
effects. Recent studies have not reduced uncertainties identified in the previous ISA, including exposure
measurement error and copollutant confounding.
Overall epidemiologic study
means: 1.3-1.7 ppb
Reproductive and Developmental Effects and Exposure (Section 5.4) Inadequate to infer a causal relationship
No change in causal determination from the 2008 SOx ISA (U.S. EPA. 2008d); new evidence is consistent with previous determination.
Key evidence0
(Table 5-38)
Consistent positive associations are observed with near-birth exposures to SO2 and preterm birth.
Although limited evidence is available, positive associations are also reported for fetal growth metrics,
birth weight, and infant and fetal mortality. There is insufficient evidence from epidemiologic studies to
support an association of SO2 exposure with detrimental effects on fertility or pregnancy. Thus, the
available studies are of insufficient consistency across outcomes. Recent studies have not reduced
uncertainties identified in the previous ISA, including exposure measurement error and copollutant
confounding. Limited evidence is available for an understanding of key reproductive and developmental
events in mode of action.
Overall epidemiologic study
means: 1.9-13 ppb
1-24
-------
Table 1-1 (Continued): Key evidence contributing to causal determinations for sulfur dioxide exposure and health
effects evaluated in the current Integrated Science Assessment for Sulfur Oxides.
SO2 Concentrations
Health Effect Category,3 Exposure Duration6, and Causal Determination Associated with Effects
Total Mortality and Short-Term Exposure (Section 5.5.1) Suggestive of. but not sufficient to infer, a causal relationship
No change in causal determination from the 2008 SOx ISA (U.S. EPA. 2008d); new evidence is consistent with previous determination.
Key evidence0 There is consistent epidemiologic evidence from multiple high-quality studies at relevant SO2 Overall epidemiologic study
(Table 5-41) concentrations demonstrating increases in mortality in multicity studies conducted in the U.S., Canada, 24-h avg means:
Europe, and Asia. There is limited coherence and biological plausibility with cardiovascular and j 5 Canada South
respiratory morbidity evidence and uncertainty regarding a biological mechanism that would explain the America Europe¦
continuum of effects leading to SC>2-related mortality; thus, chance, confounding, and other biases g 4-28 ppb
cannot be ruled out. „ .
Asia:
0.7->200 ppb
Total Mortality and Long-Term Exposure (Section 5.5.2) Inadeguate to infer a causal relationship
No change in causal determination from the 2008 SOx ISA (U.S. EPA. 2008d); new evidence is consistent with previous determination.
Key evidence0 Some epidemiologic studies report positive associations, but results are not entirely consistent, with Overall epidemiologic study
(Table 5-43) some studies reporting null associations. Additionally, there is no evidence for associations between means:
SO2 exposure and long-term respiratory or cardiovascular health effects to support an association with 1.6-24 ppb
mortality from these causes. Recent studies have not reduced uncertainties identified in the previous
ISA, including exposure measurement error and copollutant confounding.
1-25
-------
Table 1-1 (Continued): Key evidence contributing to causal determinations for sulfur dioxide exposure and health
effects evaluated in the current Integrated Science Assessment for Sulfur Oxides.
SO2 Concentrations
Health Effect Category,3 Exposure Duration6, and Causal Determination Associated with Effects
Cancer and Long-Term Exposure (Section 5.6) Inadequate to infer a causal relationship
No change in causal determination from the 2008 SOx ISA (U.S. EPA. 2008d); new evidence is consistent with previous determination.
Key evidence0 Among a small body of evidence, some epidemiologic studies report associations in lung cancer and Overall epidemiologic study
(Table 5-44) bladder cancer mortality. There is also some evidence identifying mutagenesis and genotoxicity as key means: 1.5-28 ppb.
events in a proposed mode of action linking long-term SO2 exposure and cancer; however, toxicological j0xicological studies' 5 000
studies provide limited coherence with epidemiologic studies. Recent studies have not reduced <|g 700 21 400 32 100 ppb
uncertainties identified in the previous ISA, including exposure measurement error and copollutant
confounding.
CVD = cardiovascular disease; ED = emergency department; IHD = ischemic heart disease; ISA = Integrated Science Assessment; Ml = myocardial infarction; PM = particulate
matter; S02 = sulfur dioxide; SOx = sulfur oxides.
aA large spectrum of outcomes is evaluated as part of a broad health effect category including physiological measures (e.g., airway responsiveness, lung function), clinical outcomes
(e.g., respiratory symptoms, hospital admissions), and cause-specific mortality. Total mortality includes all nonaccidental causes of mortality and is informed by the nature of the
evidence for the spectrum of morbidity effects (e.g., respiratory, cardiovascular) that can lead to mortality. The sections and tables referenced include a detailed discussion of the
available evidence that informed the causal determinations.
bShort-term exposure refers to time periods of minutes up to 1 mo, while long-term exposures are more than one mo to years.
Uncertainties remain for many of the studies included as key evidence. Uncertainty remains in some epidemiologic studies. Exposure assessments in epidemiologic studies using
fixed-site monitors may not fully capture spatial variability of S02. Spatial and temporal heterogeneity may introduce exposure error in long-term effects. For studies of reproductive
and developmental outcomes, associations with exposure to S02 at particular windows during pregnancy are inconsistent between studies. Additionally, although S02 is generally
poorly to moderately correlated with other National Ambient Air Quality Standards pollutants at collocated monitors, copollutant confounding by these and other pollutants cannot be
ruled out.
1-26
-------
1.7 Policy-Relevant Considerations
As described in the Preamble to the ISAs (U.S. EPA. 2015b) and Section 1.1. this ISA
informs policy-relevant issues that are aimed at characterizing relationships between SO2
exposure and health effects and the impact of these relationships on public health. To that
end, this section integrates information from the ISA to describe SO2 exposure durations
and patterns related to health effects, the shape of the concentration-response
relationship, regional heterogeneity in relationships, the adverse nature of health effects,
and at-risk populations and lifestages. In addressing these policy-relevant issues, this
section focuses on respiratory effects associated with short-term exposures, for which the
evidence indicates there is a causal relationship.
1.7.1 Durations and Lag Structure of Sulfur Dioxide Exposure Associated
with Health Effects
Effects have been observed in controlled human exposure studies after SO2 exposures
during exercise as brief as 5-10 minutes with limited evidence for increased airway
responsiveness to subsequent allergen challenge for at least 48 hours following SO2
exposure in combination with a copollutant (i.e., NO2). Consistent associations between
SO2 concentrations and asthma hospital admissions and ED visits that are generally
unchanged in copollutant models have been demonstrated in epidemiologic studies using
daily exposure metrics (24-h avg and 1-h daily max), although the observed effects could
be related to very short duration (5-10 minutes) peak exposures experienced during the
day.
Regarding the lag in effects, the findings from controlled human exposure studies provide
evidence of a rapid onset of effects. The limited number of epidemiologic studies that
examined lag structures reported associations within the first few days of exposure.
1.7.2 Concentration-Response Relationships and Thresholds
Characterizing the shape of concentration-response relationships for health effects
associated with SO2 helps quantify the public health impact of SO2 exposure. A key issue
is often whether the relationship is linear across the full range of policy-relevant
concentrations or whether there are deviations from linearity, and if so, at what
concentrations they occur. Another important issue is whether there is evidence of a
1-27
-------
potential threshold, indicating exposures below which adverse health outcomes are not
observed. The lack of a discernable threshold in the evidence for the health effects of
interest (i.e., respiratory effects associated with short-term exposure) precludes the
identification of an exposure level below which there is no risk of effects.
Both controlled human exposure and epidemiologic studies provide some information
with respect to the concentration-response relationship between SO2 exposures and
respiratory effects. Results from controlled human exposure studies indicate wide
interindividual variability in response to SO2 exposures, with peak (5-10 minutes)
exposures at concentrations as low as 200-300 ppb eliciting lung function decrements in
some individuals with asthma. A clear increase in the magnitude of lung function
decrements was observed with increasing exposure concentrations between 200 and
1,000 ppb during 5-10 minute SO2 exposures.
Controlled human exposure studies provide information on the direct relationship
between exposure and response over short-duration exposures (i.e., 5-10 minutes) and
changes in response with different exposure concentrations. Epidemiologic studies
evaluate whether the risk of respiratory effects changes at different ambient
concentrations, and these studies are limited to consideration of longer exposure
durations (i.e., 1-hr daily max and 24-h avg). The few epidemiologic studies that focus on
the SCh-respiratory effects concentration-response relationship examine pediatric asthma
ED visits. The limited epidemiologic evidence to date does not provide evidence for a
deviation from linearity or a discernable population-level threshold in the range of
ambient concentrations typically observed. However, epidemiologic studies have not
been conducted that provide a thorough empirical evaluation of alternatives to linearity.
The interpretation of epidemiologic study results is further complicated by potential
measurement error due to spatial and temporal variability in SO2 concentrations.
1.7.3 Regional Heterogeneity in Effect Estimates
The 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d') discussed spatial variability in SO2
concentrations and its impact on effect estimates from epidemiologic studies.
Correlations between monitors ranged from very low to very high, suggesting that SO2
concentrations at some monitoring sites may not be highly correlated with the community
average concentration. Of particular concern for SO2 is that most emissions are from
point sources, resulting in an uneven distribution of SO2 concentrations across an urban
area. Factors contributing to differences among monitoring sites include source
characteristics (e.g., stack height), proximity to sources, terrain features, and uncertainty
regarding the measurement of low SO2 concentrations.
1-28
-------
Inability to fully characterize spatial and temporal variability in SO2 concentrations can
contribute to exposure error in epidemiologic studies, whether such studies rely on
fixed-site monitor data or concentration modeling for exposure assessment. Studies using
24-h avg concentrations may not capture short-term peak exposures known to produce
health effects in controlled human exposure studies. SO2 has low to moderate spatial
correlations between ambient air monitoring sites across urban geographic scales; thus,
using fixed-site monitor data for epidemiologic exposure assessment introduces exposure
error into the resulting effect estimate. Spatial variability in the magnitude of
concentrations may affect cross-sectional and large-scale cohort studies by undermining
the assumption that intraurban concentration and exposure differences are less important
than interurban differences. This issue may be less important for time-series studies,
which rely on day-to-day temporal variability in concentrations to evaluate health effects.
Low correlations between monitors contribute to exposure error in time-series studies,
including bias toward the null and wider confidence intervals.
1.7.4 Public Health Significance
The public health significance of air pollution-related health effects is informed by the
adverse nature of the health effects that are observed, the size of the population exposed
to the air pollutant or affected by the health outcome, and the presence of populations or
lifestages with higher exposure or increased risk of air pollution-related health effects.
1.7.4.1 Characterizing Adversity of Health Effects
Both the World Health Organization (WHO) and the American Thoracic Society (ATS)
have provided guidance in describing what health effects may be considered adverse.
WHO defines health as "the state of complete physical, mental, and social well-being and
not merely the absence of disease or infirmity" (WHO. 1948). By this definition, changes
in health outcomes that are not severe enough to result in a diagnosis of a clinical effect
or condition can be considered adverse if they affect the well-being of an individual. ATS
also has considered a wide range of health outcomes in defining adverse effects.
Distinguishing between individual and population risk, ATS described its view that small
air pollution-related changes in an outcome observed in individuals might be considered
adverse on a population level. This is because a shift in the distribution of population
responses resulting from higher air pollution exposure might increase the proportion of
the population with clinically important effects or at increased risk of a clinically
important effect that could be caused by another risk factor (ATS. 2000b). Increases in
ambient SO2 concentrations are associated with a broad spectrum of health effects related
1-29
-------
to asthma, including those characterized as adverse by ATS such as ED visits and
hospital admissions.
1.7.4.2 At-Risk Populations and Lifestages for Health Effects Related to Sulfur
Dioxide Exposure
The primary NAAQS are intended to protect public health with an adequate margin of
safety. In so doing, protection is provided for both the population as a whole and those
groups potentially at increased risk for health effects from exposure to the air pollutant
for which each NAAQS is set (Preface to this ISA). Hence, the public health significance
of health effects related to SO2 exposure also is informed by whether specific lifestages
or groups in the population are identified as being at increased risk of SC>2-related health
effects.
At-risk populations or lifestages can be characterized by specific biological,
sociodemographic, or behavioral factors, among others. Since the 2008 ISA for Sulfur
Oxides (U.S. EPA. 2008d'). the U.S. EPA has used a framework for drawing conclusions
about the role of such factors in modifying risk of health effects of air pollution exposure
[Table III of the Preamble to the ISAs (U.S. EPA. 2015b)!. Similar to the causal
framework, conclusions about at-risk populations are based on judgments of the
consistency and coherence of evidence within and across disciplines (Chapter 6V Briefly,
the evaluation is based on studies that compared exposure or health effect relationships
among groups that differ according to a particular factor (e.g., people with and without
asthma) and studies conducted in a population or animal model with a particular factor or
pathophysiological condition. Where available, information on exposure, dosimetry, and
modes of action is evaluated to assess coherence with health effect evidence and inform
how a particular factor may contribute to SCh-related risk of health effects (e.g., by
increasing exposure, increasing biological effect for a given dose).
There is adequate evidence that people with asthma are at increased risk for SCh-related
health effects (Section 6.3.1). which is consistent with the findings of the 2008 ISA for
Sulfur Oxides (U.S. EPA. 2008d). The conclusions are based on findings for short-term
SO2 exposure and respiratory effects (specifically lung function decrements), for which a
causal relationship has been determined (Section 5.2.1.9). There are a limited number of
epidemiologic studies evaluating S02-related respiratory effects that include stratification
by asthma status, but there is evidence for respiratory-related hospital admissions and ED
visits (Section 5.2.1.2). Further support for increased risk in individuals with asthma is
provided by biological plausibility drawn from modes of action. Children with asthma
may be particularly at increased risk relative to adults with asthma due to their increased
responsiveness to methacholine, increased ventilation rates relative to body mass,
1-30
-------
increased time spent outdoors, and increased proportion of oral breathing, particularly
among boys. Among children in the U.S., asthma is the leading chronic illness (9.5%
prevalence) and largest reason for missed school days.
There is also evidence suggestive of increased risk for children and older adults relative
to other lifestages (Section 6.5.IV Although the 2008 ISA for Sulfur Oxides (U.S. EPA.
2008d') discussed several studies indicating stronger associations between SO2 and
respiratory outcomes for these lifestages, the recent evidence is not entirely consistent
with previous studies. For children, studies comparing SCh-associated respiratory
outcomes reported mixed results. For adults, recent evidence generally found similar
associations for SCh-related respiratory outcomes or mortality across age groups,
although those over 75 years of age were more consistently at increased risk. In addition,
there was insufficient toxicological evidence regarding the effect of lifestage on
respiratory responses to SO2 to support observations made across epidemiologic studies
that evaluated lifestage.
1.7.4.3 Summary of Public Health Significance of Health Effects Related to Sulfur
Dioxide Exposure
Several aspects of the current evidence are important for considering the public health
significance of SCh-related health effects. One aspect is adversity of the health effects,
which may include health effects that are clearly adverse such as ED visits and hospital
admissions for asthma and asthma exacerbation. The magnitude of the affected
population is also important. As noted above, in the case of S02-related health effects, the
potentially affected population is large, given the number of people with asthma in the
U.S. The roles of co-occurring risk factors or combined higher SO2 exposure and health
risk in influencing the risk of S02-related health effects is not well understood. The large
proportions of children and older adults in the U.S. population and the high prevalence of
asthma in children may translate into a large number of people affected by SO2, and thus,
magnify the public health impact of ambient SO2 exposure.
1.8 Summary and Health Effects Conclusions
This ISA is a comprehensive evaluation and synthesis of the policy-relevant science
regarding the potential health effects of sulfur oxides in the ambient air, focusing on SO2.
The ISA development process involves review of the scientific literature, selecting and
evaluating relevant studies, and evaluating the weight of evidence to reach causal
determinations regarding the likelihood of independent health effects of SO2. Information
1-31
-------
is included in the ISA on sources of SO2, atmospheric chemistry of SO2 and other
sulfur-containing compounds, ambient air concentrations of SO2 nationwide and in urban
areas, and modeling approaches for estimating SO2 concentrations. Approaches for
characterizing exposure to ambient SO2, including monitoring and modeling, together
with factors affecting ambient exposure, are described in terms of their potential impact
on epidemiologic study results. Dosimetry of SO2 and potential modes of action are
discussed to provide context in considering the potential health effects of SO2, including
respiratory effects, cardiovascular effects, reproductive and developmental effects,
cancer, and mortality.
Consistent with the findings of the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d). recent
studies support the conclusion that there is a causal relationship between short-term SO2
exposure and respiratory effects. This causal determination is based on consistency of
findings within disciplines, coherence among multiple lines of evidence, and biological
plausibility indicating that there is a causal relationship between short-term SO2 exposure
and respiratory effects in individuals with asthma. The primary evidence for this
conclusion comes from controlled human exposure studies that showed lung function
decrements and respiratory symptoms in individuals with asthma exposed to SO2 for
5-10 minutes under increased ventilation conditions. Supporting evidence was provided
by epidemiologic studies that reported positive associations between short-term SO2
exposures and asthma hospital admissions and ED visits that were generally unchanged
in copollutant models involving PM and other criteria pollutants.
For both long-term exposure and respiratory effects, as well as short-term exposure and
total mortality, the evidence is suggestive of, but not sufficient to infer, a causal
relationship. In both cases, there is some evidence of an association between SO2
exposure and health outcomes, but the evidence is inconsistent and uncertainties remain,
including exposure error and copollutant confounding. The evidence was considered to
be inadequate to infer a causal relationship for other health effects, including
cardiovascular morbidity (short- and long-term exposure), reproductive and
developmental effects, total mortality (long-term exposure), and cancer. For these
outcome categories, the evidence generally was not consistent across specific outcomes,
showed a potential for copollutant confounding, and was lacking in biological
plausibility.
In considering the effects of SO2 on various populations and lifestages, there is adequate
evidence that people with asthma are at increased risk for SC>2-related health effects, as
well as suggestive evidence for increased risk among children and older adults. The large
proportions of children and older adults in the U.S. population and the high prevalence of
1-32
-------
asthma in children may translate into a large number of people affected by SO2, and thus,
magnify the public health impact of ambient SO2 exposure.
1-33
-------
CHAPTER 2 ATMOSPHERIC CHEMISTRY AND
AMBIENT CONCENTRATIONS OF
SULFUR DIOXIDE AND OTHER
SULFUR OXIDES
2.1 Introduction
Sulfur oxides, in the context of the National Ambient Air Quality Standards (NAAQS),
are a group of closely related sulfur-containing gas-phase compounds [e.g., sulfur dioxide
(SO2), sulfur monoxide (SO), disulfur monoxide (S2O), and sulfur trioxide (SO3)]. Sulfur
oxides also appear in the particle phase, as components of particulate matter (PM), and
particle phase sulfur compounds are discussed separately as part of the Integrated Science
Assessment for PM (U.S. EPA. 2016c).
The NAAQS for SOx are currently set using SO2 as the indicator species. Of the sulfur
oxides, SO2 is the most abundant in the atmosphere, the most important in atmospheric
chemistry, and the one most clearly linked to human health effects (U.S. EPA. 2008d.
1996b; HEW. 1969). Therefore, the emphasis in this chapter is on SO2. Note that the
mechanism of particle-phase SO42 formation is briefly described in Section 2.3 [for more
detail, see Seinfeld and Pandis (2006). Finlavson-Pitts and Pitts (2000). and other
atmospheric chemistry texts]. The health effects of sulfate and other sulfur compounds in
the particle phase are discussed in the ISA for Particulate Matter (U.S. EPA. 2009a).
Sulfur dioxide is both a primary gas-phase pollutant (when formed during fuel
combustion) and a secondary pollutant [the product of atmospheric gas- or aqueous-phase
oxidation of reduced sulfur compounds (sulfides)]. Fossil fuel combustion is the main
anthropogenic source of primary SO2, while volcanoes and landscape fires (wildfires as
well as prescribed burns) are the main natural sources of primary SO2. Industrial
chemical and pulp and paper production, smelter and steel mill operations, natural
biological activity (plants, fungi, and prokaryotes), and volcanoes are among many
sources of reduced sulfur compounds that ultimately lead, through various oxidation
reactions in the atmosphere, to the formation of secondary SO2.
This chapter describes the main sources of sulfur dioxide and other gas-phase sulfur
oxides found in the atmosphere, as well as reduced sulfur gases that serve as precursors
for SO2 (Section 2.1). source emissions (Section 2.2). atmospheric chemistry and fate
(Section 2.3). measurement methods (Section 2.4). environmental concentrations
(Section 2.5). and atmospheric modeling of sulfur dioxide concentrations (Section 2.6).
This material is provided as a prologue for detailed discussions on exposure and health
2-1
-------
effects evidence in the subsequent chapters, and as a source of information to help
interpret that evidence in the context of relevant ambient concentrations.
2.2 Anthropogenic and Natural Sources of Sulfur Dioxide
This section briefly describes the main U.S. anthropogenic and natural sources of SO2
emissions. Emissions estimates for natural and anthropogenic sulfide emissions for the
U.S. alone are not available in the literature. Therefore, a brief discussion of the sulfur
cycle and estimates of the contribution of sulfides at the global scale, all of which can be
found in the literature, are provided. Section 2.2.1 describes the main categories of
anthropogenic SO2 emissions, while Section 2.2.2 presents the geographic distribution of
SO2 sources across the U.S. The declining trend in anthropogenic SO2 emissions is
discussed in Section 2.2.3. Natural sources of SO2 are discussed in Section 2.2.4. Indirect
production of SO2 through oxidation of reduced sulfur compounds emitted from geologic
and biological sources is discussed in Section 2.2.5.
Sulfur is present to some degree in all fossil fuels, especially coal, and occurs as reduced
organosulfur compounds. Coal also contains sulfur in mineral form (pyrite or other
metallo-sulfur minerals) and in elemental form (Calkins. 1994V Of the most common
types of coal (anthracite, bituminous, subbituminous, and lignite), sulfur content varies
between 0.4 and 4% by mass. Sulfur in fossil fuels is almost entirely converted to SO2 (or
SO3) during combustion, making accurate estimates of SO2 combustion emissions
possible based on fuel composition and combustion rates.
The mass of sulfur released into the environment by anthropogenic sources is comparable
to natural sources (Brimblecombe. 2003). In addition to volcanic and other geologic SO2
emissions, naturally occurring SO2 is derived from the oxidation of sulfides emitted by
low flux "area" sources, such as the oceans and moist soils. Anthropogenic emissions of
sulfur are primarily in the form of SO2, emerging from point sources in quantities that
substantially affect local and regional air quality.
2.2.1 U.S. Anthropogenic Sources
The largest S02-emitting sector within the U.S. is electricity generation based on coal
combustion (3,224,087 tons), although emissions from this sector have been declining in
recent years due to fuel substitution and emissions controls. The mass of emissions
produced by the fuel combustion in electrical utilities sector [i.e., coal-fired electric
generating units (EGUs)] exceeds those produced by the next largest sector [the fuel
2-2
-------
combustion—industrial sector (i.e., coal-fired boilers)] by nearly a factor of 5, and EGUs
emit approximately twice as much SO2 as all other sources combined. Figure 2-1
provides a sector comparison of annual emissions [in tons] found in the U.S. EPA 2014
National Emissions Inventory (NEI) (U.S. EPA. 2016b).
COMB = combustion; ELEC = electric; MFG = manufacturing; UTIL = utilities.
Note: "Fuel Comb. —Other" includes commercial, institutional, and residential sources. Metals Processing" includes copper smelting
(22,792 tpy S02) and combined iron and steel mill (28,247 tpy S02) facilities.
Source: https://www.epa.aov/air-emissions-inventories/air-pollutant-emissions-trends-data (U.S. EPA. 2016b).
Figure 2-1 Sulfur dioxide emissions by sector in tons, 2014.
Because EGUs comprise the largest NEI source category, the spatial distribution of
SCh-emitting EGUs is presented here (U.S. EPA. 2016b). Most EGU sources are located
in the eastern half of the continental U.S., as indicated in Figure 2-2. There is a
particularly high concentration of EGUs in the Ohio River valley, upper Midwest, and
2-3
-------
along the Atlantic coast. Many of the monitoring sites with elevated SO2 concentrations
are located in these same areas (Section 2.5.2).
.S. Virgin Island:
Alaska
S02 Tons Emitted from EGU
Facilities
Hawaii
Kilometers
Note: EGU = electric power generating unit; S02 = sulfur dioxide.
Source: https://www.6Da.aov/air~6missiQns-inventories: fll.S. EPA. 2016b').
Figure 2-2 Distribution of electric power generating units (438), scaled
according to annual sulfur dioxide emissions across the U.S.
based on the 2014 National Emissions Inventory.
Industrial fuel combustion is the second largest source nationwide, emitting 656,901 tpy.
followed by other fuel combustion (172,406 tpy). Miscellaneous (148,898 tpy) is the
fourth-largest source and includes SO2 emissions by fire used in landscape management
and agriculture as well as wildfires (U.S. EPA. 2016b). Wildfires, as a natural source of
SO2 emissions, are discussed in Section 2.2.4.3. The metals processing sector includes
2-4
-------
copper smelting (22,792 tpy SO2) and combined iron and steel mill (28,247 tpy SO2)
facilities. Highway vehicles emit <1% of the combined emissions shown in Figure 2-1.
The commercial marine sector falls within the off-highway category (75,712 tpy) (U.S.
EPA. 2016b). Using data from 2002, Wang et al. (2007) modeled SO2 emissions from
commercial marine activity by combining historical shipping data and marine traffic
predictions based on port sizes and probable routes. Within a 200-nautical-mile boundary
around the marine, lake, and river international borders of the U.S., the study authors
estimated that 38% of emissions related to commercial marine shipping occurred along
the East Coast of the U.S. Twenty percent of emissions were estimated for the West
Coast and 26% for the Gulf Coast. Smaller quantities were estimated elsewhere (10% for
Alaska, 3% for Hawaii, and 2% for the Great Lakes). Interior waterway activity was not
included in the Wang et al. (2007) study. In 2010, the International Maritime
Organization introduced Emissions Control Areas (ECA) around U.S., Canadian, and
French waters under the International Convention for the Prevention of Pollution from
Ships (Office of Transportation and Air Quality. 2010). The ECA is a 200-nautical-mile
buffer around the maritime borders, within which fuels cannot contain more than
1,000 ppm sulfur as of 2015. These reductions are expected to be accomplished by
having maritime vessels switch fuel sources when crossing the 200-nautical-mile buffer
to approach their port. The U.S. EPA's Office of Transportation and Air Quality (2010)
estimated that this reduction in the amount of sulfur in marine fuels used within the
200 nautical mile buffer would result in an 85% reduction in SO2 emissions from the
commercial marine sector. [Monitors located at the Port of Los Angeles and the Port of
Long Beach reflect these reductions, with latest reports from these two ports showing
SO2 concentrations well below the NAAQS (Leidos Inc. 2016)1.
National SO2 emissions sector summaries cannot offer insight concerning the local
influence of individual S02-emitting facilities. Although they may be fewer in number
than fossil fuel-fired EGUs, other types of large emissions facilities that may
substantially impact local air quality include copper smelters, kraft pulp mills, Portland
Cement plants, iron and steel mill plants, sulfuric acid plants, petroleum refineries, and
chemical processing plants. For example, the metals processing sector represents less
than 2.3% of total emissions from the 2014 NEI (U.S. EPA. 2016b). but monitoring sites
that have recorded some of the highest 1-hour daily max SO2 concentrations in the U.S.
are located near copper smelters in Arizona
(Sections 2.5.2 and 2.5.4; Figure 2-11).
2-5
-------
2.2.2
National Geographic Distribution of Large Sources
Some industrial facilities are complex sources, with S02 emissions related to multiple
processes. Figure 2-3 shows the geographic distribution of continental U.S. facilities
emitting more than 1,000 tpy SO2, with an enlargement of the midwestern states
including the Ohio River Valley, where a large number of these S02-emitting sources are
located.
U.S. EPA Sulfur Dioxide Data Requirements Rule
Another information resource regarding air quality near large sources of SO2 is the data
produced by air quality monitors required by the SO2 Data Requirements Rule (40 CFR
51.1202-51.1203; 80 FR50152, August 21, 2015), which was enacted in support of the
SO2 NAAQS. This regulation requires that, at a minimum, air agencies must characterize
air quality around sources that emit 2,000 tons per year or more of SO2. An air agency
may avoid the requirement for air quality characterization near a source by adopting
enforceable emission limits that ensure that the source will not emit more than 2,000 tpy.
This final rule gives air agencies the flexibility to characterize air quality using either
modeling of actual source emissions or using appropriately sited ambient air quality
monitors. Under this data requirements rule, in 2016, air agencies submitted to their
relevant U.S. EPA Regional Administrator a final list identifying the sources in the state
around which SO2 air quality is to be characterized. The list included sources with
emissions above 2,000 tpy SO2. The U.S. EPA Regional Offices or air agencies included
additional sources on this list that they deemed necessary. The national list contained
377 sources (https://www.epa.gov/so2-pollution/so2-data-requirements-rule-source-list').
Figure 2-4 shows the locations of those sources.
2-6
-------
(tons)
Facilities Emitting 1,000-150,000 Tons of S02
in NEI 2014
S02
U.S.
2010
Emissions by Facility
1,000 to 5,000
5,000 to 50,000
50,000 to 150,000
Counties
Population Density (persons per mi2)
0.0 to 1.0
1.0 to 20.0
20.0 to 85.0
| 85.0 to 500
| 500 to 2,000
I 2,000 to 70,000
Facilities in Midwest United States Emitting 1,000-150,000 Tons of S02
in NEI 2014
Figure 2-3 Geographic distribution of (top) continental U.S. facilities (526)
emitting more than 1,000 tpy SO2, with (bottom) an enlargement of
the midwestern states, including the Ohio River Valley, where a
large number of these sources are concentrated.
S02 Emissions by Facility (tons)
• 1,000 to 5,000
• 5,000 to 50,000
m 50,000 to 150,000
U.S. Counties
2010 Population Density (persons per mi2)
0.0 to 1-0
1.0 to 20.0
20 0 to 85 0
85.0 to 500
500 to 2,000
2,000 to 70,000
Source: https://vwwtf.epa.gov/air-emissions-inventories: U.S. EPA (2016b)
2-7
-------
Sources Subject to EPA's Data Requirements Rule (DRR)
July 18, 2016
DRR = Data Requirements Rule; U.S. EPA = U.S. Environmental Protection Agency.
Source: U.S. EPA Office of Air Quality Planning and Standards.
Figure 2-4 Sulfur dioxide sources (377) identified by state/local air agencies
under the U.S. Environmental Protection Agency's Data
Requirements Rule, as of July 18, 2016.
2-8
-------
2.2.3 U.S. Anthropogenic Emission Trends
Anthropogenic emissions of SO2 in the U.S. have shown dramatic declines since the
1970s, and emissions reductions have accelerated since the 1990 amendments to the
Clean Air Act were enacted (USC Title 42 Chapter 85). Table 2-1 gives the annual SO2
emissions, percentage of the U.S. SO2 total emissions, and change in emissions rate from
2004 to 2014. Figure 2-5 illustrates the emissions trends by sector from 1970 to 2014 in
relation to the timeline over which the NAAQS for SOx and the Clean Air Act control
programs [Acid Rain Program (ARP), NOx Budget Program (NBP), on- and off-road
diesel emissions standards, and Cross-State Air Pollution Rule (CSAPR) and other
national interstate transport rules] have been implemented (U.S. EPA. 2016b). Exceptions
to the steep decline in SO2 emissions in the listed sectors are the increases in emissions
from the waste disposal and recycling sectors, commercial storage and transport sector,
and from miscellaneous sources (e.g., landscape fires). Waste disposal and recycling
contributes only 0.8% of total 2014 SO2 emissions. Landscape fires are a larger
contributor to the NEI (3%) and are discussed further in Section 2.2.4.3.
Hand etal. (2012) studied reductions in EGU-related annual SO2 emissions during the
2001-2010 period. They found that emissions decreased throughout the U.S. by 6.2% per
year, with the largest reductions in the western U.S. at 20.1% per year. The smallest
reduction (1.3% per year) occurred in the Great Plains states.
2-9
-------
Table 2-1 Summary of current U.S. Environmental Protection Agency sulfur
dioxide trends data by emissions sector. Values shown in bold
indicate increased emissions.
Source Type
Tons SO2
Emitted, 2014
Percentage of Total*
Percent Change,
2004-2014
Fuel combustion, electric utilities
3,224,087
67%
-69%
Fuel combustion, industrial
656,901
14%
-63%
Other industrial processes
172,406
3.6%
-51%
Miscellaneous
148,898
3.1%
2.6%
Fuel comb, other
130,869
2.7%
-77%
Chemical & allied product mfg
114,980
2.4%
-55%
Metals processing
107,485
2.2%
-43%
Petroleum & related industries
102,650
2.1%
-53%
Off-highway
75,712
1.6%
-87%
Waste disposal & recycling
36,689
0.8%
32%
Highway vehicles
28,658
0.6%
-86%
Storage & transport
3,439
0.1%
2.6%
2-10
-------
35,000
S02 NAAQS (1971)
30,000
25,000
20,000
15,000
10,000
5,000
ARP 1995)
NBP (2003)
On-Road Diesel
Fuel Std (2006)
Off-Road Diesel
Fuel Std (2006)
National
Interstate
Transport
Rules
(2006)
1970
1975
1980
1985
1990
1995
2000
2005
2010
« FUEL COMB. ELEC, U71L
CHEMICAL & ALU ED PRODUCT MFG
i OTHER INDUSTRIAL PROCESSES
WASTE DISPOSAL & RECYCLING
¦ MISCELLANEOUS
i FUEL COMB. INDUSTRIAL
METALS PROCESSING
i SOLVENT UTILIZATION
HIGHWAY VEHICLES
¦ FUELCOMB. OTHER
¦ PETROLEUM & RELATED INDUSTRIES
¦ STORAGE & TRANSPORT
¦ OFF-HIGHWAY
ARP = Acid Rain Program; COMB = combustion; ELEC = electric; MFG = manufacturing; NAAQS = National Ambient Air Quality
Standards; NBP = NOx Budget Program; S02 = sulfur dioxide; STD = Standards; UTIL = utilities.
Source: https://www.epa.gov/air-emissions-inventories/air-aollutant-emissions-trends-data; U.S. EPA (2016b).
Figure 2-5 National sulfur dioxide emissions trends by sector (103 tpy),
1970-2014.
2.2.4 Natural Sources
This section provides an overview of the major natural sources of SO2 and reduced sulfur
compounds that are oxidized in the atmosphere to form SO2. Section 2.2.4.1 briefly
2-11
-------
describes the elements of the global sulfur cycle. Section 2.2.4.2 briefly discusses
volcanic sources of SO2 within the U.S. Section 2.2.4.3 discusses SO2 emissions by U.S.
wildfires. The section concludes with a brief summary of both anthropogenic and natural
emissions of reduced sulfur gases that can serve as precursors to SO2.
2.2.4.1 The Global Sulfur Cycle
The total budget for sulfur, in all its forms, at Earth's surface is on the order of 1.1 x 1016
tons S (Schlcsingcr. 1997). The sulfur cycle comprises the many chemical and biological
processes that continuously interconvert the element among its four main oxidation states
(-2, 0, +4, +6). The reduced form of sulfur is present in the environment as hydrogen
sulfide, hydrogen disulfide, and a number of organic compounds. Oxidized sulfur is
present primarily as SO2 and sulfate (SO42 ).
Volcanoes and wildfires are nonbiological natural sources that directly emit SO2 to the
atmosphere. Biological natural sources, together with volcanoes, emit reduced sulfur
compounds that subsequently oxidize in the atmosphere to form SO2. Under anaerobic
conditions, various species of plants, fungi, and prokaryotes convert oxidized sulfur into
its reduced forms (Madigan et al.. 2006). Photosynthetic green and purple bacteria and
some chemolithotrophs oxidize sulfides to form elemental sulfur. Some microorganisms
oxidize elemental sulfur to form SO42 and SO2; others reduce elemental sulfur to sulfides
(1dissimilative sulfur reduction), while others are capable of reducing SO42 all the way
down to sulfide (dissimilative SO4 reduction).
2.2.4.2 Volcanoes as a Natural Source of Sulfur Dioxide
Geologic activity, including fiimaroles, geysers, and metamorphic degassing, emits a
number of gases, including SO2, carbon dioxide (CO2), hydrogen sulfide (H2S),
hydrochloric acid, and chlorine (Simpson et al.. 1999). Eruptive and noneruptive
volcanoes are the most important sources of geologic SO2 emissions. Noneruptive, but
geothermally active, volcanoes outgas at relatively constant rates and appear to be more
significant sources of SO2 than burst emissions that occur during eruptions. The
emissions from volcanic eruptions are sporadic, and therefore, vary from year to year
(Simpson et al.. 1999).
The western U.S. border of the continental United States (CONUS) is near and in some
cases over the boundary between the North American, the Pacific, and the Juan de Fuca
tectonic plates. The region is subject to ongoing volcanic activity. In Alaska, the Aleutian
2-12
-------
volcanic arc, part of the state of Alaska, comprises 75 volcanic centers. Volcanoes in this
chain have erupted once or twice per year on average over the past 100 years with
impacts on local communities (Power. 2013). Figure 2-6 shows an image derived from
data collected by the Atmospheric Infrared Sounder (AIRS) instrument aboard NASA's
Aqua satellite during the July 12-20, 2008 eruption of the Okmok Volcano in Alaska's
Aleutian Islands. The image shows sulfur dioxide at altitudes around 16 km (10 miles)
released by the volcano over that time span, with red indicating the highest
concentrations, and pale pink indicating the lowest (Prata et al.. 2010). Sulfur dioxide has
infrared absorption features at 4 and 7.3 um. which allowed Prata et al. (2010) to
calculate the total mass of SO2 emitted during the eruption as 319,670 ± 11,023 tons.
Okmok AIRS 7.3 um Cumulative S02 12-20 July, 2008
5.0
170
4.5
\ —
65
4.0
-180
60
~ 3.5
-170
OJKok
-60
ifc* ?•
2.5
o 2.0
-150
^ %¦
rto
1.0
-120
0.5
0.0
AIRS = Atmospheric Infrared Sounder; S02 = sulfur dioxide.
Source: Image courtesy of Fred Prata of the Norwegian Institute for Air Research (NILU); NASA (2008a).
Figure 2-6 Sulfur dioxide released during the July 12-20, 2008 eruption of
the Okmok Volcano in Alaska's Aleutian Islands (image derived
from data collected by the Atmospheric Infrared Sounder
instrument aboard the National Aeronautics and Space
Administration Aqua satellite).
The line of volcanoes long the western side of North American extends from the Aleutian
Islands in Alaska south and east through the states of Washington, Oregon, California,
Arizona, and New Mexico, with outlying geologically active sites in Idaho (Craters of the
2-13
-------
Moon) and Wyoming (Yellowstone). Figure 2-7 shows the geographic location and
activity potential for these sites within the continental U.S.
Bellingham
A Glacier Peak
^ WASHINGTON sPok:arie
A Mount Rainier
Mount St Helens AA MountAdams'
Portland
¦
Three Sisters .
¦V • Am Bend
Eugene .' , _ -
A Newberry Crater
Great Fa s
MONTANA
Mount Hood
A Mount Jefferson
i ¦ !
A Yellowstone
A Craters of the Moon
A Crater Lake
OREGON
Casper
DMING
Pocate o
Cheyenne
Salt Lake City
Denver
Clear Lake A
NEVADA
Sacramento
UTAH
COLORADO
San Francisc
A\Long Valley Caldera
CALIFORNIA
Coso A
Las Vegas
Santa Fe
A San Francisco Field
Volcano active during
past 2,000 years
Other potentially active
Albuquerque
A Bandera Field
Los Angeles
San Diego
ARIZONA
vo canic aieas
Phoenix
SIEW MEXICO
0 100 200 kilometers
i—1—1
Tucson
100 miles
Topinka, USGS/CVO, 1999, Modified from: Brzntiey, 1994, Volcanoes of the United States: USGS General interest Publication
Source: USGS (1999). Map courtesy of Lyn Topinka (1999, USGS/CVO), Modified from Steve Brantley (USGS 1994), Volcanos of
the United States, USGS General Interest Publication.
Figure 2-7 Geographic location of volcanoes and other potentially active
volcanic areas within the continental U.S.
The state of Hawaii, located over a "hot spot" in the north-central portion of the Pacific
tectonic plate, is a series of volcanic islands with one of the world's most active
volcanoes, Kilauea, located on the Big Island of Hawaii. Kilauea typically emits SO2 at a
steady rate. In mid-March of 2008, the volcano exhibited a small, explosive eruption,
followed by a two-to fourfold increase in SO2 emissions. The Ozone Monitoring
Instrument (OMI) aboard the NASA Aura satellite detected this increase in SO2
emissions. Figure 2-8 shows the average concentration of SO2 in the evolving plume for
2-14
-------
the March 20-27, 2008 period. Persistent easterly trade winds moved the plume
westward, away from populated areas.
Aura/OMI - Average column for 20080320-20080327
-164 -162 -160 -158 -156
-164 -162 -160 -158 -156
S02 column [DU]
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
DU = Dobson units; OMI = Ozone Monitoring Instrument; S02 = sulfur dioxide.
Note: A DU is approximately equivalent to a total column concentration of 1 ppbv of S02. Horizontal axis is longitude with respect to
Greenwich, U.K. Vertical axis is latitude with respect to the equator.
Source: NASA (2008b).
Figure 2-8 National Aeronautics and Space Administration/Ozone Monitoring
Instrument image of the KTIauea sulfur dioxide plume during its
March 20-27, 2008 eruption.
In another study using SO2 column densities derived from the Global Ozone Monitoring
Experiment-2 satellite measurements for the penod 2007-2012, Beirle et al. (2013)
determined KTlauea's monthly mean SO2 emission rate and effective SO2 lifetime. For the
March through November, 2008 period, the authors reported KTlauea's SO2 emission rate
as 8,818-20,943 tons/day and the effective SO2 lifetime as 1-2 days. Several studies
have estimated the global SO2 emissions of sulfur by volcanoes to be in the range of
7.7 m 106-2.0 x 107 tpy (Chin et al.. 2000; Feichter et al... 1996; PhametaL 1996;
Langner and Rodhe. 1991).
2-15
-------
2.2.4.3
Wildfires as a Natural Source of Sulfur Dioxide
Sulfur is a component of amino acids in vegetation and is released during combustion,
mainly in the form of SO2. Wiedinmyer et al. (2006) determined fire location and timing,
fuel loadings, and emission factors by using the literature and satellite data from various
sources, including the Moderate Resolution Imaging Spectroradiometer (MODIS)
Thermal Anomalies Product, the Global Land Cover Characteristics 2000 data set, and
the MODIS Vegetation Continuous Fields Product. The study authors estimated SO2
emissions from fires for the U.S. as 176,370 tons in the year 2004. Canadian fires emitted
121,254 tons, and Mexican fires emitted 55,116 tons of SC^forthe same period.
However, wildfire emissions do vary from year to year. Emissions estimates for SO2
derived from global modeling studies of wildfire range between 5.1 * 106 and
6.3 x 106 tpy SO2 (Chin et al.. 2000; Feichter et al.. 1996; Pham et al.. 1996; Langner and
Rodhc. 1991). For comparison, the 2014 NEI also includes an estimate for U.S.
agricultural and prescribed burning emissions at 75,643 tpy, which is comparable to the
estimated SO2 emissions from wildfires at 71,448 tpy (U.S. EPA. 2016bV
Projected increases in wildfire frequency and intensity under warming climate conditions
imply increasing wildfire-related SO2 emissions. However, these estimates are highly
uncertain due to the lack of data on the sensitivity of emissions composition with respect
to the effects of climate change on landscape species composition and burning
conditions.
2.2.5 Reduced Sulfur Compounds as Indirect Sources of Sulfur Dioxide
Sulfides, including H2S, carbonyl sulfide (OCS), carbon disulfide (CS2),
methylmercaptan (CH3SH), dimethyl sulfide (DMS), and dimethyl disulfide (DMDS), are
emitted from energy production, industrial activities, agriculture, and various ecosystems,
especially coastal wetland systems, inland soils, and oceans. In addition to SO2,
volcanoes release sulfides, specifically H2S, OCS, and CS2. As described in Section 2.3.
all of these gases, except OCS, have short atmospheric lifetimes, given their high rates of
reaction with hydroxyl and nitrate (NO3) radicals, with SO2 as a reaction product.
Table 2-2 provides a list of the natural and anthropogenic sources of the five main
organosulfides. Dimethyl sulfide is particularly important for the large role it plays as a
source of atmospheric sulfur.
2-16
-------
Table 2-2 Largest global sulfide emissions sources, ranked according to total
sulfur emissions (103 Tpy).
Anthropogenic Sources
OCS
cs2
CHsSH
DMS
DMDS
Pulp and paper industry
107
86.5
1,852
1,612
301
Manure
NR
NR
364
728
728
Rayon/cellulosics manufacture
NR
435
56.7
39.2
NR
Pigment industry
81.6
226
NR
NR
NR
Biomass burning
47.4
1.98
NR
6.61
131
Oxidation
179
NR
NR
NR
NR
Wastewater
0.024
1.14
56.2
3.75
23.6
Biofuel combustion
80.6
3.20
NR
NR
NR
Coal combustion
66.1
0.364
NR
NR
NR
Paddy fields
0.419
29.7
0.838
27.6
0.628
Aluminum industry
33.1
4.41
NR
NR
NR
Food processing
0.694
NR
NR
4.38
31.9
Shipping
33.1
1.65
NR
NR
NR
Tire wear
12.9
17.1
NR
NR
NR
Coke production
9.93
15.4
NR
NR
N
Gas industry
0.772
NR
5.29
0.926
0.110
Vehicles
6.61
0.331
NR
NR
NR
Landfill and waste
0.087
0.209
0.375
0.287
0.009
Tire combustion
0.071
NR
0.047
NR
NR
Brickmaking
NR
0.033
NR
NR
NR
Total anthropogenic sulfur emissions, by
compound
659
823
2,336
2,422
1,216
Natural Sources
OCS
cs2
CHsSH
DMS
DMDS
Saline and ocean water
978
268
5,223
31,071
235
Vegetation and soils
NR
77
1,913
3,825
957
Oxidation
174
NR
NR
NR
NR
Volcanoes
12
19
NR
NR
NR
Total natural sulfur emissions, by compound
1,164
364
7,135
34,896
1,190
CH3SH = methylmercaptan; CS2 = carbon disulfide; DMDS = dimethyl disulfide; DMS = dimethylsulfide; NR = not reported;
OCS = carbonyl sulfide.
Adapted from (Lee and Brimblecombe. 2016).
2-17
-------
Dimethyl sulfide (DMS) is the most abundant reduced sulfur gas. It has appreciable
anthropogenic sources (pulp and paper production, agricultural operations), but these
sources emit much less than the quantity emitted by natural biological activity. Natural
emissions of dimethyl sulfide are due to the breakdown of dimethyl sulfoniopropionate, a
metabolite of the amino acid methionine, which is produced by marine organisms,
particularly in areas of oceanic upwelling and coastal zones, and by anaerobic bacteria in
marshes and estuaries. The oxidation of dimethyl sulfide contributes to low-level
background SO2 concentrations in coastal environments. Lee and Brimblecombe (2016)
provide a literature-derived global estimate of DMS emissions from seawater and
marshland of 3.1 x 07 tpy S. Earlier estimates for seawater DMS emissions range widely
from 6.1 x 106 to 2.4 x 107 tpy (Liu et al.. 2005; Chin et al.. 2000; Feichter et al.. 1996;
Pham et al.. 1996; Langner and Rodhe. 1991). A warming climate may have a complex
feedback effect on DMS emissions, influencing both ocean surface temperatures and
currents controlling nutrient dispersion and affecting the population and location of DMS
producing phytoplankton (Klostcr et al.. 2007).
2.3 Atmospheric Chemistry and Fate of Sulfur Dioxide and Other
Sulfur Oxides
The important gas-phase sulfur oxides in the troposphere are SO2 and H2SO4 (U.S. EPA.
2008d). SO3 is known to be present in the emissions of coal-fired power plants, factories,
and refineries, but it reacts with water vapor in the stacks or immediately after release
into the atmosphere within seconds to form H2SO4. This short atmospheric residence time
makes SO3 difficult to detect in the ambient atmosphere. Gas-phase H2SO4, the product
of both SO2 and SO3 oxidation, quickly condenses onto existing atmospheric particles or
participates in new particle formation (Finlavson-Pitts and Pitts. 2000). Of these species,
only SO2 is present at concentrations in the gas phase that are relevant for chemistry in
the atmospheric boundary layer and troposphere, and for human exposures. Other sulfur
oxides, including both S(IV) and S(VI) compounds, appear in the atmosphere due to
direct emissions and as the products of the oxidation of more reduced forms of sulfur.
Gas-phase precursors to SO2 include the sulfides (Section 2.2.5) and partially oxidized
sulfur-containing organic compounds. Early research on industrial air pollution reported
observations of partially oxidized sulfur compounds in industrial emissions plumes and
the ambient atmosphere as both gases and particulate matter. These compounds included
bis-hydroxymethyl sulfone (Eatough and Hansen. 1984). dimethyl- and
monomethyl-sulfate (Eatough et al.. 1986; Eatough et al.. 1981). sulfonic acids (Panter
and Penzhorn. 1980). and particle-bound iron sulfite complexes (Eatough et al.. 1978).
No other more recent studies are available that quantify these intermediate sulfur oxides.
2-18
-------
However, energy generation, industrial coal combustion, and refinement emissions
control technologies reduce the contributions of anthropogenic sources of these
compounds. Some more recent detailed studies of the atmospheric chemistry of these
compounds also indicate that these compounds may have short residence times, further
reducing their importance for human exposure.
The following subsections provide an overview of the primary atmospheric chemistry
and removal processes for SChthat are relevant to atmospheric concentrations at urban
scales. Section 2.3.1 describes the photochemical reactions that remove SO2 from the
atmosphere by converting it into compounds that condense into the particle or cloud
water phase. Section 2.3.2 describes the aqueous-phase oxidation of SO2, the major
oxidation mechanism in the atmosphere, as well as dry and wet deposition of SO2.
Section 2.3.3 summarizes the available research on the atmospheric chemistry of sulfur
oxides other than SO2 and H2SO4.
2.3.1 Photochemical Removal of Atmospheric Sulfur Dioxide
The global atmospheric lifetime (r) of SO2 with respect to reactions with the OH radical
in the troposphere is 7.2 days. The rate constant for the reaction between SO2 and NO3
radical is too small to be important in lowering SO2 concentrations at urban or regional
scales. The same is true for the reaction between SO2 and the hydroperoxyl (HO2) radical
(Sander et al.. 2011).
In the stepwise oxidation of SO2 by OH, SO2 is oxidized to form SO3, taking the sulfur
atom from the S(IV) to S(VI) oxidation state, producing the bisulfite radical (HSO3):
SO2 + OH + M -> HSOs + M
Equation 2-1
where M is an unreactive gas molecule that absorbs excess destabilizing energy from the
SO2-OH transition state. This reaction is followed by
HSOs + O2 ^ SOs + HO2
Equation 2-2
An alternative route involves a stabilized Criegee intermediate (sCI):
SO2 + sCI -> SO3 + products
Equation 2-3
2-19
-------
A Criegee intermediate is a carbonyl oxide zwitterion, i.e. a molecule with separate
positive and negative charge centers, derived from the oxidation of an alkene gas
molecule by ozone. The unspecified "products" of this reaction are other organic radicals
that result from the degradation of the Criegee intermediate (Bcrndt et al.. 2012; Mauldin
et al.. 2012; Welz et al.. 2012). Rate coefficients for the reaction of sCIs with SO2 have
been reported as 4 * 10 15 cm3/sec (Johnson et al.. 2001). approximately
3.5 x io~n cm3/sec (Liu et al.. 2014b). and 3.9 x 10~n cm3/sec (Welz et al.. 2012). Recent
studies report rate coefficients greater than 3 x 10" cm3/sec (Friedman et al.. 2016; Lee.
2015; Berndt et al.. 2012). These reaction rate coefficients far exceed those of the
reactions between these intermediates and H2O. However, hydrolysis of SO2 could be
limited if sCIs that are potential SO2 oxidants are hydrolyzed via competing reactions
(Kim et al.. 2015). The efficiency of Criegee radical hydrolysis is sensitive to the
molecular structure of the alkene. Bimolecular hydrolysis rate constants vary by a factor
of 1,000 between syn- and anti-substituted low molecular weight alkenes (Lin and
Takahashi. 2016).
Criegee radicals are produced during both night and day. The relative importance of the
OH and sCI pathways depends in large measure on the local concentration of alkenes,
such as the low-molecular-weight ones emitted by motor vehicles and industrial
processes, as well terpenoids emitted by trees. The importance of this mechanism as a
sink for SO2 is supported by observations that areas adjacent to SO2 sources, with high
biogenic or industrial volatile organic compound (VOC) concentrations, have elevated
organic PM concentrations (Friedman et al.. 2016). However, limited information on the
identity and concentrations of alkenes at urban scales prevents estimates of the impact of
this reaction pathway on urban SO2 concentrations.
The SO3 that is generated by either oxidation mechanism (i.e., reaction with OH or via
the Criegee reaction) is highly reactive. Water vapor is sufficiently abundant in the
troposphere to ensure that SO3 is quickly converted to gas-phase sulfuric acid, as shown
in the equation below (Locrting and Liedl. 2000).
SOs + H2O + M -> H2SO4 + M
Equation 2-4
Because H2SO4 is extremely water soluble, gaseous H2SO4 will rapidly dissolve into the
aqueous phase of aerosol particles and cloud droplets. In a study of SO2 plume transport
in and out of foggy conditions, Eatough et al. (1984) observed that roughly 30% of the
SO2 converts to H2SO4 particulate each hour when inside a fog bank and roughly 3.1%
per hour outside a fog bank. Khoder (2002) observed that conversion from SO2 to H2SO4
increases with increasing relative humidity and O3 levels, based on a sampling campaign
in an urban area of Egypt. Pearson correlation of S02-to-H2S04 conversion ratio with
2-20
-------
relative humidity was 0.81 in the winter and 0.89 in the summer. Hung and Hoffmann
(2015) recently conducted spray chamber experiments of SO2 to H2SO4 conversion. They
observed that SO2 deposited to the surfaces of water microdroplets and then underwent
rapid oxidation, first to HSO, and bisulfate (HSO4 ). and then to SO42 . Acidic
conditions promoted more rapid oxidation of SO2.
2.3.2 Heterogeneous Oxidation of Sulfur Dioxide
Major sulfur-containing species in clouds include the HSO, and SO,2 (sulfite) ions that
form when SO2 dissolves in cloud droplets and subsequently undergoes acid dissociation.
Both species exist in the S(IV) oxidation state and readily oxidize in the presence of
aqueous-phase oxidizing agents to form the S(VI) anions, HSO4 and SO42 . The major
species capable of oxidizing S(IV) to S(VI) in cloud water are O3, peroxides [either
hydrogen peroxide (H2O2) or organic peroxides], OH radicals, and transition metal ions
such as Fe(II) and Cu(II) that catalyze the oxidation of S(IV) to S(VI) by O2.
The basic mechanism of the aqueous-phase oxidation of SO2 can be found in numerous
texts on atmospheric chemistry [e.g., (Seinfeld and Pandis. 2006; Jacobson. 2002;
Finlavson-Pitts and Pitts. 2000; Jacob. 1999)1. Similar initial steps occur in the fluids
lining the airways (Section 4.2.1). The steps involved are summarized below (Jacobson.
2002).
Dissolution of SO2 occurs first,
S02(g) <=> S02(aq)
Equation 2-5
followed by the formation and dissociation of sulfurous acid (H2SO3).
S02(aq) + H20(1) <=> H2SO3 <=> H+ + HSOs" <=> 2H+ + SOs2"
Equation 2-6
In the pH range commonly found in rainwater (2 to 6), H2O2 will oxidize HSO , to SO42 .
HSOs" + H2O2 + H+ ^ S042" + H2O + 2H+
Equation 2-7
The rates of aqueous-phase oxidation of S(IV) to S(VI) as a function of pH are shown in
Figure 2-9. For pH values up to about 5.3, H2O2 is the predominant oxidant; above pH
5.3, O3, followed by Fe(III), becomes predominant.
2-21
-------
HO
1-12
>
W
"O
I
1-14
r1B
0
1
2
3
4
5
6
PH
Fe = iron; H202 = hydrogen peroxide; Mn2+ = manganese ion; N02 = nitrogen dioxide; 03 = ozone; S = sulfur.
Note: The rate of conversion of aqueous (droplet)-phase S(IV) to S(VI) is shown as a function of pH. Conditions assumed are:
[S02(g)] = 5 ppb; [N02(g)] = 1 ppb; [H202(g)] = 1 ppb; [03(g)] = 50 ppb; [Fe(lll)(aq)] = 0.3 jiM; [Mn(ll)(aq)] = 0.3 jiM.
Source: Seinfeld and Pandis (20061. Reprinted with permission of Wiley.
Figure 2-9 The effect of pH on the rates of aqueous-phase sulfur (IV)
oxidation by various oxidants.
Ambient ammonia (NH3) vapor readily dissolves in acidic cloud droplets to form
ammonium (NH4+). Because NH4+ is very effective in scavenging H+, which shifts the
SO2 oxidation equilibrium, amplifying the rate of oxidation of S(IV) to S(VI) and the
rate of dissolution of SO2 in particles and cloud droplets. Therefore, in environments
where NH3 is abundant, SO2 is subject to fast removal by cloud and fog droplets and
ultimately forms ammonium sulfate [(NH^SO.*].
Higher pH levels are expected to be found mainly in marine aerosols. In marine aerosols,
the chlorine radical-catalyzed oxidation of S(IV) may be more important (Hoppcl and
Caffrev. 2005; Zhang and Millero. 1991).
In the same way that it is removed from the gas phase by dissolving into cloud droplets,
SO2 can be removed by dry deposition onto wet surfaces (Shadw ick and Sickles. 2004;
2-22
-------
Clarke et al.. 1997). For example, in the eastern U.S., more than 85% of sulfur (as SO2) is
removed by dry deposition (Sickles and Shadwick. 2007). However, aqueous SO42 may
be removed through occult deposition of large fog or cloud droplets (Lillis et al.. 1999;
Pandis and Seinfeld. 1989; Pollard et al.. 1983). Scavenging by rain (wet deposition)
serves as another removal route. Modeling studies have shown that slightly more than
half of SO2 is lost by gas- and aqueous-phase oxidation, with the remainder of SO2 loss
accounted for by wet and dry deposition (Long et al.. 2013; Liu et al.. 2012a).
Sulfur dioxide is known to adhere to and then react on dust particles. Very recent
investigations have shown that, for some mineral compositions, SO2 uptake on dust
particles is sensitive to relative humidity, the mineral composition of the particle, and the
availability of H2O2, the relevant oxidant (Huang et al.. 2015b). Once SO2 is oxidized to
H2SO4 on the particle surface, glyoxal, one of the most prevalent organic compounds in
the atmosphere, will adhere to the surface and react to form oligomers and organosulfate
compounds. This process is enhanced under high humidity conditions (Shen et al.. 2016).
2.3.3 Secondary Gas-phase and Particle-phase Sulfur Oxides
Little information is available in the peer-reviewed literature concerning the atmospheric
chemistry of intermediate gas- and particle-phase sulfur oxidation products. One study
evaluated the atmospheric residence time of dimethyl sulfate against oxidation by the
gas-phase oxidants (O3, OH, and CI), along with NH3 and H2O (Japaretal.. 1990). Under
typical atmospheric conditions, dimethyl sulfate removal will occur in less than 2 days by
reaction with water vapor (other removal rates ranged from 23 days to 33 years). Most of
these species can be expected to partition into the aqueous particle phase due to their high
polarity (Barnes et al.. 2006). Rapid oxidation to H2SO4 would be expected to occur in
the aqueous phase (Japaretal.. 1990).
Given the technological improvements in burning coal and in controlling emissions in
recent decades, one may infer that these species are unlikely to exist in concentrations
significant for human exposure. However, the species reported by Eatough et al. (1986)
have since been identified as intermediates in the oxidation of dimethyl sulfide (Barnes et
al.. 2006). a ubiquitous naturally emitted sulfide species associated with coastal waters
and wetlands, vegetation, and soils (Table 2-2).
Particle-phase inorganic and organic sulfur compounds have been identified in early
studies (Eatough and Hansen. 1983. 1980; Lee et al.. 1980; Eatough et al.. 1979; Eatough
et al.. 1978; Smith et al.. 1976). These studies identified inorganic S032" complexed with
Fe(III) in the particles emitted by a smelter near Salt Lake City, UT. In a detailed
spectroscopic study of the transient complexes that form between SO2, a source of S(IV)
2-23
-------
in particles, and Fe(III) in the aqueous phase, Kraft and van Eldik (1989a) and Kraft and
van Eldik (1989b) reported that the oxidation of S(IV) by Fe(III) to form SC>42~ occurs on
the order of seconds to minutes and is further accelerated by low pH. Sulfuric acid is well
known to absorb water at even low ambient relative humidity (Seinfeld and Pandis.
2006). The highly acidic aqueous conditions that arise once smelter plume particles
equilibrate with the ambient atmosphere ensure that S(IV)-Fe(III) complexes have a
small probability of persisting and becoming a matter of concern for human exposure.
Substantial effort in recent years has been applied to understanding the mechanism for
the formation of organic sulfur compounds, and the results of this effort are described in
Section 2.3.1.
2.4 Measurement Methods
This section discusses the Federal Reference Method (FRM) and Federal Equivalent
Method (FEM) used for NAAQS compliance as well as the state, local, and tribal
monitoring networks across the U.S. used for NAAQS compliance monitoring. Detailed
information about monitoring methods, including accuracy, precision, limits of detection,
and other operational parameters was published in the 1982 Air Quality Criteria for
Particulate Matter and Sulfur Oxides Volume II (U.S. EPA. 1982a) and then updated in
Appendix B.6 of the 2008 ISA for Sulfur Oxides—Health Criteria (U.S. EPA. 2008d).
The List of Designated Reference and Equivalent Methods (U.S. EPA. 2016e) includes
all monitors approved as FRMs or FEMs and provides monitor specifications. A brief
summary of that information, along with a discussion of more recent studies evaluating
FRMs and FEMs for monitoring SO2 concentration (Section 2.4.1) or alternative SO2
monitoring methods (Section 2.4.2). is provided. Section 2.4.3 describes the sampling
network.
2.4.1 Federal Reference and Equivalent Methods
Currently, there are two FRMs for the measurement of SO2—the manual pararosaniline
wet-chemistry method and the automated pulsed ultraviolet fluorescence (UVF) method.
The manual method was approved as an FRM in the 1970s and was quickly supplanted
by the flame photometric detection (FPD) method, an FEM, because the manual method
was too complex and had a slow response even in automated form. The UVF method was
designated as an FEM in the late 1970s and ultimately supplanted the FPD method. The
UVF method became the dominant method in routine monitoring networks because it is
inherently linear and relatively safe, whereas the FPD method requires highly flammable
hydrogen gas. The UVF method has been the most commonly used method by state and
2-24
-------
local monitoring agencies since the 1980s. It was promoted to FRM status during the
promulgation of the new 1-hour SO2 primary NAAQS established in 2010 (75 FR 35520)
in light of its reliability and well-documented performance. The UVF method supports
the need for a continuous monitoring method, providing highly time-resolved data, such
as 5-minute data as well as routine 1-hour SO2 measurements. 1-hour SO2 measurements.
The existing pararosaniline manual method was retained as an FRM, and although
cumbersome, the method can provide hourly measurements to support the 1-hour
NAAQS.
In the UVF method, SO2 molecules absorb ultraviolet (UV) light at one wavelength,
elevating the molecule to a higher energy electronic state. Once electronically excited, the
molecule loses a portion of its energy by colliding with another gas molecule. The
molecule relaxes back to its electronic ground state by emitting a photon of light at a
longer wavelength (i.e. lower energy) than the light used to excite the molecule. The
intensity of the emitted light is, therefore, proportional to the number of SO2 molecules in
the sample gas. In commercial analyzers, light from a high-intensity UV lamp passes
through an optical bandwidth filter that allows only photons with wavelengths around the
SO2 absorption peak (near 214 nm) to enter the optical chamber. The light passing
through the bandwidth filter is collimated using a UV lens and passes through the optical
chamber to a reference photomultiplier detector. The detector is set perpendicular to the
illumination path to maximize the collection of SO2 fluorescence. An optical bandwidth
filter designed to block higher energy frequencies is added to further protect the detector
from stray UV lamp light. Quartz lenses are positioned between the filter and the detector
to focus the SO2 fluorescence photons onto the active area of the detector, optimizing the
fluorescence signal.
Studies have compared UVF to sampled SO2 from impregnated filters for quality
assurance. Comparison of 24-hour avg concentration measurements obtained with the
UVF method and with impregnated filters showed annual average differences within
±0.07 ppb, based on data obtained between 1993 and 2001 from four Finnish cities
(Lcppancn et al.. 2005). Ferek et al. (1997) evaluated the Teco model UVF (developed at
the University of Washington) against carbonate-impregnated filters for measurement of
SO2 concentration in laboratory studies. The Teco UVF measured SO2 concentrations
down to 16 ppt and on average, produced a positive difference of 7% compared with the
filter. The Teco UVF analyzed data at a frequency as high as 1 Hz, but noise was
curtailed by averaging up to 10 minutes. The Ferek et al. (1997) study highlighted the
Teco UVF but also included other SO2 measurement techniques in the SO2 monitor
comparison, including gas spectrometry/mass spectrometry, high performance liquid
chromatography, and a mist chamber, which produced a maximum of 30% differences
for filter-measured SO2 concentrations of 3-4 ppb averaged over 90 minutes.
2-25
-------
2.4.1.1 Minimum Performance Specifications
During the 2010 SO2 NAAQS review, minimum performance specifications [contained
in 40 Code of Federal Regulations (CFR) Part 53] were updated and became more
stringent for any new FRM and FEM automated method. The new specifications are
provided in Table 2-3. The previous specifications were based on the older, manual,
wet-chemistry FRM and were updated to reflect current technology and improved
performance in SO2 instrumentation. The lower detection limit (LDL) for a routine,
automated SO2 analyzer is required to be 2 ppb. As part of the National Core (NCore)
monitoring network, new trace-level SO2 instruments have been developed and added to
State and Local Air Monitoring Sites (SLAMS). These new trace-level (i.e., low LDL)
instruments have LDLs of 0.2 ppb or lower. Note that FRMs and FEMs may have more
stringent performance characteristics than the minimum performance specifications
presented in Table 2-3.
Depending on design, instrument settings, and operator preferences, SO2 instruments can
have different time constants or averaging times [i.e., the combined rise and fall time of
the signal detection system (75 FR 35520)] (Terek etal.. 1997). Instruments operated
with longer time constants will respond more slowly to short-lived, high concentration
spikes in SO2, reporting a broader peak with a lower maximum concentration compared
with instruments operated with shorter time constants. The delayed decline to baseline
SO2 concentrations can influence concentration measurements at adjacent 5-minute
intervals. Short concentration spikes may also not be fully detected or may be
time-shifted due to the division of each hour into twelve discrete 5-minute intervals.
2-26
-------
Table 2-3 Minimum performance specifications for sulfur dioxide established
in 40 Code of Federal Regulations Part 53, Subpart B.
Performance Parameter
Specification
Range1
0-0.5 ppm (500 ppb)
Noise
0.001 ppm (1 ppb)
Lower detection limit (two times the noise)
0.002 ppm (2 ppb)
Interference equivalent
• Each interferent
• Total, all interferents
±0.005 ppm (5 ppb)
Zero drift (12 and 24 h)
±0.004 ppm (4 ppb)
Span drift (24 h)
• 20% of upper range limit
• 80% of upper range limit
±3.0%
Lag time
2 min
Rise time
2 min
Fall time
2 min
Precision
• 20% of upper range limit
• 80% of upper range limit
2.0%
2.0%
1The CFR also provides for testing at lower ranges with special performance limit requirements.
2.4.1.2 Positive and Negative Interferences
The UVF method has a number of positive and negative interferences. The most frequent
source of positive interference is other gases that fluoresce at the same wavelength as
SO2. The most common gases include volatile organic compounds (e.g., xylenes,
benzene, toluene) and polycyclic aromatic hydrocarbons (PAHs; e.g., naphthalene). To
reduce this source of positive interference, high-sensitivity SO2 analyzers are equipped
with scrubbers or "kickers" to remove these compounds from the air stream prior to
entering the optical chamber. Luke (1997) evaluated a modified pulsed fluorescence SO2
detector and found positive interference from nitric oxide (NO), CS2, and several highly
fluorescent aromatic hydrocarbons such as benzene, toluene, o-xylene, m-xylcnc.
/•-xylene, m-cthyltolucne. ethylbenzene, and 1,2,4-trimethylbenzene. The positive
2-27
-------
artifacts could be virtually eliminated by using a hydrocarbon "kicker" membrane. At a
flow rate of 300 standard cm3/minute and a pressure drop of 645 torr across the
membrane, the interference from ppm levels of many aromatic hydrocarbons can be
eliminated.
Another source of positive interference is NO, which fluoresces in a region close to that
of SO2. However, in high-sensitivity SO2 analyzers, the bandpass filter in front of the
detector is specifically designed to prevent detection of NO fluorescence at the detector.
Care must be exercised when using multicomponent calibration gases containing both
NO and SO2, so that the NO rejection ratio of the SO2 analyzer is sufficient to prevent
NO interference.
The most common source of positive bias in high-sensitivity SO2 analyzers is stray light
in the optical chamber. Because SO2 can be excited by a broad range of UV wavelengths,
any stray light entering the optical chamber with an appropriate wavelength can excite
SO2 in the air stream and increase the fluorescence signal. Additionally, stray light
entering the optical chamber with a similar wavelength of SO2 fluorescence may impinge
on the detector and increase the fluorescence signal. Stray light is also minimized with
changes in instrument design such as use of light filters, dark surfaces, and opaque
tubing.
H2O is a common source of negative interference in high-sensitivity SO2 monitors. When
excited SO2 molecules collide with water vapor as well as other common molecules in air
(e.g., nitrogen and oxygen), nonradiative deactivation (quenching) can occur. During
collisional quenching, the excited SO2 molecule transfers energy, kinetically allowing the
SO2 molecule to return to a lower energy state without emitting a photon. Collisional
quenching decreases the SO2 fluorescence and results in an underestimation of SO2
concentration in the air sample. Of particular concern is the variable water vapor content
of air. Luke (1997) reported that the response of the detector could be reduced by an
amount of approximately 7 to 15% at water vapor mixing ratios of 1 to 1.5 mole percent
[relative humidity (RH) = 35 to 50% at 20 to 25°C and 1 atmosphere for a modified
pulsed fluorescence detector (Thermo Environmental Instruments, Model 43s)]. Systems
may minimize this interference by maintaining instrument and sampling lines at
markedly higher operational temperatures than the expected dew point and to within a
few degrees of the controlled optical bench temperature. Condensation of water vapor in
sampling lines must be avoided, as water on the inlet surfaces can absorb SO2 from the
sample air. Condensation is normally prevented by heating sampling lines to a
temperature above the expected dew point and to within a few degrees of the controlled
optical bench temperature. Some monitors are equipped with a dryer system to remove
moisture from the sample gas before it reaches the particulate filter.
2-28
-------
2.4.2
Alternative Sulfur Dioxide Measurements
A number of optical methods for measuring SO2 are available. They include laser
induced fluorescence (LIF), cavity ring-down spectroscopy (CRDS), differential optical
absorption spectrometry (DOAS), and UV light absorption. There are also methods based
on mass spectroscopy or mass spectrometry [e.g., chemical ionization mass spectroscopy
(CIMS) and atmospheric pressure ionization mass spectrometry (APIMS)]. These
methods are often too expensive and complex for routine monitoring applications and are
more suitable for source monitoring and research. However, approaches to reduce
interferences and increase SO2 selectivity could be extended to FRM and FEM
instrumentation. The LIF, CRDS, and DOAS methods will be discussed below as they
have the potential to provide trace-level SO2 measurements or have shown good
agreement with UVF instrumentation.
LIF is a technique that can provide high sensitivity for ambient SO2 measurements and
reduces interferences with species that fluoresce at the same wavelength as SO2. Both
tunable and nontunable laser sources have been evaluated. Matsumi et al. (2005)
evaluated a LIF method using a tunable laser at an SO2 absorption peak at 220.6 nm and
trough at 220.2 nm. The difference between the signals at the two wavelengths is used to
estimate the SO2 concentration. This technique has a sensitivity of 5 ppt in 60 sec.
Simeonsson et al. (2012) evaluated a direct LIF technique using a nontunable laser source
at an absorption wavelength of 223 nm, which coincides with the SO2 absorption peak.
This technique has a high sensitivity with LDL of 0.5 ppb. Both the tunable and
nontunable instruments have low LDL (<0.5 ppb); therefore, they can provide trace-level
SO2 measurements.
CRDS is an optical absorption method based on measurement of the rate of light
absorption through a sample. CRDS has successfully been used to measure ambient NO2
and NO with high sensitivity. Medina etal. (2011) compared a CRDS-tunable laser
method to the routinely used pulsed UVF method for measuring SO2. At an absorption
wavelength of 308 nm, the CRDS had an LDL of 3.5 ppb, which was higher than those
for routine and trace-level UVF SO2 monitors (e.g., Thermo Scientific 43i and Thermo
Scientific 43i-TLE). However, the response time was faster compared to the UVF
methods (a few seconds vs. 80 seconds). To reduce interferences, a ferrous sulfate
scrubber was used to remove NO2 and O3, and a denuder was used to zero SO2 levels.
Improvements could be made to increase the sensitivity to about 1 ppb by changing the
placement of the mirrors to optimize laser light reaching the cavity or using a better
detection system. Additionally, improving the mirror reflectivity could improve the
sensitivity to about 0.1 ppb, similar to the detection levels of trace-level SO2 monitors.
2-29
-------
DOAS is an optical remote sensing method based on the absorption of light in the
UV-visible wavelength region to measure atmospheric pollutants. Kim and Kim (2001)
compared SO2 concentrations measured using a DOAS system with daily mean SO2
concentrations measured by an in situ monitor in Seoul, Korea during a 13-month period.
In this study, the DOAS typically reported SO2 concentrations around 10-40% above the
in situ technique, but SO2 concentrations measured by the DOAS were sometimes
100-200% below those measured with the in situ monitor. Across all measurements, the
daily mean SO2 concentration was 36% higher from the DOAS compared with the in situ
monitor. Discrepancies between the two methods were attributed to ability to respond to
meteorological factors. The DOAS was reported to have an LDL of 0.07 ppb, compared
with 1 ppb reported for the in situ method. A newer technique called multiaxis
differential optical absorption spectroscopy (MAX-DOAS) has been developed that
offers increased sensitivity in measuring SO2 (Honningcr et al.. 2004). MAX-DOAS is
based on the measurement of scattered sunlight at multiple viewing directions and can
obtain both surface concentrations and vertical column density of SO2. Wang et al.
(2014b) compared MAX-DOAS SO2 column measurements in the 305 to 317.5 nm
absorption wavelength to surface SO2 measurements from a modified UVF SO2 monitor
(Thermo Environmental Instruments Model 43C) and found good agreement (r = 0.81,
slope = 0.90).
Remote sensing by satellites (e.g., OMI, infrared atmospheric sounding interferometer,
etc.) is an emerging technique for measuring SO2, as well as other pollutants. This
technique can be used for a variety of applications, including air quality management
(e.g., augmenting ground-based monitors, assessing emissions inventories), studying
pollutant transport, assessing emissions reductions, and evaluating air quality models.
Remote sensing methods employ a retrieval system using a combination of solar
backscatter or thermal infrared emission spectra and mathematical algorithms to estimate
pollutant concentrations. Remote sensing from space is particularly challenging for SO2
measurements for two reasons: (1) air scattering causes SO2 to have a low optical
thickness (three orders of magnitude lower than O3), so that only large SO2 sources can
be observed (Bogumil et al.. 2003) and (2) emissions reductions programs have led to
lower SO2 emissions from stationary sources, making it more difficult to see
anthropogenic SO2 emissions (Streets et al.. 2014). The majority of remote sensing
studies have focused on large natural sources (e.g., volcanoes), large anthropogenic
sources (e.g., coal-burning power plants, smelters), fuel extraction from oil sands, and
newly constructed coal-burning facilities with high, uncontrolled SO2 emissions
(Bovnard et al.. 2014; McCormick et al.. 2014; Streets et al.. 2014; Clarisse et al.. 2012;
McLinden et al.. 2012; Fioletov et al.. 2011; Nowlan etal.. 2011; Bobrowski et al.. 2010;
Li et al.. 2010; Khokhar et al.. 2008; Cam et al.. 2007).
2-30
-------
2.4.3
Ambient Sampling Network Design
Compliance with NAAQS is primarily carried out through the SLAMS network, although
modeling may also be used to characterize air quality for implementation purposes
(75 FR 35520). There were 427 SLAMS sites that reported 1-hour SO2 concentrations in
2015 to the Air Quality System (AQS), U.S. EPA's repository for detailed air pollution
data that is subject to quality control and assurance procedures. In addition to their use in
compliance evaluations, some of these sites function as central monitoring sites for use in
epidemiologic studies. The SLAMS network also reports either the maximum 5-minute
concentration in the hour (one of twelve 5-minute periods within an hour) or all twelve
5-minute average SO2 concentrations within the hour. Siting requirements for monitors in
the SLAMS network can be found in 40 CFR Part 58, Appendix E.
The SLAMS network includes the NCore monitoring network, which began January 1,
2011 and consisted of 74 mostly urban sites in 2015. NCore is a multipollutant
measurement network and includes SO2 measurements as well as measurements for other
gaseous pollutants (O3, CO, NOx, oxides of nitrogen), PM2 5, PM10-2.5, and meteorology.
NCore is focused on characterizing trends in pollutants, understanding pollutant transport
in urban and rural areas, and evaluating data with respect to the NAAQS. Figure 2-10
shows the locations of these monitoring networks across the U.S. The Clean Air Status
and Trends Network (CASTNet) also measures ambient SO2. However, these data are not
used for NAAQS compliance purposes and are obtained predominantly in National Parks
or other ecologically sensitive sites. Because CASTNet monitors are not deployed in
populated areas, they are not useful in evaluating the health effects of SO2. This network
provides weekly averages of total sulfur (dry SO2, dry SO42 , and wet SO42 ) in about
90 sites located in or near rural locations to assess long-term trends in acidic deposition
associated with emissions reduction programs. CASTNet data are presented in the
Integrated Science Assessment for Oxides of Nitrogen and Sulfur—Ecological Criteria
(U.S. EPA. 2008b).
2-31
-------
SC>2 Monitor Networks
and Reporting Times
~ NCORE, 5 min
O NCORE, 1hr
it SLAMS, 5 min
# SLAMS, 1hr
Kilometers A
0 400 800 ^
New YorkAConnecticut
Kilometers
Boston, MA
Kilometers
0 10 20
New Jersey,
Pittsburgh, PA
o
°
©
Esri, HERE. USGS.
©OpenStreetMap conlributprs
S02 Monitor
Networks and
Reporting Times
~ NCORE, S min
O NCORE, 1hr
it SLAMS, 5 min
© SLAMS, 1hr
Kilometers
NCORE = National Core; SLAMS = State arid Local Air Monitoring Sites; S02 = sulfur dioxide.
Figure 2-10 Routinely operating sulfur dioxide monitoring networks: National
Core and State and Local Air Monitoring Sites, reporting 1 hour
and 5 minute sulfur dioxide concentration data. Maps for the
Boston, MA, New York City, and Pittsburgh, PA metropolitan
areas are provided as examples of the variation in monitor
placement in the US Northeast.
2-32
-------
The minimum monitoring requirements for the SLAMS network are outlined in 40 CFR
Part 58, Appendix D. Sulfur dioxide monitors at SLAMS sites represent four main spatial
scales: (1) microscale—areas in close proximity, up to 100 m from a SO2 point or area
source, (2) middle scale—areas up to several city blocks, with linear dimensions of about
100 to 500 m, (3) neighborhood scale—areas with linear dimensions of 0.5 to 4 km, and
(4) urban scale—urban areas with linear dimensions of 4 to 50 km. Maximum hourly SO2
concentrations are established based upon measurements taken at the 3 smaller scales
[i.e. micro-, middle- and neighborhood] to account for near-source and neighborhood-
scale concentrations. Urban-scale sites are sometimes used as central monitoring sites to
characterize population exposures and trends, such as in epidemiologic studies
(Section 3.2.1). Urban-scale sites can also be used to determine background
concentrations in areas where monitors are located upwind of a local source. There are
also a number of regional-scale monitoring sites, representing length scales of tens to
hundreds of kilometers, typically in rural areas of uniform geography without large SO2
sources. These sites can be used to determine the amount of regional pollution transport
and to support secondary NAAQS.
Stationary sources are the primary emission sources of SO2. Prior to the enactment of the
revised SO2 primary NAAQS in 2010, U.S. EPA evaluated about 488 SO2 monitoring
sites in operation during 2008 and found that the network was insufficient to support the
proposed NAAQS (U.S. EPA. 20096). To partially address this deficiency, U.S. EPA
promulgated minimum monitoring requirements based on a near-source monitoring
approach. The Population Weighted Emissions Index (PWEI), which is based on
population and emissions inventory data at the core-based statistical area (CBSA) level,
was introduced to assign the appropriate number of monitoring sites in a given CBSA
(75 FR 35520). The PWEI accounts for SO2 exposure by requiring monitor placement in
urban areas where population and emissions may lead to higher potential for population
exposure to maximum hourly SO2 concentrations. The PWEI value is calculated by
multiplying the population of each CBSA by the total amount of SO2 emissions (in tons
per year) in a given CBSA, using the most recent census data (or estimates) and
combining the most recent county-level emissions data (from the National Emissions
Inventory) for each county in each CBSA, respectively. This value is then divided by
1 million, resulting in a PWEI value with units of million person-tons per year. A
minimum of three SO2 monitoring sites is required for any CBSA with a PWEI value
greater than or equal to 1,000,000. For any CBSA with a PWEI value greater than or
equal to 100,000 but less than 1,000,000, a minimum of two SO2 monitoring sites is
required. Lastly, a minimum of one SO2 monitoring site is required for any CBSA with a
PWEI value greater than or equal to 5,000 but less than 100,000. The monitors sited
within a CBSA based on the PWEI criterion should also be, at minimum, one of the
2-33
-------
following monitoring site types: population exposure, highest concentration, source
impacted, general background, or regional transport.
2.5 Environmental Concentrations
This section provides an overview of SO2 ambient and background concentrations.
Analyses are focused on characterizing recent SO2 concentration data from the U.S.
rather than the influence of atmospheric stability and meteorological conditions on
concentration distributions. Information on previous SO2 concentrations can be found in
the 2008 SOx ISA (U.S. EPA. 2008d) and earlier documents. SO2 data discussed in this
section were obtained from the AQS. Section 2.5.1 introduces different SO2 metrics used
for NAAQS compliance and epidemiologic applications. Ambient concentrations of SO2
are then discussed on various spatial and temporal scales. Spatial variability is discussed
in Section 2.5.2. which is divided into two sections discussing large-scale variability
(i.e., nationwide) and small-scale variability (i.e., urban areas). Temporal variability is
then discussed in Section 2.5.3. extending from multiyear trends to subhourly variations.
The relationships between 5-minute hourly max and 1-hour concentrations are described
in Section 2.5.4. Background SO2 concentrations from natural sources are subsequently
discussed in Section 2.5.5.
2.5.1 Sulfur Dioxide Metrics and Averaging Time
Different metrics are used to represent ambient SO2 concentrations for epidemiologic
analysis and NAAQS compliance. As discussed in Section 2.5.4. hourly and 5-minute
concentration data are routinely reported to U.S. EPA's AQS data repository by state,
local, and tribal agencies. Metrics can be derived from these hourly and 5-minute data to
represent concentration and exposure levels on different time scales. Table 2-4 provides
information on how different SO2 metrics are derived. Daily metrics include the 24-hour
avg SO2 concentration and the 1-hour daily max SO2 concentration. Hourly metrics
include the 5-minute hourly max concentration reported during a given hour and the
1-hour avg concentration. Metrics derived using maximum concentration statistics
(i.e., 1-hour daily max or 5-minute hourly max) provide insight about peak ambient
concentrations occurring over a given hour or day.
The following sections include national and urban statistics on daily and hourly metrics.
When interpreting the statistics, it is important to consider the aggregation time when
comparing the magnitude and range of ambient concentrations related to different
metrics.
2-34
-------
Table 2-4 Summary of sulfur dioxide metrics and averaging times.
Metric
Aggregation Time
Averaging Time Description
24-h avg
Daily
Daily mean of 1-h avg SO2 concentrations
1-h daily max
Daily
Maximum 1-h SO2 concentration reported during the day
1-h avg
Hourly
Hourly mean SO2 concentrations reported during the day
5-min hourly max
Hourly
Maximum 5-min SO2 concentration reported during 1-h
Avg = average; max = maximum; S02 = sulfur dioxide.
AQS SO2 data used to compute national statistics meet the data quality and completeness
criteria listed in Table 2-5. Three additional criteria were applied for the 5-minute data to
reduce the influence of outliers: (1) the 5-minute data had to correspond to an hourly data
concentration, (2) the mean of the 5-minute data could be no more than 120% of the
hourly mean, and (3) the 5-minute hourly max concentration had to fall within 1 to
12 times the 1-hour avg concentration. The AQS, by convention, accepts values that fall
within the range defined by the positive and negative of the absolute value of the
instrument's lower detection limit. This analysis included those values. Based on these
criteria, statistics were computed for data from a total of 380 sites across the U.S. for
5-minute hourly max SO2 concentrations and for data from a total of 438 sites for the
1-hour daily max, 24-hour avg, and 1-hour avg SO2 metrics. Thirteen percent of sites did
not have 5-minute data for comparison with 1-hour data.
2.5.2 Spatial Variability
This section provides a brief overview of national- and urban-scale SO2 spatial variability
and discusses how variations in ambient SO2 concentrations influence human exposure in
different geographical regions.
2-35
-------
Table 2-5 Summary of sulfur dioxide data sets originating from the Air Quality
System database.
AQS SO2 data used to compute national statistics (to meet the data quality and completeness criteria)
Years 2013-2015
Months January-December
Completeness criteria 75% of 5-min periods in an hour (where 5-min data are available)
75% of h in day
75% of days in calendar quarter
3 of 4 quarters of the yr
Number of monitoring sites meeting 380 sites reporting 5-min data (2013-2015)
completeness criteria 438 sites reporting 1-h data (2013-2015)
AQS = Air Quality System; S02 = sulfur dioxide.
2.5.2.1 Nationwide Spatial Variability
In the previous ISA for Sulfur Oxides (U.S. EPA. 2008(1). 24-hour avg, 1-hour daily max,
1-hour avg, and 5-minute hourly max SO2 concentrations measured at AQS monitoring
sites during 2003-2005 were reported. Nationwide statistics of 5-minute hourly max SO2
data were limited in the previous assessment by a scarcity of monitoring sites reporting
such data. From 2003-2005 nationwide, central statistics (mean and median) of 1-hour
daily max and 24-hour avg SO2 concentrations were generally low (less than 15 ppb),
while concentrations in the upper range of the distribution (e.g., 99th percentile) were
substantially higher (23-116 ppb), particularly for 1-hour daily max concentrations (99th
percentile: 116 ppb). In addition, 1-hour avg SO2 concentrations exhibited low mean
concentrations (4 ppb), with 99th percentile concentrations near 34 ppb. Relatively high
concentrations were typically observed at sites near stationary anthropogenic sources
(e.g., EGUs).
SO2 summary data provide a snapshot of recent concentrations and, compared with those
presented in the 2008 SOx ISA (U.S. EPA. 2008d). allow for identification of possible
trends. As shown by Table 2-6. nationwide concentrations for 2013-2015 were slightly
lower than concentrations reported in the 2008 SOx ISA. For all 24-hour avg, 1-hour
daily max, 1-hour avg, and 5-minute hourly max data pooled nationwide, mean statistics
were below 6 ppb, median statistics (50th percentile) were 2 ppb or below, and SO2
2-36
-------
concentrations in the upper range of the distribution (99th percentile) covered a wide
range of concentrations but were never greater than the primary NAAQS level of 75 ppb.
The 99th percentile 5-minute hourly max concentration was 23.8 ppb, suggesting that the
occurrence of very high 5-minute peak values is rare on the national scale. Across all
metrics, large differences were observed between mean and 99th percentile
concentrations, particularly for the SO2 1-hour daily max and 5-minute hourly max data.
Such large differences between mean and 99th percentile concentrations are consistent
with the highly variable nature of SO2, which is characterized by periodic peak
concentrations superimposed on a relatively low background concentration. Higher
concentrations in the 1-hour daily max distribution compared with the 5-minute hourly
max distribution were likely attributable to the omission of high 5-minute concentrations
from the 58 monitoring sites without 5-minute data.
The absolute highest 1-hour daily max SO2 concentration in 2013-2015 was 2,071 ppb.
Ninety-ninth percentile 1-hour daily max concentrations over 200 ppb were reported at
this site and other sites near active volcanoes in Hawaii (Table 2-6). which are discussed
further in Section 2.5.5. Other reports of 99th percentile, 1-hour daily max concentrations
greater than 200 ppb occurred at three monitoring sites near a copper smelter in Gila
County, AZ, as mentioned in Section 2.2.2. In addition, sites where the 99th percentile
1-hour daily max concentration was greater than 75 ppb were located in North Dakota,
Illinois, Iowa, Wisconsin, Arizona, Missouri, Indiana, Tennessee, Ohio, Kentucky,
Louisiana, and Pennsylvania, often near coal-fired EGUs, although the site in North
Dakota is likely influenced by a nearby shale gas processing facility. As shown in the
nationwide map in Figure 2-11. the majority of monitoring sites across the U.S. report
99th percentile, 1-hour daily max concentrations below the primary NAAQS level of
75 ppb. The 99th percentile of 24-hour avg concentrations, which are often used as
exposure metrics in epidemiologic studies, followed a similar pattern, with most elevated
values located in the industrial Midwest (Figure 2-12).
2-37
-------
Table 2-6 National statistics of sulfur dioxide concentrations (parts per billion)
from Air Quality System monitoring sites, 2013-2015.
AQS
Year NofObs Mean 5% 10% 25% 50% 75% 90% 95% 98% 99% Max Max IDa
5-min hourly max
2013 3,105,078 2.2 0.0 0.0 0.1 1.0 2.0 4.0 7.0 15.0 25.3 1,441.4 160050004
2014 3,047,302 2.2 0.0 0.0 0.2 1.0 2.0 4.0 7.0 14.7 25.0 4,208.0 160050004
2015 2,997,344 1.8 0.0 0.0 0.1 0.8 1.5 3.0 5.2 11.5 20.0 1,678.0 160050004
2013-2015 9,149,724 2.1 0.0 0.0 0.1 0.9 2.0 4.0 6.5 13.8 23.8 4,208.0 160050004
1-h avg
2013 3,105,078 1.6 0.0 0.0 0.0 0.8 1.7 3.1 5.0 9.0 15.4 2,071.0 150010007
2014 3,047,302 1.6 0.0 0.0 0.0 0.7 1.5 3.0 5.0 9.3 15.7 1,830.0 150010007
2015 2,997,344 1.3 0.0 0.0 0.0 0.6 1.1 2.5 4.0 8.0 13.0 1,779.0 150010007
2013-2015 9,149,724 1.5 0.0 0.0 0.0 0.7 1.4 3.0 4.9 9.0 15.0 2,071.0 150010007
1-h daily max
2013 134,705 5.6 0.0 0.0 0.9 2.0 4.4 10.3 18.9 37.0 62.2 2,071.0 150010007
2014 132,228 5.7 0.0 0.0 0.8 2.0 4.4 11.0 19.7 40.7 68.0 1,830.0 150010007
2015 129,789 4.7 0.0 0.0 0.6 1.4 3.3 8.1 15.7 34.2 60.0 1,779.0 150010007
2013-2015 396,722 5.3 0.0 0.0 0.7 1.8 4.0 10.0 18.0 37.4 63.5 2,071.0 150010007
2-h avg
2013 134,705 1.6 0.0 0.0 0.2 0.9 1.8 3.5 5.1 8.5 13.1 366.5 150010007
2014 132,228 1.5 0.0 0.0 0.2 0.8 1.7 3.3 5.0 8.5 13.1 317.2 150010007
2015 129,789 1.3 0.0 0.0 0.2 0.7 1.4 2.7 4.0 7.4 12.0 341.6 150010007
2013-2015 396,722 1.5 0.0 0.0 0.2 0.8 1.6 3.2 4.8 8.2 12.7 366.5 150010007
AQS = Air Quality System; avg = average; ID = identification; mean = arithmetic average; max = maximum; N = population
number; Obs = observations.
aAQS site ID number reporting the highest 3-yr concentration across the U.S.
AQS accepts as valid any reported concentration that is ± LDL. Data analyzed in this table include negative values in this range.
Note: Not all sites collect 5-min measurements. Site ID 150010007 does not collect 5-min measurements (i.e., the location of the
peak 1-hr max, 1-hr daily average and 24-hr average corresponding to a monitor adjacent to the Hawaiian volcanoes).
2-38
-------
Puerto Rico and U.S. Virgin Islands
Alaska
Monitor 1-hr Daily Max S02
2013 -2015 99th Percentile
O <= 75 ppb
O > 75 to 200 ppb
0 > 200 to 400 ppb
0 > 400 ppb
Kilometers
Max = maximum; S02 = sulfur dioxide.
Figure 2-11 Map of 99th percentile of 1 -hour daily max sulfur dioxide
concentration reported at Air Quality System monitoring sites,
2013-2015.
2-39
-------
Puerto Rico and U.S. Virgin Islands
Alaska
I Kilometers
ilnrp (iters
| Kilometers
Monitor 24-hr Daily Average
S02 2013 -2015 99th Percentile
o <= 15 ppb
o > 15 to 50 ppb
9 >50 to 100 ppb
0 > 100 ppb 0
Note: The 24-h avg concentration is a metric often used in epidemiologic studies.
S02 = sulfur dioxide.
Figure 2-12 Map of 99th percentile of 24-hour avg sulfur dioxide concentration
reported at Air Quality System monitoring sites, 2013-2015.
2.5.2.2 Urban Spatial Variability
Air quality measurements from centrally located, urban monitoring sites are often used to
represent community-scale exposure in epidemiologic analyses. However, central site
exposure estimates may not fully capture variations in pollutant concentrations over
urban scales. Urban areas differ in topography, source types, and source locations. To
illustrate the effects of these differences on urban scale concentration gradients, SO2
spatial variability was characterized in six focus areas: Cleveland, OH; Pittsburgh, PA;
New York City, NY; St. Louis, MO; Houston, TX; and Gila County, AZ. These focus
2-40
-------
areas were selected based on (1) their relevance to current health studies (i.e., areas with
peer-reviewed, epidemiologic analysis), (2) the existence of four or more monitoring sites
located within the area boundaries, and (3) the presence of several diverse SO2 sources
within a given focus area boundary.
Maps of individual focus areas indicating 99th percentile 5-minute hourly max
concentrations at monitoring sites and emissions from large point sources and their
locations are presented in Figures 2-13 through 2-18. As shown by the maps, up to
12 SO2 monitoring sites are located in individual focus areas. Monitoring sites in each
focus area are located at various distances from SO2 sources. Due to the relatively short
atmospheric lifetime of SO2, monitoring sites adjacent to large point sources
(e.g., electric generating units, industrial sources, copper smelting facilities, integrated
iron and steel mills, shipping ports) are expected to detect higher SO2 concentrations than
those farther downwind. However, other variables, particularly stack height and wind
speed and direction, influence concentrations observed near sources. For example,
Sites C and E in Cleveland are both adjacent to large sources, but Site C has a much
lower concentration than Site E despite the source near Site C emitting much more SO2
than the source near Site E.
2-41
-------
0 1.25 2.5 Kilometers
1 I
B - 61 ppb
D - 52 ppb
o
F-15 ppb
o
- 30 ppb
Esri, HERE, DeLorme.
Mapmylndia, © OpenStreetMap
contributors, and the GIS user
community
25 Kilometers
, ^ cV
>
E - 86 ppb
o
3-22125 tpy
- C - 29 ppb
1-34932 tpy
2 - 2404 tpy
S02 Emission Sources
~ Electricity Generation via Combustion
~ Steam/Heating Facility
S02 Concentration (ppb)
O 3 to 100
© GT 100 to 200
• GT 200 to 300
• GT 300 to 400
Oq
Cleveland
Esri. HERE, DeLorme, Mapmylndia, © OpenStreetMap contributors, and the GIS user community
Note: Blue circles denote monitoring sites included in the U.S. Air Quality Monitoring System. Triangles denote sources emitting
2,000 tpy or more according to the 2014 U.S. National Emissions Inventory, The inset, lower right, displays a wind rose of average
wind speed and direction for data acquired at Cleveland Hopkins international Airport over the 3-yr period 2013-2015.
Figure 2-13 Map of the Cleveland, OH focus area showing emissions from
large sources and the 99th percentile 5-minute hourly max
concentration at ambient monitors during 2013-2015.
2-42
-------
Table 2-7
Largest SO2 emissions sources, Cleveland, OH (as noted in
Fiqure 2-13).
Map Code
Facility Name
Facility Type
SO2 Emissions (tpy)
1
Avon Lake Power Plant (0247030013)
EGU—Combustion
34,932
2
The Medical Center Company
(1318003059)
Steam/Heating Facility
2,404
3
Cleveland Electric Illuminating Company,
Eastlake Plant (0243160009)
EGU—Combustion
22,126
EGU = electric power generating uni; S02 = sulfur dioxide.
1 - 3960 tpy
~
20 Kilometers
_l
F - 26 ppb
o
± 2 - 19784 tpy
~
7 -28138 tpy
3-10263tpy C
A G - 17 ppb
oA.
5 - 4445 tpy
~
E-22 ppb Q H -11 ppb
A -12 ppb
o
D -10 ppb
4-10660tpy
Pittsburgh
Bethel Park G
B - 56 ppb
C -18 ppb
Wheeling
S02 Emission Sources
~ Electricity Generation via Combustion
S02 Concentration (ppb)
O 3 to 100
• GT 100 to 200
• GT 200 to 300
• GT 300 to 400
6 - 4599 tpy
Esri. HERE. DeLorme. Mapmylndta. ffi OpenStfeetMap contributors and the GIS user community
Note: Blue circles denote monitoring sites included in the U.S. Air Quality Monitoring System, Triangles denote sources emitting
2,000 tpy or more according to the 2014 U.S. National Emissions Inventory. The inset, lower right, displays a wind rose of average
wind speed and direction for data acquired at Pittsburgh International Airport over the 3-yr period 2013-2015.
Figure 2-14 Map of the Pittsburgh, PA focus area showing emissions from
large sources and the 99th percentile 5-minute hourly max
concentration at ambient monitors during 2013-2015.
2-43
-------
Table 2-8 Largest SO2 emissions sources, Pittsburgh, PA (as noted in
Figure 2-14).
Map Code
Facility Name
Facility Type
SO2 Emissions (tpy)
1
NRG Power Midwest LP/New Castle Power Plant
EGU—Combustion
3,960
2
FirstEnergy Gen LLC/Bruce Mansfield Plant
EGU—Combustion
19,784
3
W. H. Sammis Plant
EGU—Combustion
10,263
4
Cardinal Power Plant (Cardinal Operating Company)
EGU—Combustion
10,660
5
NRG Midwest LP/Cheswick
EGU—Combustion
4,445
6
Monongahela Power Company—Fort Martin Power
EGU—Combustion
4,599
7
Genon NE Mgmt Company/Keystone Station
EGU—Combustion
28,138
EGU = electric generating units; S02 = sulfur dioxide.
2-44
-------
Paterson
10 Kilometers
A -13 ppb
o
K -10 ppb
B - 6 ppb D " 6 PPb
O.
Newark Jersey c|(y
F - 5 ppb
0 C J
Elizabeth 6 C " 9 PPb
G -10 ppb
r\
w
J - 9 ppb
New York
Esri, HERE, DeLorme, Mapmylndia, ©
OpenStreetMap contributors, and the GIS user
community
0 10 20 Kilometers
1 I
I - 4 ppb
1-3181 tpy
E - 5 ppb
Yonkers
Q
0
Newark
(P^; Vj
S02 Emission Sources
~ Electricity Generation via Combustion
S02 Concentration (ppb)
O 3 to 100
O GT 100 to 200
• GT 200 to 300
• GT 300 to 400
Bridgeport
Stamford Norwalk
L - 6 ppb
H - 5 ppb
Esri. HERE, DeLorme, Mapmylndia, © OpenStreetMap contributors, and the GIS user community
Note: Blue circles denote monitoring sites included in the U.S. Air Quality Monitoring System. Triangles denote sources emitting
2,000 tpy or more according to the 2014 U.S. National Emissions Inventory. The inset, lower right, displays a wind rose of average
wind speed and direction for data acquired at Newark International Airport over the 3-yr period 2013-2015.
Figure 2-15 Map of the New York City, NY focus area showing emissions from
large sources and the 99th percentile 5-minute hourly max
concentration at ambient monitors during 2013-2015.
2-45
-------
Table 2-9 Largest SO2 emissions source, New York, NY (as noted in
Figure 2-15).
Map Code
Facility Name
Facility Type
SO2 Emissions (tpy)
NRG REMA LLC/Portland Generating Station EGU—Combustion
3,181
EGU = electric power generating unit; S02 = suifur dioxide.
1 -7122 tpy '
C - 23 ppb
B -14 ppb
F-15 ppb
St. Louis
2 - 2867 tpy
10 Kilometers
S02
A
S02
o
Emission Sources
Breweries/Distilleries/Wineries
Electricity Generation via Combustion
Mineral Processing Plant
Concentration (ppb)
3 to 100
GT 100 to 200
GT 200 to 300
GT 300 to 400
A - 6 ppb
o E-27 ppb
Q D -15 ppb
Esri, HERE, DeLorme, Mapmylndia, ©
OpenStreetMap contributors, and the
GIS user community
3 - 33091 tpy
~ Ballwin
Wildwood
Granite City
On
Oakville
~
4-11702 tpy
G - 80 ppb
A
6 - 4406 tpy
7 - 5696 tpy
5-17444tpy
~
8 - 3285 tpy
25 Kilometers
_l
Esri, HERE, DeLorme, Mapmylndia, © OpenStreetMap contributors, and the GIS user community
Note: Blue circles denote monitoring sites included in the U.S. Air Quality Monitoring System. Triangles denote sources emitting
2,000 tpy or more according to the 2014 U.S. National Emissions Inventory. The inset, upper right, displays a wind rose of average
wind speed and direction for data acquired at Lambert—St. Louis International Airport over the 3-yr period 2013-2015.
Figure 2-16 Map of the St Louis, MO-IL focus area showing emissions from
large sources and the 99th percentile 5-minute hourly max
concentration at ambient monitors during 2013-2015.
2-46
-------
Table 2-10 Largest SO2 emissions sources, St. Louis, MO-IL (as noted in
Fiqure 2-16).
Map Code
Facility Name
Facility Type
SO2 Emissions (tpy)
1
Dynegy Midwest Generation LLC
EGU—Combustion
7,122
2
Anheuser-Busch Inc-St. Louis
Breweries/Distilleries/Wineries
2,867
3
Ameren Missouri-Labadie Plant
EGU—Combustion
33,091
4
Ameren Missouri-Meramec Plant
EGU—Combustion
11,702
5
Ameren Missouri-Rush Island Plant
EGU—Combustion
17,444
6
Dynergy Midwest Generation
EGU—Combustion
4,406
7
Prairie State Generating Station
EGU—Combustion
5,696
8
Mississippi Lime Company—Saint
Genevieve
Mineral Processing Plant
3,285
EGU = electric generating units; S02 = sulfur dioxide.
2-47
-------
0 15 30 Kilometers
1 l
O
H - 9 ppb " I - 6 ppb
0
Sugar Land
E - 30 ppb 2 - 2203 tpy
Houston ^
w Bay town
O
D- 25 ppb O
B -12 ppb Pasadena F-9ppb
f~*,i A
C -12 ppb
O
G - 6 ppb
Pea rland
1 - 43980 tpy
A League City
A -16 ppb
NORTH ,.v
- 20*
/ '16%
'N A - 8%
\VS5T »«8L. EAST.
S02 Emission Sources
Jk
A Electricity Generation via Combustion
A Petroleum Refinery
S02 Concentration (ppb)
O 3 to 100
WNDSPS)
(m's)
¦ 8.8 -11.1
SOUTH- ¦ 5'7-8-8
¦ 3.3- 5.7
2.1 - 3.6
O GT 100 to 200
__ 0.5 - 2.1
Calms: 16.11%
• GT 200 to 300
• GT 300 to 400
Esri. HERE, DeLorme, Mapmylndia, © OpenStreetMap contributors, and the GlS user community
Note: Blue circles denote monitoring sites included in the U.S. Air Quality Monitoring System. Triangles denote sources emitting
2,000 tpy or more according to the 2014 U.S. National Emissions Inventory. The inset, lower right, displays a wind rose of average
wind speed and direction for data acquired at George Bush Intercontinental Airport over the 3-yr period 2013-2015.
Figure 2-17 Map of the Houston, TX focus area showing emissions from large
sources and the 99th percentile 5-minute hourly max
concentration at ambient monitors during 2013-2015.
Table 2-11 Largest SO2 emissions source, Houston, TX (as noted in Figure 2-17).
Map Code Facility Name
Facility Type
SO2 Emissions (tpy)
1 WA Parish Electric Generating Station
EGU—Combustion
43,980
2 Baytown Refinery
Petroleum Refinery
2,203
EGU = electric generating units; S02 = sulfur dioxide.
2-48
-------
25 Kilometers
0 1.25 2.5 Kilometers
1 I
1 - 4505 tpy
A
Esri, HERE, DeLorme. Mapmylndia, ©
OpenStreetMap contributors, and the
GIS user community
Kearny
S02 Emission Sources
A Primary Copper Smelting/Refining Plant
S02 Concentration (ppb)
© 3 to 100
O GT 100 to 200
• GT 200 to 300
• GT 300 to 400
B - 282 ppb
%
2-17432 tpy
WIND SP3
(m/s)
m
Esri, HERE, DeLorme, Mapmylndia, © OpenStreetMap contributors, and the GIS user community
Note: Blue circles denote monitoring sites included in the U.S. Air Quality Monitoring System. Yellow triangle denote sources
emitting 2,000 tpy or more according to the 2011 U.S. National Emissions Inventory. The inset, lower right, displays a wind rose of
average wind speed and direction for data acquired at the Phoenix Sky Harbor Intercontinental Airport over the 3-yr period
2013-2015.
Figure 2-18 Map of the Gila County, AZ focus area showing emissions from
large sources and the 99th percentile 5-minute hourly max
concentration at ambient monitors during 2013-2015.
2-49
-------
Table 2-12
Largest SO2 emissions sources, Gila County, AZ (as noted in
Fiqure 2-18).
Map Code
Facility Name
Facility Type
SO2 Emissions (tpy)
1
Freeport McMoran Miami Smelter
Primary Copper
Smelting/Refining
Plant
4,505
2
Asarco, LLC—Hayden Smelter
Primary Copper
Smelting/Refining
Plant
17,432
S02 = sulfur dioxide.
Table 2-13 provides the distribution of 1-hour daily max SO2 concentrations and monitor
type (standard vs. trace level monitor) reported at individual AQS sites in the six focus
areas. Concentrations reported at these sites were similar to nationwide SO2
concentrations discussed earlier in this section (Section 2.5.2.1V For all but one
individual monitoring site, median concentrations were below 15 ppb. The one exception
was the monitoring site in the Gila County, AZ focus area, for which the median
concentration was 39 ppb. This particular monitoring site (Site B) is located within 1 km
of a copper smelting plant with markedly high annual SO2 emissions [greater than
17,000 tpy S02; (U.S. EPA. 2016b)l.
Table 2-13 1-h daily max sulfur dioxide concentration distribution by Air Quality
System monitoring site in six focus areas, 2013-2015.
AQS
Monitoring N of
Site Label Site ID Obs Mean Min 10% 25% 50% 75% 90% 99% Max Monitor Type
Cleveland-Elyria-Mentor, OH
A 390350065 709 6.4 0.0 0.0 1.0 3.0 7.0 13.2 55.9 125.0 Standard
B
390350060
887
11.5
0.0
0.0
2.0
6.0
16.0
32.0
62.1
92.0
Standard
C
390850003
758
7.6
0.0
2.0
3.0
6.0
10.0
15.0
37.4
95.0
Standard
D
390350038
786
14.0
0.0
1.0
4.0
10.0
20.0
32.5
61.3
105.0
Standard
E
390850007
901
11.2
0.0
2.0
3.0
6.0
11.0
22.0
117.0
201.0
Standard
F
390350045
630
3.9
0.0
0.0
0.0
2.0
5.0
9.0
30.0
51.0
Standard
2-50
-------
Table 2-13 (Continued): 1 h daily max sulfur dioxide concentration distribution by
Air Quality System monitoring site in six focus areas,
2013-2015.a
AQS
Monitoring N of
Site Label Site ID Obs Mean Min 10% 25% 50% 75% 90% 99% Max Monitor Type
Pittsburgh, PA
A 421255001 1,020 3.6 0.0 0.0 0.0 3.0 5.0 9.0 17.0 53.0 Standard
B 420030064 1,076 16.6 0.0 2.0 4.0 11.0 21.0 39.5 90.8 244.0 Standard
C 421250005 1,044 6.1 0.0 2.0 3.0 4.0 7.0 11.0 33.6 61.0 Standard
D 420030067 1,069 3.4 0.0 0.0 1.0 2.0 4.0 7.0 19.0 55.0 Standard
E 420030002 1,090 5.9 0.0 1.0 2.0 4.0 7.0 12.0 41.0 75.0 Standard
F 420070005 1,014 7.0 0.0 0.0 1.0 4.0 10.0 17.0 40.0 80.0 Standard
G 420070002 1,028 5.6 0.0 1.0 2.0 4.0 8.0 12.0 24.7 45.0 Standard
H 420030008 706 4.0 0.0 0.9 1.7 2.8 4.5 7.7 20.2 100.3 Trace
New York-Northern New Jersey-Long Island, NY-NJ-PA
A 360050133 1,089 4.0 0.2 0.9 1.5 2.8 5.3 8.9 16.5 26.5 Standard
B 340130003 1,089 1.8 0.0 0.3 0.6 1.3 2.4 3.9 7.8 13.0 Trace
C 340170006 725 1.4 0.0 0.0 0.0 1.0 2.0 4.0 9.0 11.0 Standard
D 340171002 1,090 1.4 0.0 0.0 0.0 1.0 2.0 4.0 8.0 11.0 Standard
E 340273001 1,065 1.4 0.0 0.0 0.0 1.0 2.0 3.0 9.0 20.0 Standard
F 340390003 1,089 1.3 0.0 0.0 0.0 1.0 2.0 3.0 6.0 12.0 Standard
G 340390004 1,081 2.3 0.0 0.0 1.0 1.0 3.0 5.0 13.2 109.0 Standard
H 360590005 1,001 2.0 0.2 0.8 1.1 1.5 2.3 3.6 8.3 14.6 Standard
I 360790005 1,083 1.2 0.1 0.4 0.6 0.8 1.3 2.2 5.8 10.3 Standard
J 360810124 1,086 2.5 0.0 0.5 0.9 1.7 3.3 5.4 11.0 18.5 Trace
K 360050110 1,077 3.1 -0.2 0.8 1.2 2.2 4.1 6.7 14.2 32.1 Standard
L 361030009 938 1.6 -0.6 0.1 0.4 1.0 2.3 4.0 8.7 15.8 Standard
St. Louis, MO-IL
A 171170002 646 2.2 0.0 0.8 1.0 2.0 3.0 4.0 8.5 21.0 Standard
B 171191010 1,023 4.1 0.0 0.9 1.3 3.0 5.0 9.0 18.0 40.0 Standard
2-51
-------
Table 2-13 (Continued): 1 h daily max sulfur dioxide concentration distribution by
Air Quality System monitoring site in six focus areas,
2013-2015.a
AQS
Monitoring N of
Site Label Site ID Obs Mean Min 10% 25% 50% 75% 90% 99% Max Monitor Type
C 171193007 1,041 5.6 0.0 1.0 2.0 4.0 7.0 11.6 24.4 42.0 Standard
D 171630010 1,018 4.7 0.0 1.0 2.0 3.6 6.0 10.0 20.8 30.0 Standard
E 295100085 921 7.2 0.0 1.3 2.4 4.2 9.1 16.5 40.2 51.4 Trace
F 295100086 1,077 4.5 0.5 1.2 1.8 3.3 5.6 9.5 19.6 31.8 Standard
G 290990027 1,089 11.6 0.3 1.1 2.2 4.2 8.8 36.3 94.5 252.7 Standard
Houston-Sugar Land-Baytown, TX
A 481670005 736 3.6 0.3 1.0 1.5 2.4 3.8 6.8 26.5 50.6 Standard
B 482010051 214 3.1 0.0 0.7 1.0 1.9 3.4 6.1 22.2 44.4 Standard
C 482010062 160 3.7 0.4 1.0 1.7 2.4 4.4 7.9 18.0 19.3 Standard
D 482010416 313 5.5 0.3 0.9 1.6 3.4 6.9 12.1 33.6 54.0 Standard
E 482011035 71 4.9 0.3 0.5 1.5 2.4 5.4 13.1 25.9 29.8 Standard
F 482011039 590 2.2 0.0 0.2 0.7 1.6 2.9 5.2 11.0 16.0 Trace
G 482011050 885 1.9 0.2 0.5 0.7 1.4 2.4 3.8 9.0 16.4 Standard
H 482010046 15 3.5 1.8 1.9 2.3 2.8 3.2 4.7 12.0 13.1 Standard
482011017 415 1.5 0.0 0.4 0.6 1.0 1.9 3.3 8.3 10.6 Standard
Gila County, AZ
A 40070009 1,080 24.9 0.0 2.0 3.0 12.0 34.3 64.0 153.2 259.0 Standard
B 40071001 889 50.8 0.0 1.0 13.0 39.0 71.0 114.2 247.2 368.0 Trace
C 40070011 739 28.5 0.0 1.0 2.0 9.0 36.0 84.0 204.9 380.0 Trace
D 40070012 630 31.3 0.0 1.0 2.0 8.0 39.8 95.0 230.7 324.0 Trace
AQS = Air Quality System; ID = identification; max = maximum; mean = arithmetic average; min = minimum; N = population number;
Obs = observations.
AQS accepts, as valid, reported concentrations that are ± LDL. Data analyzed in this table include negative values in this range.
2-52
-------
More substantial site-to-site differences were observed in the 99th percentile of SO2
concentrations. Across these monitoring sites, 1-hour daily max 99th percentile
concentrations ranged from 5.8 to 247.2 ppb, with the majority of sites exhibiting 99th
percentile concentrations at or below 40 ppb. Relatively high 99th percentile
concentrations were reported at monitoring sites within 5 km of a large SO2 point source,
particularly in Gila County, AZ near copper smelters. Relatively high 99th percentile
concentrations were also observed in the Cleveland, OH and Pittsburgh, PA focus areas.
These data were in agreement with previous studies, which generally observed higher
urban SO2 concentrations near local industrial/combustion sources related to oil-burning
units, smelters, and EGUs (Brand et al.. 2016; Cloughertv et al.. 2013; Wheeler et al..
2008).
Over the past decade, the number of AQS monitoring sites reporting 5-minute SO2
concentrations has substantially increased. At the time of the 2008 SOx ISA (U.S. EPA.
2008d), a total of 98 monitoring sites periodically reported 5-minute hourly max
concentrations. To date, approximately 380 sites report 5-minute data, including urban
sites within focus areas, sites near city centers, and sites near SO2 sources (Figure 2-10 in
Section 2.4.3).
Similar analyses of 5-minute hourly max concentrations were performed on more recent
data reported at individual monitoring sites in the six focus areas. Table 2-14 shows the
range in 5-minute hourly max SO2 concentrations reported at individual monitors, within
the six focus areas in the 2013-2015 time frame. Median 5-minute hourly max
concentrations are below 5 ppb, while maximum concentrations range from 15 to
1,241 ppb.
Table 2-14 5-minute hourly max sulfur dioxide concentrations by Air Quality
System monitoring sites in select focus areas, 2013-2015.
AQS
Site Monitoring
Label Site ID
N of
Obs
Mean
Min
10%
25%
50%
75%
90%
99%
Max
Monitor Type
Cleveland-Elyria-Mentor, OH
A 390350065
16,201
3.7
0.0
0.0
0.0
2.0
5.0
8.0
27.0
397.0
Standard
B 390350060
18,585
4.9
0.0
0.0
0.0
1.0
4.0
13.0
53.0
159.0
Standard
C 390850003
15,966
3.6
0.0
0.0
1.0
2.0
5.0
8.0
26.0
241.0
Standard
D 390350038
17,321
6.0
0.0
0.0
0.0
2.0
7.0
16.0
49.0
180.0
Standard
2-53
-------
Table 2-14 (Continued) 5 minute hourly max sulfur dioxide concentrations by Air
Quality System monitoring sites in select focus areas,
2013-2015
AQS
Site Monitoring N of
Label Site ID Obs Mean Min 10% 25% 50% 75% 90% 99% Max Monitor Type
E 390850007 19,297 5.6 0.0 0.0 1.0 3.0 5.0 9.0 69.0 428.0 Standard
F 390350045 13,720 1.5 0.0 0.0 0.0 0.0 2.0 4.0 15.0 131.0 Standard
Pittsburgh, PA
A 421255001 24,367 1.5 0.0 0.0 0.0 0.0 2.0 4.0 12.0 73.0 Standard
B 420030064 25,602 6.1 0.0 0.0 1.0 2.0 7.0 16.0 56.0 493.0 Standard
C 421250005 24,930 3.3 0.0 1.0 1.0 2.0 4.0 6.0 18.0 137.0 Standard
D 420030067 25,480 1.4 0.0 0.0 0.0 1.0 2.0 4.0 10.0 89.0 Standard
E 420030002 26,001 2.4 0.0 0.0 0.0 1.0 3.0 6.0 22.0 112.0 Standard
F 420070005 24,264 3.1 0.0 0.0 0.0 1.0 3.0 8.0 26.0 155.0 Standard
G 420070002 24,572 2.2 0.0 0.0 0.0 1.0 3.0 6.0 17.0 64.0 Standard
H 420030008 16,095 1.7 -0.2 0.1 0.4 1.0 2.2 3.8 10.7 158.3 Trace
New York-Northern New Jersey-Long Island, NY-NJ-PA
A 360050133 25,699 2.5 0.0 0.4 0.8 1.5 3.2 5.8 13.0 32.3 Standard
B 340130003 25,928 0.9 0.0 0.1 0.2 0.5 1.2 2.3 5.7 23.1 Trace
C 340170006 17,200 0.8 0.0 0.0 0.0 0.0 1.0 3.0 9.0 29.0 Standard
D 340171002 25,826 1.0 0.0 0.0 0.0 1.0 1.0 2.0 6.0 34.0 Standard
E 340273001 24,451 1.2 0.0 0.0 1.0 1.0 1.0 2.0 5.0 58.0 Standard
F 340390003 25,887 1.2 0.0 0.0 0.0 1.0 2.0 3.0 5.0 47.0 Standard
G 340390004 25,748 1.4 0.0 0.0 0.0 1.0 2.0 3.0 10.0 317.0 Standard
H 360590005 23,683 1.4 0.1 0.6 0.8 1.1 1.6 2.3 5.3 21.5 Standard
I 360790005 25,630 0.9 0.0 0.4 0.5 0.7 1.0 1.4 3.6 16.1 Standard
J 360810124 25,557 1.5 -0.1 0.1 0.3 0.8 1.9 3.8 9.0 26.8 Trace
K 360050110 25,333 2.0 -1.2 0.3 0.7 1.4 2.6 4.3 10.0 46.6 Standard
L 361030009 22,128 1.2 -0.7 0.1 0.4 0.9 1.7 2.9 6.4 30.5 Standard
St. Louis, MO-IL
A 171170002 14,260 1.5 0.0 0.5 1.0 1.2 2.0 2.7 6.0 56.0 Standard
2-54
-------
Table 2-14 (Continued) 5 minute hourly max sulfur dioxide concentrations by Air
Quality System monitoring sites in select focus areas,
2013-2015
Site
Label
AQS
Monitoring
Site ID
N of
Obs
Mean
Min
10%
25%
50%
75%
90%
99%
Max
Monitor Type
B
171191010
22,801
1.7
0.0
0.0
0.0
0.9
2.0
4.0
15.0
240.0
Standard
C
171193007
23,684
2.7
0.0
0.0
0.8
1.3
3.0
6.0
24.0
94.0
Standard
D
171630010
22,691
1.9
0.0
0.0
0.0
1.0
2.0
4.2
15.0
87.4
Standard
E
295100085
20,653
3.3
0.0
0.6
1.2
2.0
3.3
6.3
26.5
93.7
Trace
F
295100086
25,720
2.4
0.2
0.8
1.1
1.5
2.5
4.5
15.2
53.0
Standard
G
290990027
26,002
5.7
0.2
0.5
0.9
2.1
3.6
8.0
80.4
657.1
Standard
Houston-Sugar Land-Baytown, TX
A
481670005
16,307
1.9
0.0
0.4
0.6
1.1
2.1
3.6
15.8
84.9
Standard
B
482010051
4,523
1.1
0.0
0.2
0.3
0.6
1.2
2.3
10.3
65.9
Standard
C
482010062
3,399
1.6
0.0
0.3
0.5
1.0
1.8
3.1
12.5
33.4
Standard
D
482010416
6,982
2.4
0.0
0.3
0.6
1.0
2.3
5.2
24.1
90.9
Standard
E
482011035
1,482
2.4
0.0
0.3
0.5
1.0
2.3
4.4
26.3
75.8
Standard
F
482011039
12,547
0.9
0.0
0.0
0.0
0.5
1.1
2.2
6.8
25.7
Trace
G
482011050
19,894
1.0
0.0
0.3
0.4
0.6
1.1
2.1
5.7
21.3
Standard
H
482010046
313
1.8
0.0
0.3
0.5
1.5
2.6
3.3
7.2
15.2
Standard
I
482011017
8,728
0.7
0.0
0.0
0.2
0.4
0.8
1.5
5.0
25.3
Standard
Gila County, AZ
A
40070009
25,732
9.2
0.0
1.0
1.0
3.1
4.5
21.6
115.5
461.0
Standard
B
40071001
20,222
19.6
0.0
0.0
1.0
2.0
10.6
55.0
252.2
1,241.2
Trace
C
40070011
16,630
9.1
0.0
0.0
0.0
1.0
3.0
22.0
142.1
694.0
Trace
D
40070012
14,156
7.6
0.0
0.0
1.0
1.0
2.0
11.0
148.0
993.0
Trace
AQS = Air Quality System; ID = identification; max = maximum; mean = arithmetic average; min = minimum; N = population number;
Obs = observations.
AQS accepts as valid reported concentrations that are ± LDL. Data analyzed in this table include negative values in this range.
2-55
-------
To evaluate the extent of SO2 spatial variability over urban geographical scales,
concentration correlations between monitoring site pairs were calculated in each of the
six focus areas. To estimate the degree to which concentrations at two different
monitoring sites followed similar temporal trends, pairwise comparisons were evaluated
using Pearson correlations. Across the six focus areas, Pearson correlations ranged from 0
to 1.0 for 24-hour avg data. Correlations close to 1 represent strong correspondence over
time between pairwise monitoring site concentrations, while values close to 0 represent
poor correspondence between concentrations. Figures 2-19 and 2-20 show scatterplots of
pairwise correlations of 24-hour avg and 5-minute hourly max SO2 concentrations,
respectively, versus distance between monitoring site pairs. The 24-hour avg
concentrations are presented due to their frequent use in epidemiologic studies, while
5-minute hourly max concentrations are a metric of interest for short-duration exposures.
Given the meandering nature of SO2 plumes and potential for plume touchdown several
kilometers from the stack (Turner. 1970). low correlation among monitoring sites would
be expected in most cases for the 5-minute hourly max data.
Intersite pairwise comparisons in Figure 2-19 suggest high spatial variability of the
24-hour avg SO2 concentration time series, consistent with long-standing observations of
the movement of emissions plumes. In every focus area except for New York (discussed
below), low to moderate intersite pairwise correlations of 24-hour avg SO2 concentration
data were observed, with the majority of Pearson correlations below 0.6. Intersite
pairwise correlations tended to decrease with distance. Even within relatively short
distances (up to 15 km), most intersite pairwise correlations were low, reflecting the
variable nature of ambient SO2 across urban spatial scales, possibly due to short
atmospheric residence time, variable meteorology, and the episodic nature of the
emissions as discussed in Section 2.2.
In comparison, 5-minute hourly max SO2 concentrations had somewhat higher spatial
variability across urban spatial scales (Figure 2-20). In most cases, intersite pairwise
correlations of 5-minute hourly max concentrations are lower (less than 0.4) and decline
more dramatically with distance than intersite pairwise correlations of 24-hour avg
concentrations.
2-56
-------
Cleveland
100
Distance (femt
O d
Pittsburgh
/« • ,
• •
—i—
50
100
Distance (km)
150
New York
c
-------
00
o
I
a
O
o
b
Cleveland
50 100
Distance (km)
o o
O Tf
O b
o
o
Pittsburgh
50 100
Distance (km)
New York
-------
While spatial variability is evident to some degree in all urban areas, the extent of this
variability is location dependent. For example, pairwise correlations in Cleveland, OH
and St Louis, MO indicate strong SO2 spatial heterogeneity. In comparison, pairwise
correlations in New York City, NY are generally high and uniform across more than
100 km despite sometimes large distances between monitoring sites. Stronger pairwise
correlations in New York City, NY may be related to similar temporal source patterns,
given that the focus area's smaller power plants (<2,000 tpy SO2 emissions), including
gas-coal cogeneration facilities in Brooklyn, NY and Sayreville, NJ; an oil-burning
facility in Queens, NY; a coal-fired power plant in Jersey City, NJ; and numerous homes
using oil-burning heat likely have similar periods of high operation across the
metropolitan area. This is analogous to observations about similarities in traffic patterns
across large distances that promote higher correlation despite distance between the
sources (Sarnat et al.. 2010). Conversely, high spatial variations in Cleveland, OH and St.
Louis, MO may be explained by the presence of a limited number of sources (>2,000 tpy)
located at unevenly distributed sites across the metropolitan area.
In summary, SO2 concentrations vary substantially across urban spatial scales as
evidenced by poor to moderate intersite pairwise correlations observed in SO2 data in six
focus areas. Spatial heterogeneity in urban-scale SO2 concentrations and their temporal
patterns may be explained by the presence of multiple, unevenly distributed SO2 sources,
meteorological factors that lead to varying degrees of SO2 dilution, or removal through
cloud/fog chemistry and deposition. Additionally, in this analysis, metrics representing
maximum SO2 concentrations generally exhibited more spatial heterogeneity than
24-hour avg metrics.
2.5.3 Temporal Variability
Temporal variations in outdoor SO2 concentrations affect the magnitude, duration, and
frequency in which humans are exposed to SO2. In this section, different types of
temporal trends are discussed, spanning long-term temporal trends on an annual basis to
short-term trends on a subhourly basis.
2.5.3.1 Long-Term Trends
Trends in SO2 concentrations reported at AQS monitoring sites across the U.S. from 1980
to 2015 are shown in Figure 2-21 for the annual 99th percentile of the 1-hour daily max
SO2 concentration. Information on SO2 concentration trends at individual, local air
2-59
-------
monitoring sites can be found at https://www. epa. gov/ai r-trends/su 1 fur-dioxidc-trends
(U.S. EPA. 2012b).
20101-hr NAAQS
Note: The solid line shows the mean concentrations and the upper and lower dashed lines represent the 10th and 90th percentile
concentrations, respectively. The red line indicates the current NAAQS for sulfur oxides. NAAQS = National Ambient Air Quality
Standards.
Source: https://www.epa.gov/air-trends/sulfur-dioxide-trends.
Figure 2-21 National sulfur dioxide air quality trend, based on the 99th
percentile of the 1-hour daily max concentration for 163 sites,
1980-2015. A 76% decrease in the national average was observed
from 1990-2015.
The steady decline in SO2 concentrations over the past 25 years is largely attributed to
emissions reductions at EGUs due to the Acid Rain and NOx Budget Trading programs,
CSAPR and other national interstate transport rules that have been implemented under
the Clean Air Act Amendments of 1990 (USC Title 42 Chapter 85). Reductions in SO2
emissions commenced in 1996 and continued into the 2000s, resulting in dramatic
decreases in total, nationwide SO2 emissions and concentrations (Figure 2-5). The NOx
Budget Trading Program and national interstate transport rules led to further reductions in
2-60
-------
SO2 emissions. From 1990-2014, the annual 99th percentile average of 1-hour daily max
SO2 concentration has decreased by 76% nationally.
Substantial declines in SO2 concentration over the past decades have also been observed
on regional scales. Blanchard et al. (2013) reported an average decline of 7.6% per year
(±1.6%) in SO2 emissions from 1999-2010 across four southeastern U.S. states
(Alabama, Florida, Georgia, Mississippi), primarily due to reductions in power plant
emissions, which account for approximately 75% of total SO2 emissions in the
southeastern U.S. region. This decline corresponded to large reductions in annual SO2
concentrations (between 5.1 and 9.7% per year) reported at monitoring sites across these
four states.
2.5.3.2 Seasonal Trends
In the 2008 SOx ISA (U.S. EPA. 2008d). month-to-month trends for SO2 were observed
across a number of metropolitan areas, and these seasonal profiles varied by location.
Some cities, such as Steubenville, OH and Phoenix, AZ showed clear wintertime
maxima, while other urban areas (Philadelphia, PA; Los Angeles, CA; Riverside, CA)
exhibited higher SO2 concentrations during summer months. Differences in seasonal
profiles were attributed to variations in source emissions, topography, and meteorological
conditions among different areas.
Month-to-month variability based on more recent 1-hour daily max concentrations
(2013-2015) is shown for the six focus areas introduced earlier in this chapter
(Section 2.5.2.2). Figure 2-22 displays the range of SO2 concentrations reported at all
monitoring sites within each focus area.
The data indicate that 1-hour daily max SO2 concentrations vary across seasons,
especially in the higher concentrations within monthly SO2 concentration distributions.
Among the five urban focus areas, mean concentrations (red circle) varied by no more
than 10 ppb throughout the year, while the median concentration in the Gila County, AZ
focus area varied by 30 ppb. Large variations across all focus areas are observed in the
upper end (greater than 75th percentile) of SO2 concentrations. Notably, mean monthly
SO2 concentrations were higher and more variable than median values, indicating that the
distribution is skewed by high, infrequent observations.
2-61
-------
Pittsburgh
Cleveland
JFMAMJ JASOND
8 T
New York
St. Louis
S
I
S
e
3
JFMAMJJASOND
Houston
8 t
• •
(i #
FMAMJ JASOND
Gila County
Note: For every month, arithmetic mean concentrations are displayed as red circles. The whiskers represent the 95% confidence
interval.
Figure 2-22 Sulfur dioxide month-to-month variability based on 1-hour daily
max concentrations at Air Quality System sites in each focus
area, 2013-2015.
2-62
-------
Recent data further demonstrate that seasonal profiles vary by location. While each focus
area exhibits some degree of seasonal variation, no consistent seasonal profile was
observed across the focus areas. For example, springtime maxima in 1-hour daily max
SO2 are evident in Cleveland, OH and Gila County, AZ, corresponding to focus areas
with the highest SO2 concentrations. Alternatively, New York City, NY; Houston, TX;
and Pittsburgh, PA show clear wintertime maxima.
Month-to-month variations in SO2 concentrations are consistent with month-to-month
emissions patterns (Lee etal.. 201 la) and the atmospheric chemistry of SO2.
Summertime minima, observed in the New York City, NYC and Houston, TX focus
areas, may correspond to enhanced oxidation of SO2 to SO42 by photochemically derived
atmospheric oxidants that are more prevalent during the humid summer (Khoder. 2002).
The difference in seasonality among these cities suggest that SO2 can be substantially
variable across local and regional scales.
2.5.3.3 Diel Variability
The 2008 SOx ISA (U.S. EPA. 2008d) explored nationwide patterns in diel variability of
SO2 concentrations (i.e., variability of SO2 concentrations across a 24-hour period), and
found clear daytime maxima and nighttime minima, with larger day-night differences
with increasing SO2 concentrations. Daytime maxima were attributed to entrainment of
SO2 from elevated point sources (e.g., power plants and industrial sources) into the mixed
boundary layer, which expands due to rising surface temperatures.
Diel patterns (i.e., patterns of change in ambient SO2 concentrations over a 24-hour
period) were investigated in the focus areas using 1-hour avg and 5-minute hourly max
SO2 data for the 2013-2015 time frame. Figures 2-23 and 2-24 show variations in 1-hour
avg and 5-minute hourly max SO2 concentrations, respectively, in the six focus areas.
Consistent with the nationwide diel patterns reported in the 2008 SOx ISA (U.S. EPA.
2008d). SO2 concentrations in the six focus areas were generally low during nighttime
and approach maxima values during daytime hours (Figures 2-23 and 2-24). In
Pittsburgh, PA; New York City, NY; St. Louis, MO; Houston, TX; and Gila County, AZ,
daytime maxima occurred during early morning hours (6:00 to 9:00 a.m. local standard
time). In Cleveland, OH, SO2 tended to peak later in the morning or in some cases early
to mid-afternoon.
The timing and duration of daytime SO2 peaks vary across the focus areas. These
variations were likely a result of a combination of source emissions and meteorological
parameters. The 2008 SOx ISA (U.S. EPA. 2008d) concluded that higher daytime SO2
2-63
-------
likely reflected an increase in power plant emissions coupled with an increase in
entrainment of these elevated emissions into the lower atmosphere as the mixed layer
expands throughout the day. Distinct morning peaks may have been related to stable
atmospheric conditions, which tend to trap atmospheric pollution near the ground,
resulting in an overall increase in ground-level pollution.
Notably, SO2 concentrations were all well below the primary NAAQS level during all
hours of the day in every focus area except Gila County, AZ. In the five urban focus
areas, mean 5-minute hourly max and 1-hour avg concentrations were almost all less than
5 ppb. All measured SO2 concentrations were for the most part below 15 ppb for all but
Gila County, AZ, even when examining the upper end of the distribution of 5-minute
hourly max concentrations. For Gila County, AZ, the upper bound of the 95% confidence
interval for 5-minute hourly max and 1-hour avg SO2 concentrations exceeded 100 ppb
and 50 ppb, respectively. At this location the large copper smelter sources nearby
contributed to a strong morning peak in the diel pattern.
2-64
-------
Cleveland
Is
e
5
O
888888888888888888888888
£ °
c
.2
8
s
3
Pilts burgh
888888888888888888888888
C
5
O
New York
888888888888SSS8S8888S8S
8 *
C
St. Louis
*««*
888888888888888888888888
Houston
e
~
o
888888888888888888888888
Gila County
888888888888888888888888
Note: For every hour, arithmetic mean concentrations are displayed as red circles. The whiskers represent the 95% confidence
interval. Hours are shown in local standard time.
Figure 2-23 Diel variability based on 1-hour avg sulfur dioxide concentrations
in the six focus areas, 2013-2015.
2-65
-------
Cleveland
Pilts burgh
% °
C
J?
TO
IS
e
5
O
888888888888888888888888
8
e
3
888888888888888888888888
New York
C
5
O
888888888888888888888888
a S
8
1 2
o
St. Louis
888888888888888888888888
e
~
o
Houston
Gila County
888888888888888888888888
888888888888888888888888
Note: For every hour, arithmetic mean concentrations are displayed as red circles. The whiskers represent the 95% confidence
interval. Hours are shown in local standard time.
Figure 2-24 Diel trend based on 5-minute hourly max data in the six focus
areas, 2013-2015.
2-66
-------
Diel SO2 concentration patterns may be influenced by seasonal factors. Diel plots of
5-minute hourly max for winter and summer are presented for Cleveland, OH and Gila
County, AZ in Figure 2-25. Concentration patterns are clearly different between the two
locations. While the peak of the mode increased in the summer months indicating higher
concentrations within the diel pattern, Cleveland, OH exhibited very little change in the
location of the mode for the diel patterns between the cold and warm seasons. In contrast,
the mode of the diel pattern occurred earlier in summer compared with winter for Gila
County, AZ. Factors that may influence the mode of the diel pattern include peak smelter
operation times and atmospheric mixing. For example, seasonal differences in solar
radiation prolong nighttime inversion periods during the winter. Transport to downwind
monitoring sites may be impeded by stable conditions. Moreover, increased solar
radiation during the summer enhances mixing, increasing the probability of plume
touchdown (Sladc. 1968b). The average 5-minute hourly max SO2 concentrations were
also somewhat lower during the summer compared with winter in Gila County, AZ. O3
production in the summer may have promoted oxidation of SO2 (Khodcr. 2002) to
produce the observed losses.
2-67
-------
s o
c <*>
9
re-
C
a
u
O
Cleveland Winter
888388888888888888888888
= H
¦B °
Gila County Winter
888888888888888888888388
n
a.
¦S o
c o
,2
c
r.
O
Cleveland Summer
O ...
• o+no?"
Q^0rfrt*rTirr5'u2>t-^C00iC?'— ao-OS O -t- CM-rt
*-C^t
-------
2.5.4
Relationships between Hourly Mean and Peak Concentrations
Peak concentrations within an SO2 plume can greatly exceed the mean concentration at
the plume centerline, so that exposure to the peak may be much greater than an hourly or
daily SO2 measurement. Plume dispersion is a Gaussian process, but the plume meanders
so that the peak at any instant in time exceeds the mean of the plume centerline found by
averaging over some longer time period, such as 1 hour or 1 day (Slade. 1968a; Gifford.
I960). Several studies (Dourado et al.. 2012; Schauberger et al.. 2012; Venkatram. 2002;
Turner. 1970) have characterized the peak-to-mean ratio (PMR), showing that the ratio
increases with longer concentration averaging time (where average concentration is found
in the denominator) compared with the time over which the peak concentration is
measured (where peak concentration is found in the numerator). Venkatram (2002) used
dispersion modeling to illustrate the stochasticity of the dispersion process, where the
mean over a longer time period is determined by an ensemble average across simulations.
At a fixed location, the results of Venkatram (2002) imply that exposure to the plume
peak occurs with varying probabilities based on the time scale used to represent the
instantaneous plume, the time scale over which the average is computed, the
intermittency of atmospheric turbulence, and atmospheric stability, where higher
atmospheric turbulence would lead to lower PMR.
The PMR has been computed in the literature as a function of the ratio of the
mean-to-peak concentration integration times raised to some power in the range of 0.2
(Venkatram. 2002; Turner. 1970) to 0.5 (Venkatram. 2002) or 0 to 0.68 (Schauberger et
al.. 2012). with the increasing exponent corresponding to increased atmospheric
instability. When 5-minute hourly max data are compared with 1-hour avg data, the
mean-to-peak integration time ratio is 60 minutes-to-5 minutes = 12. This inverted
relationship implies that a larger averaging time generally produces a larger PMR. A
peak-to-mean ratio of 1 to 5.4 would be expected using the wider range of exponents
(i.e., 12° to 12°6S). PMR varies over space and time due to differences in distance from
sources, source characteristics (e.g., stack height), wind speed, and changes in
atmospheric stability during the day.
Scatterplots of collocated 5-minute hourly max and 1-hour avg measurements are
displayed for all monitors in Figure 2-26 and by focus area in Figure 2-27. Data for the
PMR analyses were subject to the same completeness criteria outlined in Table 2-5
(Section 2.5.1).
PMRs were used in the previous SO2 NAAQS review to evaluate the distribution of
5-minute hourly max concentrations corresponding to a given 1-hour avg SO2
concentration (U.S. EPA. 2009c). PMRs are determined by dividing the 5-minute hourly
max concentration by the 1-hour avg concentration. Using this approach, a PMR of 1
2-69
-------
demonstrates that 5-minute hourly max and 1-hour avg concentrations are equivalent. A
high PMR value (up to a maximum value of 12 in this case) indicates that the 5-minute
hourly max concentration is higher than the 1-hour avg concentration. For example, a
PMR of 2 indicates that 5-minute hourly max concentration is 2 times higher than the
1-hour avg concentration. PMR values of 1 and 5.4, the upper value from the literature
(Schaubcrgcr et al.. 2012). are displayed as lines in Figures 2-26 and 2-27. Median PMRs
obtained from comparing the 5-minute hourly max with the 1-hour avg AQS data at sites
where both measures were available simultaneously had a range of 1 to 5.5 with a median
of 1.3 (99th percentile = 4.7), in reasonable agreement with the predicted range of 1 to
5.4 for the PMR. Concentrations at the 99th percentile 1-hour daily max of 63.5 ppb
correspond to a 5-minute hourly max of 200 ppb with a PMR of 3.2.
0 200 400 600 800
1-hr Average S02 Concentration (ppb)
S02 = sulfur dioxide.
Note: Peak-to-mean ratios are displayed on each scatter plot as 1:1 (5-min hourly max = 1-h avg) and 5.4:1 (5-min hourly max is
5.4 times higher than 1-h avg).
Figure 2-26 Scatterplot of 5-minute hourly max versus 1-hour avg sulfur
dioxide concentrations, 2013-2015.
2-70
-------
Xs o
a. oo
G,
c
o
O
O
to
-£= <=>
~ i"vj
Cleveland
Pittsburgh
200 400 600
1-hr Average SC2 Concentration (ppb)
200 400 600
1-hr Average S02 Concentration (ppb)
c
o
O
O
to
2
2* o
£
(-<
New York
200 400 600
1-hr Average S02 Concentration (ppb)
-C o
a. «s
Q,
e
o
o
o
CO
5.4:1/
/
/ •
S
*• / . v.y/
1 " /
•?! „
¦k:y
.JLTfrx *• /
mf St. Louis
200 400 600
1-hr Average S02 Concentration (ppb)
800
X5 O
a. oo
Q-
C
O
O
O
to
2* o
~ i*SJ
Houston
Q.
£=
O
O
O
to
£
i-.
200 400 600 800
¦hr Average S02 Concentration (ppb)
Gila County
200 400 600 800
1-hr Average S02 Concentration (ppb)
S02 = sulfur dioxide.
Note: Peak-to-mean ratios are displayed on each scatter plot as 1:1 (5-min hourly max = 1-h avg) and 5.4:1 (5-min hourly max is
5.4 times higher than 1 -h avg).
Figure 2-27 Scatterplot of 5-minute hourly max versus 1-hour avg sulfur
dioxide concentrations by focus area, 2013-2015.
2-71
-------
Table 2-15 displays the range of temporal correlations between corresponding 5-minute
hourly max and 1-hour avg concentrations and the range of PMRs computed from SO2
measurements reported at these monitoring sites within the six focus areas shown in
Figure 2-27. Similar to results in the 2008 SOx ISA (U.S. EPA. 2008d'). 5-minute hourly
max concentrations tend to correlate well with 1-hour avg metrics, suggesting that 1-hour
avg metrics, in most cases, adequately represent changes in 5-minute hourly max data
over time. However, 5-minute hourly max concentrations tend to be higher than 1-hour
avg concentrations. PMRs were skewed higher for the Gila County focus area and
slightly higher for the New York City focus area. However, overall 1-hour daily max
concentrations in New York were relatively low (highest 99th percentile 1-hour daily
max was 16.5 ppb), so a PMR of 2 or 3 would lead to a 5-minute hourly max of 33 or
49.5 ppb. In contrast, the 1-hour daily max concentrations in Gila County were much
higher (highest 99th percentile 1-hour daily max was 247 ppb), which would lead to
5-minute hourly max concentrations of 494 ppb if the PMR were 2 and of 741 ppb if the
PMR were 3.
Table 2-15 Pearson correlation coefficient comparing 1-hour avg with 5-minute
hourly max and peak-to-mean ratio for maximum sulfur dioxide
concentrations in the six focus areas, 2013-2015.
Focus Area
N Monitoring Sites
Correlation Coefficient
Median PMRa
Cleveland, OH
7
0.89-0.93
1.00-1.85
Pittsburgh, PA
9
0.91-0.97
1.00-1.40
New York City, NY
12
0.66-0.98
1.28-2.33
St Louis, MO
7
0.88-0.94
1.17-1.38
Houston, TX
9
0.91-0.95
1.33-1.69
Gila County, AZ
4
0.84-0.93
3.24-6.15
N = population number; PMR = peak-to-mean ratio.
aMedian PMR = 5 min max/1-h avg. The range of data represents median PMR across each site within the focus area.
2-72
-------
2.5.5
Background Concentrations
With the exception of periodic volcanic eruptions in Hawaii, natural and international
transboundary sources of SO2 make only minor contributions to the total atmospheric
burden of SO2 in the U.S. Sections 2.2.4 and 2.2.5 describe those sources contributing to
background SO2.
No new studies have appeared that attempt to estimate background SO2 concentrations
since the 2008 SOx ISA (U.S. EPA. 2008d'). The 2008 SOx ISA discussed a global scale
three-dimensional modeling study that estimated annual mean SO2 concentrations in
surface air including both anthropogenic and natural sources, using the MOZART-2
(Model of Ozone and Related Chemical Tracers) (Horowitz et al.. 2003). Sources
included in the study were emissions from fossil and biofuel combustion, biomass
burning, biogenic and soil emissions, and oceanic emissions. Background SO2
concentration estimates were below 0.01 ppb over much of the U.S. Aside from a few
areas influenced by near-border sources, maximum background concentrations of SO2 are
0.03 ppb. In the U.S. Northwest, geothermal sources of SO2 are responsible for 70 to 80%
of the background SO2 concentration; even so, total SO2 concentrations are still on the
order of ~2 ppb or less. In these simulations, background contributed less than 1% to SO2
concentrations in surface air in 2001 throughout much of the contiguous U.S.
Satellite-borne instruments have mapped large SO2 sources globally and have obtained
data showing intercontinental transport. Fioletov et al. (2013) identified a number of
"hotspots" for continuous SO2 emissions, both anthropogenic and volcanic
(e.g., industrial sources in China, Russia, the U.S., the Gulf of Mexico, and Saudi Arabia;
volcanic sources in Kilauea, HI and Anahatan in the Marianas). Clarisse et al. (2011)
showed evidence for transport of SO2 from Asia to Alaska and Canada. In one such
episode in November 2010, there was a clearly defined plume crossing the Pacific.
As described in Section 2.2.4.2. volcanic sources of SO2 in the U.S. are found in the
Pacific Northwest, Alaska, and Hawaii. The most important domestic effects from
volcanic SO2 occur on the Hawaiian Islands. Nearly continuous venting of SO2 from
Mauna Loa and Kilauea produces SO2 in high concentrations that can affect populated
areas on the Big Island of Hawaii (as well as others in the chain, depending on wind
conditions). Figure 2-28A shows the 2008-2013 time series for 1-hour daily max SO2
concentrations at Hilo, HI, (population of approximately 40,000), which is located about
50 km northeast of Kilauea. Figure 2-28B shows the same time series at Pahala
(population -1,300), which is located about 30 km southeast of Kilauea (Longo et al..
2010). As demonstrated by these figures, 1-hour daily max SO2 concentrations can reach
levels greater than 1,000 ppb. Figure 2-29 shows a 6-month concentration time series for
2-73
-------
the Ka' u District, one of the other communities scattered throughout the southern half of
the island that are also exposed to high SO2 concentrations (Longo et al.. 2010).
1800
1600
Daily MAX 1 -hr S02 at Hilo, HI
— 1400
>
1 1200
I 1000
| 800
§ 600
o
<5" 400
I
200 I
> IlliiJi I.J..I .lit ill
1/1/2008 1/1/2009 1/1/2010 1/1/2011 1/1/2012 1/1/2013
Date
B
Daily MAX 1-h S02 at Pahala, HI
5 1400
1 1000
O 600
S02 = sulfur dioxide.
Figure 2-28
1-hour daily max sulfur dioxide concentrations measured at
(A) Hilo, HI and (B) Pahala, HI.
2-74
-------
300
250
§
c
o
= 200
ID
Q,
m
cl
c
o
o5
£Z
(D
O
c
O
CO
150
100
50
Low Exposure Period
High Exposure Period
~
~
c ~
B. ~
2 ~ ~
LU
E
1 *
C/3 ~
~ ~ ~
~ ~ ~
•
~
~
~ ~
%~ ~~
~
~ ~ ~ *
; • ~ ~ %4
~ % * * ~
* * ~ ~ . A ~ ~, ~'
~~ ~
~ « * «~~ ~ •
* — ~ ~
* * *~~~ *
h- h- h-
O O O
<3 co
CM
o
h- h-
O O
O N ^f
CM 2" 31 £M Q
CM CM CM CM
COCOCOCOCOCOQOCOOOoOQOCOQOCOCOQOQOCOOO CO CO CO
0O0000O000O0O0OOO00 ooo
s
CM CM
5
CM
c^ cn3
CM
CNj
CO
CO
CM CO
c?5 co
^ 5
CM CN
? $
LO
CM
l?5 i?5 i?5
CD
S02 = sulfur dioxide.
Note. The dashed line represents the World Health Organization 24-h avg S02 guideline = 7.5 ppbv (WHO. 2006V
Data source: S02 measured continuously by a TECO pulsed-fluoresce nee monitor, State of Hawaii Air Quality Division.
Source Lonao et al. (2010). Reprinted with permission of Taylor and Francis.
Figure 2-29 Average 24-hour ambient sulfur dioxide concentrations during
low and high (volcanic gas) concentration study periods
(November 26, 2007 to June 6, 200B) for Ka'u District, located
downwind of KTIauea Volcano.
2.6 Atmospheric Modeling
This section discusses various modeling techniques to estimate ambient concentrations of
SO2. Different types of models are discussed in terms of their capabilities, strengths, and
limitations. The section focuses on recent models that have been widely used in U.S.
applications. Section 2.6.1 focuses on dispersion models, which are the most widely used
and the most relevant for modeling the influence of large point sources on local-scale SO2
concentrations in the urban and other near-field environments. Section 2.6.2 briefly
2-75
-------
discusses chemical transport models (CTMs) that can be used to model SO2
concentrations at regional and national scales.
2.6.1 Dispersion Modeling
Atmospheric transport and dispersion (ATD) models are important mathematical tools for
simulating the fate of air pollutants in support of a wide variety of environmental
assessments. ATD models can be used to estimate SO2 concentration for regulatory
purposes if monitoring data are not available or sufficient (75 CFR 35520). Using
equations that represent the physical and chemical atmospheric processes that govern
dispersal and fate, ATD models provide an estimate of the concentration distribution,
both temporally and spatially, of pollutants emitted from sources such as industrial
facilities, roadways, and urban areas. The processes that are most important vary
depending on the particular model application. The models must specifically account for
the characteristics of the source or sources of the pollutant (e.g., buoyant releases), the
meteorological conditions, the surrounding surfaces and complexities (e.g., buildings,
terrain, and trees), the background concentrations from sources not considered directly in
the modeling and the chemical transformations of the pollutant in the atmosphere. The
use of ATD models in health studies for SO2 is described in Section 3.3.2.4.
Dispersion models are particularly important to pollutant studies where monitoring is not
practical or sufficient. For pollutants such as SO2 where spatial distributions of 1-hour
avg concentrations associated with large sources often contain extreme gradients, the
siting of individual monitors to capture high ground-level concentrations over a wide
variety of sources and meteorological conditions would be challenging at best. Extensive
arrays of monitors are impractical. Thus, the implementation program for the 2010
primary SO2 NAAQS allows for air quality modeling to be used in place of monitoring to
characterize air quality, and for such air quality information to be used in the process for
informing final designation decisions (75 FR 35520). The SO2 NAAQS is currently the
only criteria pollutant standard for which modeling may be used to characterize air
quality for the purpose of the area designation process. In addition, modeling is critical to
the assessment of the impact of future sources or proposed modifications where
monitoring cannot inform. Also, modeling is helpful in the design and implementation of
mitigation techniques for addressing existing pollution problems and for compliance
evaluations.
ATD models take many forms. They include steady-state (emissions and meteorology),
Gaussian-based formulations [e.g., AERMOD, (Cimorelli et al.. 2005)1; Lagrangian
models [e.g., SCIPUFF, (Svkcs ct al.. 2007): HYSPLIT, (Draxlcr. 1999): (NOAA.
2-76
-------
2014)1. which are particularly useful when emissions and meteorological conditions are
variable over the modeling increment, and Eulerian photochemical grid-based models
[e.g., Community Multiscale Air Quality (CMAQ), (Bvun and Schere. 2006)1. which
explicitly model chemical processes and have modeling resolution ranges from about one
to tens of kilometers. Additionally, there are stochastic or statistical approaches using, for
example, Monte Carlo techniques (Hanna ct al.. 1982) or those using simple regression
approaches (Baneriee etal.. 2011). For very complex flows such as a release within an
urban canopy of a city, computational fluid dynamics models are considered. Hanna et al.
(2006) demonstrated that these models are capable of reproducing the general flow and
measured tracer dispersion patterns when very detailed source and three-dimensional
building information are available.
In the U.S., steady-state Gaussian models are the most common dispersion models used
for primary pollutants like SO2 (U.S. EPA. 2010a). These models may be used to
determine compliance with standards and primary pollutant impacts from new or
proposed sources. The same is true for these types of analyses in other countries. For
example, the Advanced Dispersion Modeling System (ADMS) (Carruthcrs et al.. 1995).
the Hybrid Plume Dispersion Model (HPDM) (Hanna and Chang. 1993). the Danish
model, Operationelle Meteorologiske Luftkvalitetsmodeller (OML) (Olesen et al.. 1992).
and several other steady-state Gaussian-based models have been recommended by the
European Environment Agency (van Aalst et al.. 1998) for applications involving SO2
from smoke stacks. Other examples in which Gaussian-type models are found to be
applicable for near-field applications are by the U.K. Department of Environment, Food,
and Rural Affairs (Williams et al.. 2011) and by the New Zealand Ministry of the
Environment (Bluett et al.. 2004). The primary concerns for many of these
compliance-type applications are the magnitude, location, and frequency of high
concentrations and the strong gradients of concentrations found near sources. Often the
highest concentrations are found within a few kilometers and sometimes within tens of
meters of the source. Near-field or near-to-the-source dispersion is the real strength of
steady-state modeling.
The American Meteorological Society (AMS)/EPA Regulatory Model (AERMOD) is the
preferred model of U.S. EPA for the vast majority of near-field applications with the Off-
shore and Coastal Dispersion model (OCD) being used for offshore emissions and
alternative models used for unique situations [e.g. the California Puff model (CALPUFF)
for Class I area screening application) where justified. AERMOD represents a
modernization of applied Gaussian models with advances in areas such as: boundary
layer scaling formulations; dispersion rates for both surface and elevated releases; plume
interactions with buildings and complex terrain; and characteristics of point, area, and
volume source types. In convective conditions, where dispersion produces a distinctly
2-77
-------
non-Gaussian vertical pollutant distribution, AERMOD provides a three-part formulation
(each Gaussian) that when combined yield distributions representative of those observed
(Weil et al.. 1997; Briggs. 1993). The challenges faced by Gaussian models in very light
wind conditions are addressed in AERMOD by simulating a meandering plume, and
providing turbulence-based lower limits on the transport wind speed and an empirically
based correction for the surface friction velocity. In recent years, U.S. EPA has been
working to improve AERMOD predictions under light wind conditions, including an
adjustment of surface friction velocity under stable light wind conditions (80 FR 45340).
For modeling applications where light and variable winds are dominant and reliable wind
field estimates are available, models such as SCIPUFF or HYSPLIT provide estimates of
plume trajectories and more temporally resolved concentration distributions
[e.g., Wannberg et al. (2010)1.
AERMOD and models like it are designed to simulate concentrations on an hourly
increment, and model evaluations are focused on averaging times of 1 hour or greater
(Perry et al.. 2005). although any averaging times may be used, including 5-minute
averages, provided that dispersion parameters used in the model have been estimated
from available data for the selected averaging time (Pasquill and Smith. 1983V Longer
term concentrations are obtained by averaging the 1-hour concentrations. Spatial
resolution is simply determined by the density of receptors included in the analysis
(i.e., very high resolution possible). For each hour, emissions and other source
characteristics, land surface characteristics, and meteorological conditions are provided to
the model. Additionally, the model requires a description of buildings and complex
terrain within the modeling domain that are expected to influence pollutant dispersion.
The model can simulate hundreds of sources and receptors, providing for analyses in
urbanized and industrialized areas.
One limitation of the Gaussian approach is the assumption of steady conditions over a
1-hour modeling period and over the plume transport distance to the receptors. The model
is recommended for receptors up to 50 km from a source when steady conditions are
appropriate (U.S. EPA. 2005b'). However, this can be challenging, especially for light
winds. Under low wind conditions, there are concerns that AERMOD can overestimate
measured SO2 concentrations without adjustment for empirical relationships between
wind and concentration (Paine et al.. 2015). Recent updates to AERMOD have been
made by the U.S. EPA to address those concerns (80 FR 45340). AERMOD is also
limited in its treatment of SO2 chemistry, using a method much simpler than the more
rigorous simulation of atmospheric transformation of SO2 found in models such as
CMAQ or SCICHEM (Chowdhury et al.. 2012). AERMOD uses a simple 4-hour half-life
assumption for reducing SO2 concentration in the plume with travel time (Turner. 1964).
This approach yields results consistent with the SO2 residence time estimates by Hidv
2-78
-------
(1994) and Seinfeld and Pandis (2006). Therefore, for conditions and sources where the
highest hourly concentrations are expected to be relatively close to the source, chemistry
is not expected to play a major role in determining compliance with primary standards.
Lagrangian puff dispersion models, such as CALPUFF, have been developed as an
alternative to Gaussian dispersion models, such as AERMOD. CALPUFF models SO2 as
particles and then uses a Lagrangian step algorithm to model nonsteady-state dynamics,
using time-varying winds specified by meteorological models, such as MM5 [e.g., Atabi
et al. (2016). Abdul-Wahab et al. (2011). Souto et al. (2014). Lee et al. (2014). Zhang et
al. (2015a)l. The nonsteady-state approach offered by Lagrangian puff dispersion models
may be considered an alternative to Gaussian dispersion models that do not account for
time dependence. Comparisons have been conducted between Lagrangian models such as
CALPFUFF and Gaussian plume models such as AERMOD. CALPUFF predictions of
24-hour SO2 concentrations at an oil refinery in Sohar, Oman compared within 36% of
measurements (Abdul-Wahab et al.. 2011). Comparison of CALPUFF and AERMOD to
SO2 measurements at a gas refinery in South Pars, Qatar showed that, while CALPUFF
and AERMOD both typically underestimated SO2 measurements, CALPUFF predictions
were usually closer to measured SO2 concentrations compared with AERMOD (Atabi et
al.. 2016). However, Rood (2014) observed that Lagrangian puff models and Gaussian
dispersion models both underpredicted 1-hour and 9-hour avg concentrations, but the
magnitude of bias was larger in the Lagrangian puff models applied at a field site in
Colorado with variable winds and natural topography. Holnicki et al. (2016) noted that
the model performance improved with longer averaging times and that the 1-hour avg
concentration predicted by CALPUFF was less accurate than predictions for annual
average concentrations, when compared to SO2 measurements. However, recent
dispersion modeling results were compared between CALPUFF and AERMOD for the
Section 126 Petition from New Jersey for the Portland Generating Station (76 FR 69052)
where CALPUFF overestimated 1-hour daily max SO2 observations taken in Columbia,
NJ by 226%, while AERMOD overestimated the same observations by 14%.
Uncertainty in the model predictions is influenced by the uncertainty in model input data
(in particular emission or source characterization and meteorological conditions) as well
as by inadequacies in model formulations. Uncertainty related to model input variables is
generally estimated by propagating the expected errors in the individual input variables
(e.g., wind speed, emission rate) through the model using Monte Carlo techniques
(Dabberdt and Miller. 2000). In addition, there is uncertainty related to the fundamental
difference between modeled and measured concentrations. Monitored data (within
sampling error) represents actual realizations of events, while modeling estimates
represent ensemble mean concentrations (Rao. 2005). Based on a study comparing a
variety of models (including Gaussian) to a number of tracer field study results, Hanna et
2-79
-------
al. (1993) found that for continuous point releases and receptors within a kilometer of the
source, uncertainty in model inputs in combination with the stochastic nature of the
atmosphere result in typical mean biases on the order of 20 to 40% and normalized mean
square errors up to 70%. The author points out that these levels of difference between
model and monitor results would likely exist even for more sophisticated models. Hanna
(2007) provided a comprehensive review of methods for determining sensitivity and
uncertainty in ATD models.
Focusing on the uncertainties in model inputs, it is easy to see that an individual model
estimate paired in time and space with a monitored concentration will likely differ,
sometimes substantially. Weil (1992) pointed out that wind direction uncertainties alone
can cause disappointing results in space and time pairings from otherwise
well-performing dispersion models. With wind direction errors, the plume footprints from
the model and that from the observations may not overlap. However, a model that is
based on appropriate characterizations of the important physical processes should be able
to reproduce the distribution of observed concentrations assuming that the distributions of
model inputs is similar to that of the observed conditions (Venkatram et al.. 2001).
Meteorological inputs coupled with AERMOD can impact the results, and the output may
depend on the use of recorded meteorological observations or meteorological models
(e.g., Weather Research and Forecasting (WRF) model). Meteorological models may add
error to the dispersion simulation, and that error is impacted by model selection and
resolution (Isakov et al.. 2007). Therefore, in evaluating a model's ability to predict
concentrations within the modeling domain, it is important to include an analysis of
modeled and monitored concentration distributions for any location studied, and possibly
to adjust the results based on any observed bias. As part of the proposed update to the
Guideline on Air Quality Models, U.S. EPA proposed to allow the use of prognostic
meteorological data for regulatory applications of AERMOD (80 FR 45340). U.S. EPA
conducted several assessments comparing observed meteorological data to prognostic
meteorological data and found that the prognostic data performed adequately (U.S. EPA.
2015a).
Chang and Hanna (2004) provided a comprehensive discussion of methods for evaluating
the performance of air quality models. They discuss a series of performance measures
that included statistical metrics such as fractional bias (FB), geometric mean bias,
normalized mean squared error and the fraction of estimates within a factor of two
observations. These and other measures are included in the commonly used BOOT
software (Chang and Hanna. 2005). which also allows for estimation of confidence limits
on the concentrations computed and provides insight about the sources of bias in the
model (Irwin. 2014). Chang and Hanna (2004) also discussed exploratory analysis
methods of plotting and analyzing the modeled and measured concentrations. They
2-80
-------
pointed out that the most useful model evaluation studies are those that examine a
number of models and compare them with a number of field studies.
For models intended for application to compliance assessments (e.g., related to the 1-hour
daily max SO2 standard), the model's ability to capture the high end of the concentration
distribution is important. Measures such as robust highest concentration (RHC) (Cox and
Tikvart. 1990). and exploratory examinations of quantile-quantile plots (Chambers et al..
1983) are useful. The RHC represents a smoothed estimate of the top values in the
distribution of hourly concentrations. In contrast, for dispersion modeling in support of
health studies where the model must capture concentrations at specified locations and
time periods, additional measures of bias and scatter are important.
The intended use of a model and the objective of a model evaluation guide the selection
of evaluation criteria. Frost (2014) evaluated the model performance for AERMOD when
it was applied to the study of 1 year of SO2 emissions from three coal-fired EGUs. The
study authors found agreement within a factor of two of the 99th percentile SO2 design
value for the majority of the data but noted performance outside a factor of two for the
top 5% of measured 1-hour avg concentrations. However, Rehbein et al. (2014) found
that the model fell within a factor of two of the monitoring data even at high
concentrations for a model validation outside a nickel smelting facility in Sudbury,
Ontario, Canada. U.S. EPA also conducted evaluations of prognostic meteorological data
in AERMOD (U.S. EPA. 2015a). including the facility modeled by Frost (2014). These
evaluations included data analysis adhering to the U.S. EPA Protocol for Best Performing
Models, which includes a scientific and operational component of model performance
(U.S. EPA. 1992). SO2 concentrations modeled by AERMOD were within a factor of two
observations in all but one simulation when using the metrics of the protocol.
Meteorological parameters were modeled with FB within 20% of observations (U.S.
EPA. 2015a).
At the time of its inclusion into the U.S. EPA Guideline on Air Quality Models (U.S.
EPA. 2005b). the performance of AERMOD was evaluated against seventeen field-study
databases over averaging times from 1 hour to 1 year (Perry et al.. 2005). In each case,
the emissions characteristics and background concentrations were well known,
meteorological data were available on site, and tracer concentrations were measured at
multiple locations where high plume impacts were expected. Four of the studies involved
very dense sampler arrays. For the four intensive studies, Perry et al. (2005) found the
ratio of modeled 1-hour avg RHC to monitored RHC ranged from 0.77 to 1.18
[i.e., relatively unbiased in estimating extreme (high) values]. For studies involving tall
buoyant stacks with more limited monitoring locations, 1-hour ratios were not reported,
but the 3-hour avg ratios ranged from 1.0 to 1.35 (i.e., a slight tendency to overpredict the
2-81
-------
high concentrations). Examination of quantile-quantile plots supported the findings that
the model was capturing the upper end of the 1- and 3-hour avg concentration
distribution. Hannaetal. (2001) evaluated the AERMOD and ADMS Gaussian
dispersion models with five field study databases including area sources, low releases,
and tall power plant stacks in rural, flat, and complex terrain. Among the median
performance measures they reported, the ratio of maximum modeled to maximum
observed concentrations was 0.77 for AERMOD and 0.80 for ADMS, each a small
underprediction. The median value over the five databases of the geometric mean (MG, a
measure of the ratio of averaged modeled to monitored concentration) was 1.7 for
AERMOD and 1.22 for ADMS. With 1.0 as the ideal value, both models were found to
overpredict (with ADMS less biased). Unlike the ratio of maximum values, MG is a
measure of performance over the entire distribution of concentrations. Hurley (2006) also
evaluated AERMOD and two Australian models against seven field studies and found no
database against which AERMOD performed poorly.
With the adoption of the 2010 1-hour daily max SO2 standard, there is renewed interest in
AERMOD's abilities to simulate near-field maximum short-term concentrations. A
number of specific areas for model improvement were discussed at the 10th and 11th
Modeling Conference on Air Quality in 2012 (U.S. EPA. 2012a) and 2015 (U.S. EPA.
2016a). Among them were concerns about simulations in stable conditions with light and
meandering winds, use of prognostic meteorological data, modeling of emissions from
haul roads, plume chemistry, and building downwash. Proposed improvements include an
adjusted friction velocity model for stable/low wind conditions in AERMET, a new
model for dispersion options in AERMOD, and an option for buoyant line sources in
AERMOD (U.S. EPA. 2016a'). Routine reporting of 5-minute average concentrations by
air agencies may also facilitate model evaluation. Research in many of these areas is
underway, and improvements to AERMOD have been made based on the outcomes of
those conferences, largely as part of U.S. EPA rulemaking to revise the Guideline. While
the stochastic nature of the atmosphere will always preclude the development of a perfect
model, improvements to the model formulations will continue with the goal of estimating
hourly average concentrations while reducing model uncertainty and expanding
applicability.
2.6.2 Chemical Transport Models
Chemical transport models are an important tool for characterizing regional- and
national-scale air quality. The scales at which they typically operate are too large to
satisfactorily capture meteorological and chemical processes involving SO2 at the local or
2-82
-------
near-source scale. The dispersion models discussed previously are thus preferable for
characterizing SO2 concentrations at these scales.
Chemical transport models such as the CMAQ model, are deterministic models of
chemical transport that account for physical and chemical processes, including advection,
turbulence, diffusion, deposition, gas-phase and heterogeneous chemistry, and convective
cloud transport, while following the constraint of mass conservation (Bvun and Schcrc.
2006). CTMs provide regional concentration estimates and are typically run with
horizontal grid resolutions of 4, 12, or 36 km. Temporal resolutions are typically 1 hour,
although larger temporal aggregation often occurs for the purpose of maintaining
reasonable data file size. CTMs are used to compute interactions among primary
atmospheric pollutants and their transformation products, the production of secondary
aerosols, the evolution of particle size distribution, and transport and deposition of
pollutants. CTMs are driven by emissions inventories for primary species such as SO2,
NO2, NH3, VOCs, and primary PM, and by meteorological fields produced by other
numerical weather prediction models. Values for meteorological variables such as winds
and temperatures are taken from a meteorological model that is nudged by operational
analyses, reanalyses, or general circulation models. In most cases, these are off-line
meteorological predictions; thus, they are not modified by radiatively active species
generated by the air quality model. Work to integrate meteorology and chemistry was
initiated in the mid-1990s [by Lu et al. (1997a) and Lu et al. (1997b) and references
therein], although limits to computing power prevented widespread application. More
recently, new integrated models of meteorology and chemistry are available; see, for
example, the Weather Research and Forecast model with chemistry (WRF-Chem;
http://ruc.noaa.gov/wrf/wrf-chem/') and WRF-CMAQ (Wong et al.. 2012).
Biases in SO2 concentrations predicted by CTMs can occur as a result of error in model
representation of atmospheric processes converting SO2 to H2SO4 and in removal
processes. For example, overestimates of cloud-based reactions converting SO2 to H2SO4
have been shown to negatively bias SO2 concentration estimates in CMAQ v4.6 (Mueller
et al.. 2011). Improvements to modeling these processes, such as capturing metal
catalysis of the SO2 H2SO4 conversion process, have been included in CMAQ v5.0.2
to improve model estimates of SO2 and SO42 (Alexander et al.. 2009). Therefore, when
using CMAQ to estimate exposure to SO2, attention must be given to the version of the
model so that any inherent biases are understood.
The Air Quality Model Evaluation International Initiative (AQMEII) was developed by
scientists in Europe and North America to evaluate several CTMs against each other
using common input data sets (Rao etal.. 2011). Pouliot et al. (2015) assembled
emissions input data for European and North American simulations performed over two
2-83
-------
phases of the AQMEII study and found a 12% reduction in SO2 emission estimates for
2006 in both Europe and North America. These differences were attributed to differences
in methodologies used to estimate emissions and to differences in input data that
influence the CTM output. In a comparison of CTM models of SO2 with surface
measurements in Europe, the Modeling Atmospheric Composition and Climate (MACC)
model reanalysis overestimated surface SO2 concentrations by 40% in winter and
underestimated surface SO2 levels by 63% in summer (Giordano et al.. 2015). In North
America, MACC underestimated SO2 in summer by 81%. MACC results were higher
than regional CTMs in the winter for North America, and seasonal variability was not
well captured (r = 0.16 in summer and r = 0.19 in winter). These errors were thought to
relate to the differences in the lifetime of SO2 transported from the domain borders to the
domain center being shorter than the time scale of the model bias.
2.7 Summary
Of the sulfur oxides, SO2 is the most abundant in the atmosphere, the most important in
atmospheric chemistry, and the one most clearly linked to human health effects
(Chapter 5). Thus, the NAAQS are currently set using SO2 as the indicator species. As a
consequence of several U.S. air quality regulatory programs, emissions of SO2 have
declined by approximately 79% for all NEI source categories during the time period 1990
to 2014 (Section 2.2). Coal-fired EGUs remain the dominant anthropogenic source,
emitting 3.2 x 106 tons SO2 annually according to the 2014 NEI. Natural sources include
volcanoes, wildfires, and biogenic sulfides that are intermittent and of limited spatial
extent.
Beyond the size of the emissions source, the important variables that determine the
concentration of SO2 downwind of a source are the photochemical removal processes
occurring in the emissions plume (Section 2.3) and local meteorology. The gas-phase
oxidation of SO2 by hydroxyl radical is slow in comparison to aqueous-phase oxidation
in cloud and fog droplets. Clouds and fog can reduce local SO2 concentrations by
converting it to H2SO4 in the droplet phase. Another gas-phase oxidation mechanism
involves a Criegee intermediate biradical that participates in converting SO2 to SO3. The
Criegee-based SO2 oxidation mechanism may amplify the rate of SO2 removal in areas
with high concentrations of Criegee precursors (i.e., low molecular weight organic gases,
such as biogenic compounds, and unsaturated hydrocarbons) present downwind of
industrial sites and refineries. The atmospheric SO2 oxidation processes, coupled with
variable meteorological conditions, including wind, atmospheric stability, humidity, and
cloud/fog cover, influence the observed SO2 concentrations at urban monitoring sites.
2-84
-------
Changes were undertaken to the existing U.S. EPA monitoring network as a result of the
new 1-hour daily max primary NAAQS standard promulgated in 2010 (Section 2.4).
First, the automated pulsed UVF method, the method most commonly used by state and
local monitoring agencies for NAAQS compliance, was designated as an FRM. Second,
new SO2 monitoring guidelines require states to report 5-minute data in light of health
effects evidence on lung function decrements among exercising individuals with asthma
following a 5-10 minute exposure of SO2 above 200 ppb (Section 5.2.1.2). There are
380 monitoring sites across the U.S. reporting 5-minute data. Analysis of environmental
concentrations of SO2 data reported in this chapter reflect the monitoring network
changes, particularly the analysis of the recent 5-minute data.
On a nationwide basis, the average 1-hour daily max SO2 concentration reported during
2013-2015 was 5.3 ppb (Section 2.5.2.1). The national 99th percentile 5-minute hourly
max concentration was 23.8 ppb, suggesting that the occurrence of very high 5-minute
peak values is rare. However, peak concentrations (99th percentile) of the 1-hour daily
max SO2 concentrations were greater than 75 ppb at some monitoring sites located near
large anthropogenic sources (e.g., power plants). Volcanoes produce large amounts of
SO2, and hourly concentrations in their vicinity can be greater than 2,000 ppb. SO2
concentration is highly variable across urban spatial scales (Section 2.5.2.2). exhibiting
moderate to poor correlations between SO2 concentrations measured at different
monitoring sites across a metropolitan area. This high degree of urban spatial variability
may not be fully captured by central site monitoring estimates.
Long-term concentration trends show a steady decline in the mean, 10th, and 90th
percentile of the site-specific 99th percentile of the 1-hour daily max SO2 concentrations
(Section 2.5.3). The data show a 76% decline in 99th percentile 1-hour daily max SO2
concentration over the period 1990-2015. Seasonal trends were examined for six focus
areas, and only New York and, to a lesser extent, Houston exhibited strong intra-annual
trend in which cool season 1-hour daily max SO2 concentrations were higher than warm
season 1-hour daily max SO2 concentrations. Diel patterns in 1-hour avg SO2
concentration mostly shows daytime concentrations peak in the morning or midday, and
the time of the peak can vary by location and may be influenced by seasonal conditions.
Peak concentrations within an SO2 plume can greatly exceed the mean concentration at
the plume centerline, so that exposure to the peak may greatly exceed an hourly or daily
SO2 measurement (Section 2.5.4). PMRs obtained from comparing the 5-minute hourly
max with the 1-hour avg AQS data at sites where both measures were available
simultaneously had a range of 1 to 5.5 with a median of 1.3. In a city with low SO2
concentrations, a high PMR may still be related to elevated 5-minute hourly max SO2
concentration. For example, overall 1-hour daily max concentrations in the New York
2-85
-------
focus area were relatively low (highest 99th percentile 1-hour daily max was 16.5 ppb),
so a PMR of 2 or 3 would lead to a 5-minute hourly max of 33 or 49.5 ppb. In contrast,
the 1-hour daily max concentrations in Gila County were much higher (highest 99th
percentile 1-hour daily max was 247 ppb), which would suggest 5-minute hourly max
concentrations of 504 ppb if the PMR were 2 and of 741 ppb if the PMR were 3.
Contributions to background concentrations include natural emissions of SO2 and
photochemical reactions involving reduced sulfur compounds of natural origin, as well as
the transport of sulfur compounds from outside of the U.S. (Section 2.5.5). In the U.S.
Northwest, geothermal sources of SO2 are responsible for 70 to 80% of the background
SO2 concentration; even so, total SO2 concentrations are still on the order of ~2 ppb or
less. In model simulations, background contributed less than l%to SO2 concentrations in
surface air in 2001 throughout much of the contiguous U.S. Even with ambient
concentrations for 2013-2015 that were roughly half the magnitude of those measured
around 2001, the estimated background SO2 would contribute only 2% to ambient SO2
concentrations in most of the contiguous U.S.
Atmospheric modeling includes dispersion and chemical transport models to estimate
SO2 concentrations in locations where monitoring is not practical or sufficient
(Section 2.6). Because existing ambient SO2 monitors may not be sited in locations to
capture peak 1-h daily max concentrations, the implementation program for the 2010
primary SO2 NAAQS allows for air quality modeling to be used to characterize air
quality for informing designation decisions (75 FR 35520). Modeling is critical to
assessing the impact of future sources or proposed modifications when monitoring cannot
be informative, and for designing and implementing mitigation techniques.
2-86
-------
CHAPTER 3 EXPOSURE TO AMBIENT SULFUR
DIOXIDE
3.1 Introduction
The 2008 SOx Integrated Science Assessment (ISA) (U.S. EPA. 2008d) evaluated
ambient sulfur dioxide (SO2) concentrations and exposure assessment in multiple
microenvironments, presented methods for estimating personal and population exposure
via monitoring and modeling, analyzed relationships between personal SO2 exposure and
ambient SO2 concentrations, and discussed the implications of using ambient SO2
concentrations to estimate exposure in epidemiologic studies. This chapter summarizes
that information and presents new information regarding exposure to ambient SO2. The
chapter will focus on the inhalation exposure route for SO2 from the key sources
described in Chapter 2 because the presence of other SOx species in the atmosphere has
not been demonstrated, as discussed previously. Exposure to particulate sulfate formed
by oxidation of SO2 is addressed in the particulate matter (PM) ISA (U.S. EPA. 2009a).
Sections within the chapter are organized to first present broad exposure concepts
applicable to air pollution in general, followed by S02-specific material. Topics
addressed in the chapter include methodological considerations for use of exposure data,
and exposure assessment and epidemiologic inference. Many new studies are included in
this chapter to better characterize exposure and understand exposure error. This material
provides context for interpreting the epidemiologic studies described in Chapter 5.
3.2 Conceptual Overview of Human Exposure
3.2.1 Exposure Terminology
A variety of metrics and terms are used to characterize air pollution exposure. They are
described here at the beginning of the chapter to provide clarity for the subsequent
discussion.
The concentration of an air pollutant is defined as the mass or volume of the pollutant in
a given volume of air (e.g., |ig/m3 or ppb). Concentrations observed in outdoor locations
are referred to as ambient concentrations. The term exposure refers to contact with a
specific pollutant concentration over a certain period of time (Zartarian et al.. 2005). in
single or multiple locations. For example, contact with a concentration of 10 ppb SO2 for
3-1
-------
1 hour would be referred to as a 1-hour exposure to 10 ppb SO2, and 10 ppb is referred to
as the exposure concentration. As discussed in Chapter 4. dose incorporates the concept
of intake into the body (via inhalation). Exposure concentrations are particularly relevant
for interpreting controlled human exposure studies, where participants are exposed to a
well-defined pollutant concentration, or panel epidemiologic studies that use personal
exposure monitors. Ambient concentrations are more relevant to epidemiologic studies
using measured or modeled concentrations.
A location where exposure occurs is referred to as a microenvironment, and an
individual's daily exposure consists of the time-integrated concentrations in each of the
microenvironments visited during the day. Ambient air pollution may penetrate indoors
(see Section 3.4.1.1 on infiltration), where it combines with air pollution from indoor
sources (nonambient air pollution) to produce the total measured indoor concentration.
Exposure to the ambient fraction of this indoor concentration, together with exposure to
ambient concentrations in outdoor microenvironments, is referred to as ambient exposure
(Wilson et al.. 2000).
Because personal exposures are not routinely measured, the term surrogate is used in this
chapter to describe a quantity meant to estimate or represent exposure, such as an SO2
concentration measured at a fixed-site monitor (Sarnat et al.. 2000). When surrogates are
used for exposure assignment in epidemiologic studies, exposure misclassification or
exposure error can result. Exposure misclassification refers to exposure error for
categorical variables, such as diseased and nondiseased individuals. Exposure
measurement error due to exposure assignment methods and spatial and temporal
variability in pollutant concentrations may be either differential or nondifferential.
Differential error occurs when the measured exposure and associated error differ across
groups such that the mismeasured exposure contains information about the health
outcome other than that associated with the true exposure (Armstrong. 2008). An
example of differential exposure error is the use of geocoding to estimate air pollution
exposure by proximity to roadways, because concentrations are different upwind and
downwind of a major roadway (Lane et al.. 2013; Singer et al.. 2004). Nondifferential
exposure error refers to the situation where exposure characterization is similarly
accurate across all groups (Armstrong et al.. 1992).
Exposure misclassification and exposure error can result in bias and reduced precision of
the effect estimate in epidemiologic studies. Bias refers to the difference between the
observed and true association, while precision is typically represented by the width of the
confidence interval around the effect estimate. Bias toward the null, or attenuation of the
effect estimate, indicates an underestimate of the magnitude of the effect and is
characteristic of nondifferential measurement error, particularly classical measurement
3-2
-------
error. Bias away from the null can occur through differential exposure measurement error
or under certain exposure scenarios (Armstrong et al.. 1992).
Exposure error refers to the bias and uncertainty associated with using measured or
estimated concentrations to represent the actual exposure of an individual or population
(Lipfert and Wvzga. 1996). Exposure error has two components: (1) exposure
measurement error derived from uncertainty in the quantity being used to represent
exposure, whether measured or estimated concentration or exposure, and (2) use of a
surrogate in the epidemiologic study in lieu of the true exposure, which may be
unobservable. Section 3.4.4 provides additional definitions for specific types of exposure
error and discusses the potential impact of such errors on epidemiologic study results.
3.2.2 Conceptual Model of Personal Exposure
A theoretical model of personal exposure is presented in this section to highlight
measurable quantities and uncertainties. This model has been developed and presented in
previous ISAs, most recently in the 2016 ISA for Oxides of Nitrogen (U.S. EPA. 2016d).
and it is reproduced here to provide context for the current document.
An individual's time-integrated total exposure to SO2 can be described based on a
compartmentalization of the person's activities throughout a given time period:
ET = j Ctdt
Equation 3-1
n
Et = X Cjtj
y=i
where Et = total exposure over a time period of interest, Ct = airborne SO2 concentration
during time t spent in a given microenvironment, and dt = portion of the time period
spent in a given microenvironment,/ Total exposure can be decomposed into a model
that accounts for exposure to SO2 of ambient (/¦.' ,) and nonambient (Em) origin of the
form:
Et — Ea + Ena
Equation 3-2
Although indoor combustion of sulfur-containing fuels, particularly kerosene, is a
nonambient source of SO2 (see Section 3.4.1). these sources are specific to individuals
3-3
-------
and may not be important sources of population exposure. This ISA focuses on the
ambient component of exposure because this is more relevant to the National Ambient
Air Quality Standards (NAAQS) review. Ambient sources of SO2, such as electric power
generating units (EGUs) and industrial fuel combustion, are described in Section 2.2.1.
Assuming steady-state outdoor conditions, A', can be expressed in terms of the fraction of
time spent in various outdoor and indoor (including enclosed microenvironments such as
vehicles) microenvironments (U.S. EPA. 2006; Wilson et al.. 2000):
Ea = £f0C0 + ZfiFinf.iCo
Equation 3-3
where/= fraction of the relevant time period (equivalent to dt in Equation 3-1); subscript
o denotes outdoor microenvironments; subscript /' denotes indoor microenvironments;
and = infiltration factor for indoor microenvironment Equation 3-3 is subject to the
constraint T.f„ + I/,' = 1 to reflect the total exposure over a specified time period, and each
term on the right-hand side of the equation has a summation because it reflects various
microenvironmental exposures. Here, "indoors" refers to being inside any aspect of the
built environment, [e.g., home, office buildings, enclosed vehicles (automobiles, trains,
buses), or recreational facilities (movie theaters, restaurants, bars)]. "Outdoor" exposure
can occur in parks or yards, on sidewalks, and on bicycles or motorcycles. Assuming
steady-state ventilation conditions, the infiltration factor r) is a function of the
penetration (/') of SO2 into the microenvironment, the air exchange rate (AER) (a) of the
microenvironment, and the rate of SO2 loss (k) in the microenvironment:
Pa
Finf = (a + k)
Equation 3-4
In epidemiologic studies, it is often assumed that the fraction of time spent outdoors can
be expressed cumulatively as the indoor terms still retain a summation because
infiltration differs for different microenvironments. If an epidemiologic study employs
only one measure for C0, then the assumed model of an individual's exposure to ambient
SO2, given in Equation 3-3. is re-expressed solely as a function of C0:
Ea = (fo + ^ fjFjnf.i) C0
Equation 3-5
The spatial variability of outdoor SO2 concentrations and epidemiologic study design
determine whether Equation 3-5 is a reasonable approximation for Equation 3-3. Spatial
variability of outdoor SO2 is influenced by proximity to sources, source characteristics,
meteorology, built and natural topography, and oxidation rates. These equations also
3-4
-------
assume steady-state microenvironmental concentrations. Errors and uncertainties inherent
in using Equation 3-5 in lieu of Equation 3-3 are described in Section 3.4.4 with respect
to implications for interpreting epidemiologic studies. Epidemiologic studies may use
concentration measured at a fixed-site monitor to represent ambient concentration; thus a,
the ratio between personal exposure to ambient SO2 and the ambient concentration of
SO2, is defined as:
_ Ea
a C
Equation 3-6
Combining Equations 3-5 and 3^6 yields:
a — fo EfiFinf.i
Equation 3-7
where a varies between 0 and 1. Estimates of a for SO2 are provided in Section 3.4.1.3. If
a person's exposure occurs in a single microenvironment, the ambient component of a
microenvironmental SO2 concentration can be represented as the product of the ambient
concentration and I'm. Time-activity data and corresponding estimates of I'm for each
microenvironmental exposure are needed to compute an individual's a with accuracy
(U.S. EPA. 2006). In epidemiologic studies, a is assumed to be constant in lieu of
time-activity data and estimates of Fm, which varies with building- and
meteorology-related air exchange characteristics (Section 3.4.1.1). If important local
outdoor sources and sinks exist that are not captured by fixed-site monitors, then the
ambient component of the local outdoor concentration may be estimated using dispersion
models, land use regression (LUR) models, receptor models, fine-scale chemical
transport models (CTMs), or some combination of these techniques, which are described
in Section 3.3.2.
3.2.3 Exposure Considerations Specific to Sulfur Dioxide
The inhalation exposure pathway relevant for SO2 is influenced by sources, chemistry,
meteorology, and ambient concentrations, described in detail in Chapter 2 and
summarized briefly here. The majority of SO2 is emitted by coal-fired EGUs
(Section 2.2); the point source nature of these emissions contributes to the relatively high
spatial variability of SO2 concentrations (both ambient and exposure) compared with
pollutants such as PM and ozone (O3) (Sections 2.5 and 3.4.2.2). Another contributing
factor to spatial variability is the dispersion and oxidation of SO2 in the atmosphere
(Section 2.3). resulting in decreasing ambient SO2 concentrations with increasing
3-5
-------
distance from the source. SO2 from point sources travels as a plume, which may or may
not impact portions of an urban area depending on meteorological conditions. Ambient
SO2 concentrations do not exhibit consistently strong temporal variability over daily or
seasonal time scales (Section 2.5): however, in some areas, concentrations are low during
nighttime and show a daytime maximum, affecting temporal exposure patterns. Due to
the relative lack of indoor SO2 sources, personal SO2 exposure is expected to be
dominated by ambient exposure (Section 3.4.1.3).
3.3 Methodological Considerations for Use of Exposure Data
This section describes techniques that have been used to measure or model concentrations
of SO2 that serve as surrogates for personal SO2 exposures in epidemiologic studies.
Common methods for assigning an exposure surrogate from monitoring data include
using ambient SO2 concentration measured at a single fixed-site monitor to represent
population exposure (Section 3.3.1.1). averaging ambient SO2 concentrations from
multiple fixed-site monitors, or using personal monitoring data (Section 3.3.1.2).
Fixed-site monitoring data are often used as exposure concentration surrogates in
time-series epidemiologic studies to examine how changes in SO2 exposure over time are
associated with changes in a health outcome (Section 3.4.4.1). Panel epidemiologic
studies may use personal monitors to estimate personal exposure (Section 3.4.4.3).
Modeling methods vary in complexity from source proximity models (SPM)
(Section 3.3.2.1) to monitoring data-based methods [LUR (Section 3.3.2.2) and inverse
distance weighting (IDW) (Section 3.3.2.3)1 to physics-based models [dispersion models
(Section 3.3.2.4) and CTMs (Section 3.3.2.5)1 to microenvironmental exposure models
(Section 3.3.2.6) and are often used to produce exposure concentration surrogates for
long-term epidemiologic studies (Section 3.4.4.2). Previous studies from the 2008 SOx
ISA (U.S. EPA. 2008d) are described along with newer studies.
3.3.1 Measurements
3.3.1.1 Fixed-Site Monitoring
Fixed-site monitors are sited for the purpose of determining whether attainment goals are
met under the Clean Air Act. However, fixed-site monitoring ambient SO2 concentration
data are also often used in epidemiologic studies as a surrogate for exposure to SO2, as
discussed in Section 3.4.4. Methods, errors, and uncertainties regarding measurements
made by fixed-site monitors are described in Section 2.4. The effect of errors and
3-6
-------
uncertainties due to instrumentation issues depends on epidemiologic study design, as
described further in Section 3.4.4. Various uses of these data are possible depending on
the design of the epidemiologic study. Short-term (e.g., daily, hourly) data can be used
for time-series studies and long-term (e.g., annual average) data for longer term studies.
For a given core-based statistical area (CBSA), fixed-site monitors are sited at a fixed
location based on the number of people living in the CBSA and the sources of SO2
emissions (40 CFR 58, Appendix D). Even in CBSAs with multiple monitors, the
monitors do not fully capture spatial variability in SO2 concentration across the study
area. Moreover, fixed-site monitors that only log hourly average SO2 concentration may
not fully capture temporal variability; monitors that log 5-minute avg or 5-minute hourly
max SO2 concentration data better characterize temporal variability.
3.3.1.2 Personal and Microenvironmental Monitoring Techniques
Personal and microenvironmental SO2 monitors have been used in studies characterizing
relationships between indoor and outdoor SO2 concentrations (Section 3.4.1.2) and
relationships between personal exposure to SO2 and ambient SO2 concentrations
(Section 3.4.1.3). Additionally, personal monitoring is used infrequently in the
epidemiologic studies described in Chapter 5. As described in the 2008 SOx ISA (U.S.
EPA. 2008d). both active and passive samplers have been used to measure personal and
microenvironmental SO2 exposures. The Harvard-Environmental Protection Agency
(EPA) annular denuder system is a stationary active sampler initially developed to
measure microenvironmental concentrations of particles and acidic gases simultaneously
(Koutrakis et al.. 1988); Braueretal. (1989) modified it to serve as a personal exposure
monitor. The system draws air past an impactor to remove particles and then through an
annular denuder coated with sodium carbonate to trap SO2 and other acidic gases. Gases
collected within the denuder are extracted with ultrapure water and analyzed by ion
chromatography. The detection limit depends on the sensitivity of the ion
chromatography analysis as well as the volume of air sampled, and is typically below
1 ppb (Braueretal.. 1989). with a collection efficiency of 99.3% for a 24-hour
measurement (Koutrakis et al.. 1988V Similar denuder-type systems have been used in
other microenvironmental monitoring studies [e.g., (Patterson and Eatough. 2000)1.
Another active sampler, developed for a scripted personal exposure study in Baltimore,
MD, used a hollow glass denuder coated with triethanolamine, with SO2 detection by ion
chromatography (Chang et al.. 2000). At a sampling rate of 100 mL/minute for 1 hour,
the detection limit was 62 ppb, resulting in many of the 1-hour SO2 samples being below
the detection limit; see Section 2.5 for a summary of typical ambient SO2 concentrations.
For microenvironmental sampling, Federal Reference Method (FRM) and Federal
3-7
-------
Equivalent Methods (FEMs), described in Section 2.4. have also been deployed
[e.g., (Maggos et al.. 2016; Bozkurt et al.. 2015; Halios et al.. 2014)1.
Passive badge-type samplers have also been developed to eliminate the need for a
powered sampling pump. A common version is manufactured by Ogawa USA, Inc. and
consists of a cellulose fiber filter coated with triethanolamine (Ogawa & Co. 2007). SO2
is detected via ion chromatography with a reported detection limit for a 24-hour sample
of 2-6 ppb (Sarnat et al.. 2006; Sarnat et al.. 2005; Sarnat et al.. 2000). Passive badge
samplers can also be combined with active particle samplers to create a multipollutant
sampler [e.g., (Dcmokritou et al.. 2001)1. Another type of passive sampler, useful for
microenvironmental measurements of SO2, involves a sampling cassette with a sodium
carbonate-soaked glass fiber filter, again analyzed for SO2 with ion chromatography
(Triche et al.. 2005). Passive samplers for measuring SO2 concentrations are not very
sensitive to potential interferants, such as temperature or relative humidity (Swaans et al..
2007). The cumulative sampling approach and the relatively high detection limit of the
passive badges makes them mainly suitable for monitoring periods of 24 hours or greater.
This limits their ability to measure short-term daily fluctuations in personal SO2
exposures. If the passive sampling data below detection limits are censored, then the
estimated mean and distribution of the SO2 concentrations may be biased. Maximum
likelihood estimation and bootstrap methods can be employed to estimate unbiased
means and data distributions for censored data-sets (Zhao and Frev. 2006; Frev and Zhao.
2004; Zhao and Frev. 2004).
3.3.2 Modeling
Many models can be used to predict the outdoor concentration of SO2 across geographic
regions or at specific locations of interest where people spend time (e.g., outdoors at
homes, schools, workplaces, roadways). The modeled concentration can be used as a
surrogate for human exposure to SO2. Most models do not estimate exposures to ambient
SO2 directly, because those models are not designed to include time-activity patterns and
indoor concentrations of ambient SO2 in various microenvironments. Approaches to
modeling exposure concentration described below include source proximity models
(SPM), LUR, IDW models, dispersion models, and CTM. These models can be employed
at urban, regional, or national scales to estimate daily, or longer, average ambient SO2
concentrations as an exposure surrogate. Short-term (e.g., daily) ambient SO2
concentration estimates are needed for ambient SO2 exposure surrogates in acute
exposure assessments, whereas long-term (e.g., annual) ambient SO2 concentration
estimates can be used for ambient SO2 exposure surrogates in chronic exposure
assessments. Additionally, microenvironmental models, which incorporate time-activity
3-8
-------
data to estimate exposures directly, are included in this discussion. Table 3-1 provides an
overview of the modeling approaches discussed in this section.
Table 3-1
Comparison of models used for estimating exposure concentration
or exposure. Factors available in each model are checked.
Type of Model
Factors
SPM IDW LUR
Dispersion
CTM
Microenviron mental
Distance from
source
X
X
X
X
X
X
Emission rate
X X
X
X
X
Terrain or land
use
X
X
X
X
Dispersion
X
X
X
Chemistry
X
X
X
Human activity
X
Inhalation
X
CTM = chemical transport model, IDW = inverse distance weighting, LUR = land use regression, SPM
= source proximity models.
3.3.2.1 Source Proximity Models
SPMs provide a simple method to estimate ambient SO2 concentration as a surrogate for
ambient SO2 exposure concentration. These models calculate the distance from receptors
(e.g., homes, schools) to a source of SO2 emissions (e.g., industrial facilities). It is
assumed that ambient SO2 concentration is some function of distance from the source.
SO2 emitted from a point source is thought to disperse as a meandering plume, such that
average ambient SO2 concentration decreases with distance from the source
(Section 2.6.1). Source proximity models do not account for emissions, stack
characteristics, plume dispersion, meteorology, or atmospheric chemistry.
These models do not necessarily account for the effect of stack height to limit ambient
SO2 concentrations in the immediate vicinity of the point source, which produces zero
SO2 concentration close to the source and then a peak ground-level concentration at the
location of plume touchdown. Burstvn et al. (2008) avoided the stack height issue by
3-9
-------
modeling ambient SO2 concentration as a function of the inverse distance within 2- and
50-km buffers of each gas plant and oil well. In another study, proximity to source was
treated as a Boolean variable as a surrogate for high and moderate ambient SO2 exposure
(Cambra et al.. 2011). Likewise, Liu et al. (2012b) computed relative risk (RR) of
respiratory disease using ZIP codes with fuel-fired power plants compared with the
reference of ZIP codes without fuel-fired power plants.
SPMs are widely applied for exposure assessments because few input data are required.
The main limitation of an SPM is the potential for error in the exposure surrogate because
none of the factors other than distance from the source affecting emission rates,
dispersion, and photochemical activity of pollutants (e.g., emission rates, atmospheric
physics, chemistry, meteorology) are included as model variables [e.g., Zou et al.
(2009a)l.
To improve the accuracy of SPMs in providing a surrogate for exposure, an
emission-weighted proximity model (EWPM) was developed that considers the emission
rate and duration of each ambient SO2 point source, in addition to the distance from the
source. The model still does not account for stack characteristics, plume dispersion,
meteorology, or atmospheric chemistry. Zou et al. (200%) evaluated the SPM and
EWPM to estimate ambient SO2 concentrations in Dallas and Ellis counties, TX.
Normalized ambient SO2 concentration estimates based on SPM and EWPM were
compared to normalized ambient SO2 concentration measurements at three monitoring
sites. EWPM-based ambient SO2 concentration estimates agreed more closely to the
observed ambient SO2 concentrations than SPM-based ambient SO2 concentration
estimates. Epidemiologic estimates of risk also were in closer agreement between EWPM
and American Meteorological Society/Environmental Protection Agency Regulatory
Model (AERMOD) compared with the comparison of results using SPM and AERMOD
(Zou et al.. 2011). In addition, surface maps of EWPM- and SPM-predicted ambient SO2
concentrations across two counties showed that with SPM risk of exposure is usually
overestimated in the region of dense emission sources and underestimated where
emission sources were sparse (Zou et al.. 2009b).
SPMs provide a simple estimation of SO2 concentrations based on location information.
However, they are fundamentally limited because they are not designed to account for
emissions, stack characteristics, plume dispersion, meteorology, or atmospheric
chemistry. EWPMs provide an alternative to SPMs that account for emissions for
improved prediction but still suffer from most of the limitations of SPMs.
3-10
-------
3.3.2.2
Land Use Regression Models
LUR models are used to estimate ambient SO2 concentration as a surrogate for exposure
in some large health studies of long-term SO2 exposure, because they provide spatial
variability in estimates of ambient SO2 concentration across the geographic area of the
study population. A detailed description of LUR models is provided in Chapter 3 of the
2016 ISA for Oxides of Nitrogen (U.S. EPA. 2016dV Briefly, LUR fits a multiple linear
regression model of concentration based on local data (e.g., proximity to SO2 emissions
sources, road length, land use, population density) and then applies that model to
locations without monitors to predict the ambient SO2 concentration field. LUR allows
the ability to characterize more completely the spatial variation by predicting at arbitrary
locations, compared with methods where data are available at a limited number of points,
such as fixed-site monitoring (Marshall et al.. 2008). Metrics used for comparing
modeling approaches include spatial scale, averaging time, and out-of-sample coefficient
of variation, and root mean squared error (RMSE) as measures of prediction error to
cross-validate the model. In-sample coefficient of variation and RMSE are sometimes
reported to illustrate the model training error. However, studies in the literature of LUR
model results do not consistently report all of these parameters. The discussion of LUR
models below includes the metrics provided in specific papers.
Models are typically calibrated using ambient SO2 concentration data from passive
sampler measurements and several local predictor variables. LUR models are typically
based on several days, weeks, or years of data and thus do not account well for short-term
temporal variability in the ambient SO2 concentration estimates. Hence, LUR is typically
used to estimate air pollution exposure in long-term epidemiologic studies. Although
LUR is usually employed for nitrogen dioxide (NO2), it has also been used to study
spatial variability in ambient SO2 concentration in a small number of studies [e.g., Atari
et al. (2008)1. Important methodological issues for interpreting LUR model results
include number of measurement sites used to fit the statistical model, predictor variable
selection, and comparison of LUR performance among model formulations and with
other models. These issues affect how well the spatial variability of ambient SO2
concentration in a city is represented by the LUR. For example, Gulliver et al. (2011)
observed a small fractional bias in the models of-4% to -6%, depending on year, for a
retrospective LUR model of annual average SO2 concentration over the U.K. for years
1962-1991. The authors note that the model tended to overpredict the SO2 concentrations
in rural areas and underpredict in urban areas without a high monitor density. Leave one
out cross-validation R2 declined from 0.71 in 1962 to 0.31 in 1991, which the authors
attribute to lower spatial variation in SO2 concentration in later years. Therefore, LUR is
not appropriate for representing average conditions over time periods long enough for
source conditions to change.
3-11
-------
LUR models have been applied to estimate ambient SO2 concentrations in close
proximity to industrial SO2 sources. Atari et al. (2008) developed an LUR model to
predict ambient SO2 concentrations in Sarnia, Ontario, Canada, an area known as
"Chemical Valley" for its high density of chemical industries. Ambient SO2
concentrations measured over a 2-week period by passive badge monitors were used to
"train" the model, and the explanatory variables for the LUR model were distance to an
industrial zone, location within 1,200 m of industrial areas, and location within 100 m of
major roads. Measurements of ambient SO2 concentration for model training were
collected with passive samplers at 37 locations across the city for 2 weeks in the fall of
2005, with an average concentration of 3.4 ppb. Although the sampling time was
relatively short, the authors suggested that it was sufficient to capture intra-urban
variability given the spatial distribution of SO2 sources. The in-sample coefficient of
determination was R1 = 0.66. An out-of-sample coefficient of determination was
calculated to cross-validate the model for three separate tests: removing 5, 10, and 50%
of the samplers and then predicting the SO2 concentration at those sites. The
out-of-sample coefficient ranged from R2 = 0.62 to If = 0.73, and the RMSEs of the
out-of-sample predictions were 0.3 to 1 ppb. Atari et al. (2008) attributed this moderate
validation to a skewed ambient SO2 concentration distribution, although skewness
metrics were not provided. These findings suggest that LUR simulates the spatial
variability of SO2 from a point source with reasonable accuracy but may not fully capture
the distribution of SO2 data.
LUR has also been applied to predict ambient SO2 concentrations in the vicinity of urban
sources. Cloughcrtv et al. (2013) modeled concentrations of ambient SO2, NO2,
particulate matter with a nominal aerodynamic diameter less than or equal to 2.5 |im
(PM2 5), and black carbon (BC), measured using passive samplers during 2-week periods
at different locations across New York City, NY. Ambient SO2 concentration was
predicted by the reference site mean (partial R2 = 0.35), number of oil-burning units
(partial R2 = 0.36), and nighttime population within 1 km (partial R2 = 0.06) to give an
overall out-of-sample model fit of R2 = 0.77, where R2 was based on the comparison
between raw ambient SO2 concentrations and model predictions. Traffic covariates were
not included in the model. The study authors thought these findings reflected the presence
of large combustion boilers in Manhattan and western Bronx, where ambient SO2
concentrations were predicted to be highest because sulfur content in residential heating
fuel is high. Ambient SO2 concentration was not influenced by vehicle traffic, unlike the
other air pollutants studied. Beelen et al. (2007) modeled ambient SO2, NO2, nitric oxide,
and black smoke (BS) concentrations as the sum of regional, urban, and local
components. For the urban-scale model, land use variables, such as location in a nonrural,
urban, or industrial area, were included. For the local-scale model, traffic intensity
variables were included. The model produced an in-sample R2 = 0.56. The analysis used
3-12
-------
passive sampling data from four 2-week periods in 1999-2000, when diesel fuel
contained higher concentrations of sulfur before the fuel standards (66 FR 5002) took
effect to reduce sulfur concentrations in diesel fuel for highway vehicles and heavy-duty
vehicles. The out-of-sample RMSE was 1.6 ppb for the regional model and 1.2 ppb for
the urban model; RMSE was not reported for the local model. Ambient SO2
concentrations modeled in the Beelen et al. (2007) study were used as exposure estimates
in a longitudinal cohort study of vascular damage among young adults [see
Section 5.3.2.5 and Lenters et al. (2010)1. Wheeler et al. (2008) applied LUR for a study
of ambient SO2 concentration to estimate exposure in Windsor, Ontario and found that
distance to the Ambassador Bridge, housing density, and SO2 emission sources from
Detroit within 3 km were all significant predictors of ambient SO2 concentration with
in-sample R2 = 0.69 and out-of-sample If = 0.65. Wheeler et al. (2008) also evaluated
LUR performance for predicting ambient SO2 concentration, measured using passive
samplers over 2-week periods for each season, by comparing the LUR results with
measurements to estimate air pollutant exposure in Windsor, Ontario. They found that
correlation of summer predictions of ambient SO2 concentrations with those from other
seasons was lower, suggesting that photochemistry might not be well represented in the
LUR model. These studies illustrate several factors that may affect the accuracy of the
LUR for SO2, including source proximity, source type, and season. Therefore, the
individual study characteristics and model validation results must be evaluated to
determine if LUR provides an accurate exposure prediction.
Spatial variability in ambient SO2 concentrations predicted by LUR has been used to
estimate inter-individual variability in exposure by assuming the ambient SO2
concentration modeled at the study participants' homes matched their exposure. Ambient
SO2 concentrations computed using LUR by Atari et al. (2008) were used by Atari et al.
(2009) to correlate modeled ambient SO2 concentrations with individual and community
perceptions of odor, by Oiamo and Luginaah (2013) to study whether males and females
are affected differently by ambient SO2 exposure, and by Oiamo et al. (2011) to
investigate the relationship between estimated ambient SO2 exposure and access to a
general practitioner. Kanaroglou et al. (2013) used a spatial autocorrelation LUR model
to estimate ambient SO2 concentrations, in which the spatial autocorrelation component
of the model's residuals was removed. Kanaroglou et al. (2013) applied the spatial
autocorrelation LUR model in the vicinity of an industrial area in Hamilton, Ontario,
Canada based on monitoring data obtained on a mobile platform during nonpeak traffic
times (10:00 a.m.-4:00 p.m.) for 62 days over a 5-year period (2005-2010) and observed
that location and difference between wind direction and direction of the industrial area to
the receptor each predicted ambient SO2 concentration (RMSE = 1.24). Inclusion of
autocorrelation did not improve the model substantially. These findings suggest that LUR
3-13
-------
captures inter-individual variability of the study population exposed to SO2 from a point
source with reasonable accuracy in these studies.
The LUR studies evaluated indicate that LUR can do a reasonable job of estimating SO2
exposure concentrations, and LUR is capable of capturing spatial variability of SO2
concentration well. However, some modeling decisions can detract from LUR modeling
accuracy. Use of very long averaging periods (in the example cited above, 30 years) may
lead to inaccurate predictions. Likewise, model variables may affect accuracy of
predictions. These include source proximity, source type, and season. Therefore, the
individual study characteristics and model validation results must be evaluated for each
study to determine if LUR provides an accurate prediction of exposure.
3.3.2.3 Inverse Distance Weighting
IDW, in which ambient SO2 concentration at a receptor point is calculated as the
weighted average of ambient SO2 concentration measured at monitoring locations, has
been used to estimate exposure based on ambient SO2 concentration surfaces. Several
recent studies using IDW have been published. The weighting factor is an inverse
function of distance between the receptor and the monitor. For example, Braueretal.
(2008) and Maclntyre et al. (2011) estimated exposure to ambient SO2 and other
industrial pollutants within 10 km of point sources using an IDW sum of ambient SO2
concentration and the three closest monitors within 50 km for application in
epidemiologic models (Clark et al.. 2010). Often, the weighting factor is the inverse
distance raised to some power, and a higher power is applied to increase the weight on
monitors that are closer to the receptor. Rivera-Gonzalez et al. (2015) applied an
inverse-distance-squared weighting and compared the results with a citywide average,
use of the nearest monitor, or kriging to develop an ambient SO2 concentration surface.
The results from IDW were correlated with the other citywide averaging, nearest monitor,
and ordinary kriging (r = 0.88-0.97), and the mean ambient SO2 concentration estimated
with IDW was within 10% of the mean computed with the other methods. Neupane et al.
(2010) estimated the ambient SO2 concentration surface using both bicubic spline
interpolation and IDW for a study of long-term exposure to air pollutants and risk of
hospitalization for pneumonia in Hamilton, Ontario, Canada in a case-control study
design. Bicubic spline interpolation produced a lower mean ambient SO2 concentration
and larger interquartile range (IQR) compared with IDW. Because there is no reference
value in these studies, it is difficult to conclude that IDW presents any substantial
improvement in prediction accuracy compared with other methods. These findings do
indicate that the results of IDW are comparable to averaging and smoothing methods
when estimating SO2 concentration.
3-14
-------
3.3.2.4 Dispersion Models
Gaussian dispersion models have been applied to estimate ambient SO2 concentration as
a surrogate for human exposure to SO2 [e.g., (Smargiassi et al.. 2009)1. A detailed
description of Gaussian dispersion modeling, along with its strengths and limitations for
modeling ambient SO2 concentrations, can be found in Section 2.6. When used in health
studies, evaluation of the model outputs in comparison with monitoring data can help
determine the applicability of the model. Adjustment of the model surface using monitor
observations to account for biases is another technique to improve population exposure
concentration estimates.
Zou et al. (2009c) developed a hybrid modeling system to estimate source-specific
ambient SO2 concentration across space as a surrogate for population exposure to
ambient SO2 in Dallas County, TX. First, an AERMOD dispersion model was run for
three source scenarios (vehicle only, industrial only, and combined vehicle and
industrial), and kriging interpolation was applied to the modeling results to produce a
monthly average ambient SO2 concentration grid map (100 x 100 m). The population
exposure was next estimated by multiplying the ambient SO2 concentration value
modeled by AERMOD and the corresponding population density value for each grid cell
(100 x 100 m) and for the three source classifications. The results showed that monthly
simulated population SO2 exposure concentrations were moderately correlated with
simulated ambient SO2 concentrations from vehicle sources (r = 0.440) and weakly
correlated with ambient SO2 concentrations from industrial sources (r = 0.069). This
study used emissions data from the year 2000 before the ultra-low sulfur diesel fuel
regulations were enacted.
Lagrangian particle modeling has been used to estimate ambient SOx concentration as a
surrogate for ambient SOx exposure from specific sources (Ancona et al.. 2015) to study
the relationship of long-term exposure to SOx with mortality for all-causes
(Section 5.5.2.2). cardiovascular disease (Section 5.3.2.2). and cancer (Section 5.6.1).
The Lagrangian particle model tracks the movement of SOx as nonreactive parcels
(i.e., massless particles), considering SOx to be a marker of the emission source
representing some combination of directly emitted SO2 and sulfate formed in the
atmosphere (Section 2.3). Gariazzo et al. (2004) compared this type of Lagrangian
particle model against ambient SO2 concentration measurements and observed reasonable
agreement, although the observations seemed to lag the modeled ambient SO2
concentration at times. The results suggest that the model would have provided a
reasonable estimate of exposure in the Ancona et al. (2015) study, especially given the
long-term nature of the study.
3-15
-------
3.3.2.5
Chemical Transport Models
Ambient SO2 concentrations calculated with CTMs, such as the Community Multiscale
Air Quality (CMAQ) model, are sometimes used to estimate human exposure to ambient
SO2 (Section 2.6). CTMs can be applied in epidemiologic studies of either short- or
long-term exposure to ambient SO2 but are more commonly used in long-term ambient
SO2 exposure studies. Given observed biases in the CTMs [e.g., (U.S. EPA. 2008c)l.
much attention has been given to bias correction of these models for application in
exposure assessment. Chen et al. (2014a') evaluated CMAQ v4.7.1 results for several
pollutants and found that ambient SO2 concentration was underpredicted by roughly a
factor of two, but this problem was largely ameliorated through bias correction
techniques. Improvements to modeling ambient SC>2-related reactions have been
corrected in CMAQ v 5.0.2, so that ambient SO2 concentrations used for exposure
surrogates from this or later versions would have smaller exposure errors.
One major limitation of CTMs for estimating ambient SO2 concentrations as exposure
surrogates is that the grid resolution, typically between 4 and 36 km, can be much larger
than the length scale of the meandering plume upon touchdown. This limitation presents
the possibility that ambient SO2 concentrations can be underestimated along the plume
path when localized peaks are averaged over space. Baldasano et al. (2014) recognized
this limitation and merged HYSPLIT with a CTM simulation of ambient SO2 and
particulate matter with a nominal aerodynamic diameter less than or equal to 10 ^m
(PM10) transport in the vicinity of a refinery. HYSPLIT models dispersion of pollutants,
such as ambient SO2, as particle trajectories; the Weather Research and Forecast
meteorological model is coupled with the particle trajectory model to account for wind
speed, wind direction, and atmospheric turbulence. China et al. (2006) nested smaller
grids (1,4, 12 km) within larger grids (36 km) to improve spatial variability of the
simulation. Similarly, Karamchandani et al. (2010) coupled a plume-in-grid model with
CTM that treats dispersion as a Gaussian process with parameters that are set using
micrometeorological conditions. Inclusion of subgrid-scale modeling enables calculation
of the ambient SO2 plume at finer spatial scales so that maximum ambient SO2
concentration, and potentially maximum exposures, can be estimated by the model suite
(Baldasano et al.. 2014).
3.3.2.6 Microenvironmental Exposure Models
Microenvironmental exposure models are designed to account for variations in the
amount of time people spend in different locations by using time-weighted SO2
concentrations in each microenvironment (e.g., outdoors; indoors at home, school,
3-16
-------
workplace; in-vehicle) for the exposure surrogate. Models such as the Stochastic Human
Exposure and Dose Simulation (SHEDS) and Air Pollutants Exposure Model (APEX)
have been used occasionally for exposure assessment in epidemiologic studies (Dionisio
et al.. 2014; Mannshardt et al.. 2013; Chang et al.. 2012a'). and they are also used for the
risk assessment performed as part of the NAAQS review process, as was done for the risk
and exposure assessment during the last review of the SO2 NAAQS (U.S. EPA. 2009c).
The fundamental principles of stochastic population-based exposure models are described
in detail in the 2008 NOx ISA Annex 3.6 (U.S. EPA. 2008a). Briefly, the models
combine ambient concentration data with information on infiltration into enclosed
microenvironments, such as buildings and vehicles (see Section 3.4.1.1). to estimate
microenvironmental concentrations. The models then use demographic variables such as
age and sex to select appropriate activity patterns from a database. For the risk
assessment done during the last review of the SO2 NAAQS, the U.S. EPA used the
Consolidated Human Activity Database (CHAD), which is described in Section 3.4.2.1
and in the 2016 NOx ISA (U.S. EPA. 2016d'). Inhalation rates are determined from the
level of effort associated with each activity (e.g., sitting, walking, or running). Inhalation
rates and microenvironmental concentrations are combined to estimate dose. Depending
on the availability of controlled human exposure data, response functions based either on
microenvironmental exposure concentrations or inhaled dose are used to characterize
expected health effects. For population-level exposure assessments, stochastic exposure
models such as SHEDS and APEX estimate the distribution of exposures across the
population of interest (U.S. EPA. 2012c; Burke et al.. 2001).
3-17
-------
3.3.3
Choice of Exposure Surrogates in Epidemiologic Studies
Epidemiologic studies use a variety of methods to assign a surrogate for ambient SO2
exposure. Study design, data availability, and research objectives are all important factors
when selecting an exposure assessment method. Epidemiologic study design may
influence the choice of exposure surrogate. The influence of exposure error on effect
estimates from epidemiologic studies of different designs is discussed in detail in
Section 3.4.4. Table 3-2 and the following text summarizes various metrics used in
epidemiologic studies of ambient SO2 exposure, appropriate applications for the metrics,
and errors and uncertainties that may be associated with the metrics. Elements included
on the table are based on the studies described in Section 3.3.2. Elements presented are
exposure concentration assignment method (which can include monitor deployment or
making predictions using different types of models), description of the method,
description of the type of epidemiologic applications where this method has been used in
Chapter 5. strengths of the method, limitations of the method, and exposure errors
associated with the exposure concentration data collection or model prediction method.
3-18
-------
Table 3-2 Summary of exposure assignment methods, their typical uses in
sulfur dioxide epidemiologic studies, strengths, limitations, related
errors, and uncertainties.
Exposure
Concentration
Assignment Method Description
Epidemiologic
Application
Strengths
Limitations
Exposure
Errors
Measurement Methods
Fixed-site monitors An FRM or FEM
(Section 3.3.1.1). monitor located at
a fixed location to
measure ambient
SO2
concentration.
Potential for
bias if ambient
SO2
concentration
at a receptor
location is
higher or lower
than the
ambient SO2
concentration
measured at
the monitor;
potential for
imprecision
from
assumption of
constant SO2
concentration
within some
radius ofthe
monitor.
Short-term
community
time-series
studies: surrogate
for ambient SO2
exposure
concentration of a
population within a
city.
Ambient SO2
concentration
measurements
undergo rigorous
quality assurance.
Measurements of
ambient SO2
concentration
made at a fixed
location may differ
from an exposed
individual's true
exposure, and no
spatial variation is
assumed.
Correlation
between
outdoor SO2
concentrations
proximal to the
receptors and
ambient SO2
concentration
measurements
typically
decreases with
increasing
distance from
the monitor,
potentially
leading
simultaneously
to decreased
precision and
to bias towards
the null, as
increased
noise drives
the slope
towards zero.
Long-term
epidemiologic
studies: surrogate
for ambient SO2
exposure
concentration to
compare
populations among
multiple cities.
3-19
-------
Table 3-2 (Continued): Summary of exposure assignment methods, their typical
use in sulfur dioxide epidemiologic studies, strengths,
limitations, and related errors and uncertainties.
Exposure
Concentration
Epidemiologic
Exposure
Assignment Method
Description
Application
Strengths
Limitations
Errors
Microenvironmental
Typically an FRM
Panel
Ambient SO2
Instrument
High detection
monitors
or FEM monitor
epidemiologic
concentration
expense may
limit may lead
(Section 3.3.1.2).
located in an
studies: SO2
measurements
make it difficult to
to bias if
outdoor or indoor
exposure
undergo rigorous
perform sampling
appropriate
microenvironment
(e.g., personal or
quality assurance.
simultaneously in
statistical
to measure
residential
multiple
methods are
ambient SO2
samples) within a
environments.
not used for
concentration.
geographic area.
handling
biased data;
nonambient
SO2 exposure
sampling may
lead to bias.
Active personal
Air is pulled
Panel
SO2 concentrations
High detection
High detection
exposure monitors
through a pump
epidemiologic
are obtained at the
limit.
limit and
(Section 3.3.1.2).
and sampled for
studies: SO2
site of the exposed
potential for
ambient SO2
exposure
person.
nonambient
concentration
(e.g., personal or
SO2 exposure
using ion
residential
sampling may
chromatography
samples) within a
lead to bias if
to measure
geographic area.
appropriate
personal SO2
statistical
exposure.
methods are
not used for
handling
biased data.
Passive personal
exposure monitors
(Section 3.3.1.2).
SO2 is captured
on a coated filter
via passive
exposure for a
time period to
measure a
personal or area
sample.
Panel studies:
ambient SO2
exposure within a
city or among
multiple cities.
SO2 concentrations
are obtained at the
site of the exposed
person.
Integrated sample
does not allow for
time-series
analysis; high
detection limit.
High detection
limit may lead
to bias if
appropriate
statistical
methods are
not used for
handling
biased data;
nonambient
SO2 exposure
sampling may
lead to bias.
3-20
-------
Table 3-2 (Continued): Summary of exposure assignment methods, their typical
use in sulfur dioxide epidemiologic studies, strengths,
limitations, and related errors and uncertainties.
Exposure
Concentration
Epidemiologic
Exposure
Assignment Method
Description
Application
Strengths
Limitations
Errors
Modeling Methods
Source proximity
Ambient SO2
Long-term
Few input data
Does not account
Potential for
model
concentrations
epidemiologic
required.
for emission rate
bias if ambient
(Section 3.3.2.1).
are estimated
studies: surrogate
and duration,
SO2
from distance of
for ambient SO2
stack parameters,
concentration
receptor from
exposure
plume dispersion,
at a receptor
source.
concentration
meteorology,
location is
within a city or
oratmospheric
higher or lower
among multiple
chemistry; over-
than the
cities or regions.
smoothing based
average
on assumption
ambient SO2
that ambient SO2
concentration
concentration is
between the
constant for a
source and
given distance
receptor;
from the source or
potential for
based on
imprecision
smoothing
from overly
function between
smoothed SO2
monitors.
concentration.
Emission weighted
proximity model
(Section 3.3.2.1).
Ambient SO2
concentrations
are estimated
from distance of
receptor to
pollution source,
emission rate,
and duration.
Long-term
epidemiologic
studies: surrogate
for ambient SO2
exposure
concentration
within a city or
among multiple
cities or regions.
Considers emission Does not account Potential for
rate and duration. for stack bias if ambient
parameters, SO2
plume dispersion, concentration
meteorology, or at a receptor
atmospheric location is
chemistry; over- higher or lower
smoothing based than the
on assumption average
that ambient SO2 ambient SO2
concentration is concentration
constant for a between the
given distance source and
from the source or receptor;
based on potential for
smoothing imprecision
function between from overly
monitors. smoothed SO2
concentration.
3-21
-------
Table 3-2 (Continued): Summary of exposure assignment methods, their typical
use in sulfur dioxide epidemiologic studies, strengths,
limitations, and related errors and uncertainties.
Exposure
Concentration
Assignment Method
Description
Epidemiologic
Application
Strengths
Limitations
Exposure
Errors
Land use regression
model
(Section 3.3.2.2).
Measured
ambient SO2
concentrations
are regressed on
local variables
(e.g., land use
factors), and the
resulting model is
used to estimate
ambient SO2
concentrations at
specific locations.
Long-term
epidemiologic
studies: surrogate
for ambient SO2
exposure
concentration,
usually across a
city but sometimes
among multiple
cities.
High spatial
resolution.
Does not account
for emission rates,
stack parameters,
plume dispersion,
or atmospheric
chemistry and
may account for
meteorology only
in terms of wind
speed and wind
direction,
depending on
model formulation;
potential for model
misspecification;
has limited
generalizability,
and moderate
resources are
needed.
Potential for
bias if grid is
not finely
resolved or if
the model is
misspecified or
applied to a
location
different from
where the
model was fit.
Inverse distance
weighting
(Section 3.3.2.3).
Measured
Long-term
High spatial
Does not fully
Potential for
ambient SO2
epidemiologic
resolution.
capture spatial
negative bias if
concentrations
studies: surrogate
variability of
ambient SO2
are interpolated
for ambient SO2
ambient SO2
sources are
to estimate
exposure
concentration
not captured or
ambient SO2
concentration,
among monitors,
SO2
concentration
usually within a
and does not
concentration
surfaces across
city or geographic
account for
is overly
regions. IDW
region.
emissions, stack
smoothed;
uses an inverse
parameters, and
potential for
function of
dispersion; only
positive bias if
distance to
accounts for
SO2 deposition
monitors.
meteorology and
or other loss
chemistry to the
processes;
extent that it is
potential for
calibrated to data
imprecision
with similar
from overly
meteorology and
smoothed SO2
chemistry; over-
concentration.
smoothing based
on assumption
that ambient SO2
concentration is
constant for a
given distance
from the source or
based on
smoothing
function between
monitors.
3-22
-------
Table 3-2 (Continued): Summary of exposure assignment methods, their typical
use in sulfur dioxide epidemiologic studies, strengths,
limitations, and related errors and uncertainties.
Exposure
Concentration
Epidemiologic
Exposure
Assignment Method
Description
Application
Strengths
Limitations
Errors
Dispersion modeling
Ambient SO2
Long-term
High spatial and
Resource
Potential for
(Section 3.3.2.4).
concentrations at
epidemiologic
temporal
intensive, very
bias where the
specific locations
studies: surrogate
resolution,
limited
dispersion
are estimated
for ambient SO2
accounts for
representation of
model does
from emissions,
exposure
atmospheric
atmospheric
not capture
meteorology, and
concentration
physics from local
chemistry or
boundary
atmospheric
within a city or
emission sources
background SO2
conditions and
physics.
geographic region.
concentrations.
resulting fluid
dynamics well
(e.g., in large
cities with
urban
topography
affecting
dispersion).
Chemical transport
Grid-based
Long-term
Strengths include
Limited grid cell
Potential for
model
ambient SO2
epidemiologic
accounting for
resolution
bias if grid
(Section 3.3.2.5).
concentrations
studies: surrogate
stack parameters,
(i.e., grid cell
cells are too
are estimated
for ambient SO2
emission rates,
length scale is
large to
from emissions,
exposure
mixing height,
typically 4-36 km
capture spatial
meteorology, and
concentration,
atmospheric
and much larger
variability of
atmospheric
sometimes within
stability,
than plume width),
ambient SO2
chemistry and
a city but more
meteorology,
resource-
exposures.
physics.
typically across a
larger region.
atmospheric
chemistry, and
complex terrain.
intensive, spatial
smoothing of local
SO2 emissions
sources.
Microenvironmental
Estimates
Panel
Accounts for
Models simulate
Potential for
model (e.g., APEX,
distributions of
epidemiologic
variability of SO2
individuals and
bias when the
SHEDS)
micro-
studies; no
exposures across
their exposures,
modeled
(Section 3.3.2.6).
environmental
epidemiologic
large populations,
but they do not
distributions of
SO2
studies cited here
accounts for
represent an
ambient SO2
concentrations,
use micro-
different
actual population.
concentration,
exposures, and
environmental
concentrations in
indooroutdoor
doses for
models.
different
pollutant
populations
microenvironments,
ratios, and
(e.g., census
accounts for
time-activity
tracts) based on
location-activity
patterns differ
air quality data,
information.
from the true
demographic
distributions.
variables, and
activity patterns.
APEX = Air Pollutants Exposure model; FEM = Federal Equivalent Method; FRM = Federal Reference Method; IDW = inverse distance
weighting; SHEDS = Stochastic Human Exposure and Dose Simulation; S02 = sulfur dioxide.
3-23
-------
Time-series epidemiologic studies examine how changes in SO2 exposure (or a surrogate
for SO2 exposure concentration) are associated with changes in a health outcome over
time. Fixed site monitors that measure SO2 concentration or personal SO2 monitors that
provide an estimate of exposure are often used for that purpose (Section 3.3.1.1V The
advantage of fixed-site monitors is that the monitors undergo rigorous quality assurance,
while the limitation is that the ambient SO2 concentration at the monitor location may
differ from an exposed individual's true exposure. Low correlation between SO2
concentrations at the fixed-site monitor and exposure concentrations at the location of a
receptor can lead to decreased precision of the effect estimate and to bias of the effect
estimate towards the null.
Personal SO2 monitors are useful because obtaining SO2 concentrations at study
participants' changing locations allows for an estimate of exposure, but these monitors
are limited by a high detection limit (Section 3.3.1.2). High detection limit and potential
for nonambient SO2 exposure sampling may both lead to bias in effect estimates where
the exposure was estimated by personal samples. Panel studies often use personal
monitors as well and can suffer from bias for the same reasons.
Long-term average epidemiologic studies examine the influence of SO2 exposure to a
population over a time period of at least a month and often of years. Long-term studies
are often used to compare health effect estimates in populations for different cities or to
examine the impact of spatial variability of SO2 concentrations within a city or region.
Fixed-site monitors are often used for long-term studies that compare health effect
estimates among cities (Section 3.3.1.1). As for time-series studies, bias and reduced
precision can result from differences between SO2 exposure concentrations and SO2
concentrations measured at the monitor. Passive monitors have been employed for
long-term average epidemiologic studies (Section 3.3.1.2). The integrated design of
passive samplers does not permit time-series analysis and also has a high detection limit,
which could lead to bias if appropriate statistical techniques are not applied to the data.
Most of the modeling approaches described in Section 3.3.2 are commonly used for
studies of the health effects of SO2 exposure. Data requirements of the models increase
with model complexity, from SPM to CTM. The strength of SPM (Section 3.3.2.1).
EWPM (Section 3.3.2.1). and IDW (Section 3.3.2.3) is their simplicity. However, the
limitation of these models is that they do not account for emissions, dispersion,
meteorology, or atmospheric chemistry. This can lead to either positive or negative
biases. LUR also does not account for emissions, dispersion, or atmospheric chemistry,
but its strength is that it has high spatial resolution, which reduces but does not eliminate
the potential for bias in in health effect estimates (Section 3.3.2.2). Model
misspecification may also bias long-term health effect estimates using LUR to estimate
the exposure concentration. Dispersion models (Section 3.3.2.4) and CTMs
3-24
-------
(Section 3.3.2.5) rely on atmospheric chemistry and physics to develop estimates of SO2
exposure concentration. Limited representation of the fluid dynamics is a limitation of
dispersion models, and spatial smoothing due to grid resolution can be a limitation of
CTMs. Both factors can lead to bias in the health effect estimates for long-term studies.
Microenvironmental models (Section 3.3.2.6) incorporate time-activity data to overcome
some limitations of spatial smoothing in grid-based models, but they are rarely used in
epidemiologic models.
3.4 Exposure Assessment, Error, and Implications for
Epidemiologic Inference
This section describes exposure assessment issues related to the use of surrogates for
ambient SO2 exposure in epidemiologic studies that may influence or introduce error into
the observed health effect estimate. The section reviews several factors that may
influence exposure to SO2 (Section 3.4.1) and errors in its estimation (Section 3.4.2).
copollutant relationships that may confound the relationship between SO2 exposure and
health effects (Section 3.4.3). and how epidemiologic study results may be influenced by
these factors and relationships (Section 3.4.4).
Section 3.4.4 focuses on three types of epidemiologic studies that are discussed
frequently in Chapter 5: community time-series studies, long-term cohort studies, and
panel studies. Community time-series studies assess the daily health status of a population
of thousands or millions of people over the course of multiple years by estimating
population exposure concentrations across an area using a short monitoring interval
(hours to days). In these studies, the community-averaged concentration of an air
pollutant measured at fixed-site monitors is typically used as a surrogate for individual or
population ambient exposure. Long-term cohort studies, such as the American Cancer
Society (ACS) cohort study, usually involve hundreds or thousands of subjects followed
over several years or decades [e.g., (Jerrettetal.. 2009)1. Concentrations are generally
aggregated over time and by community to estimate exposures. Panel studies, which
consist of a relatively small sample (typically tens) of study participants followed over a
period of days to months, have been used to examine the health effects associated with
short-term exposure to ambient concentrations of air pollutants [e.g., (Delfino et al..
1996)1. Panel studies may also apply a microenvironmental model to represent exposure
concentrations for an air pollutant.
Some information presented in this section, including parameters influencing infiltration
factors (Section 3.4.1.1) and activity patterns (Section 3.4.2.1). can be used in health risk
assessment. They are discussed here, based on their inclusion in a small number of
3-25
-------
epidemiologic studies (Dionisio et al.. 2014; Mannshardt et al.. 2013; Chang et al..
2012a). The more detailed information provides a more comprehensive exposure
characterization that can help explain population-level variability in the exposure and
health effect estimates (Mannshardt et al.. 2013V
3.4.1 Relationships between Personal Exposure and Ambient Concentration
Several factors influence the relationship between personal SO2 exposure and ambient
SO2 concentration. Indoor SO2 concentrations are highly dependent on AER due to the
lack of indoor SO2 sources and the rapid deposition of ambient SO2 after it penetrates into
enclosed microenvironments (Section 3.4.1.1). Generally, indoor SO2 concentrations are
lower than ambient SO2 concentrations measured outdoors. Because people spend the
bulk of their time indoors (Section 3.4.2.1). personal SO2 exposures are often much lower
than ambient SO2 concentrations. For example, Brown et al. (2009) reported the mean
winter personal SO2 exposure concentrations in Boston to be 1.8 ppb, while the ambient
SO2 concentration was 11.3 ppb. Both personal SO2 exposure concentration and ambient
SO2 concentration were even lower in summer, with mean values of near zero and
3.6 ppb, respectively. The following sections describe studies evaluating AER,
relationships between indoor and outdoor SO2 concentrations, and personal-ambient
relationships for SO2.
3.4.1.1 Parameters Influencing Infiltration Factors
Air Exchange Rate
AER, which is the airflow into and out of a building and is represented by a in the
conceptual model presented in Section 3.2.2. influences the rate of entry of ambient SO2
and hence personal exposure to SO2, because people spend an average of more than 80%
of their time indoors (Spalt et al.. 2015; Klepeis et al.. 2001). Several factors affect the
AER, including the physical driving forces of the airflows (e.g., pressure differences
across the building envelope from wind, indoor-outdoor temperature differences, and
mechanical ventilation), building characteristics (e.g., local wind sheltering, tightness of
the building envelope), and occupant behavior (e.g., opening windows, operating
outdoor-vented fans, adjusting thermostat temperature during heating and cooling
seasons). Therefore, substantial spatial and temporal AER variations can occur due to
temporal and geographical differences in weather conditions, building characteristics, and
occupant behavior. The resulting spatial-temporal variations in ambient SO2 exposure
may help explain possible differences in epidemiologic associations between ambient
3-26
-------
SO2 concentrations and health effects in different U.S. communities (Baxter and Sacks.
2014V
Field studies indicate that the AER of U.S. residences varies by season and region, with
substantial variability among different residences within a region, and variability across
regions due to differences in occupant behavior, building age, air conditioning
prevalence, and building type. Yamamoto et al. (2010) reported AER measured at
residences in Los Angeles, CA, Elizabeth, NJ, and Houston, TX as part of the
Relationship Among Indoor, Outdoor, and Personal Air (RIOPA) Study conducted
between 1999 and 2001. Among the three cities and across seasons, median AER was
0.71/hour. Regional differences can be seen when breaking the data down by season and
location. Median AERs in Los Angeles, Elizabeth, and Houston were 0.87/hour,
0.88/hour, and 0.47/hour. Differences between AER for Houston and AER for Los
Angeles and Elizabeth may in part be related to larger home sizes (average home volume
was 304 m3 for Houston, compared with 163 m3 in Los Angeles and 252 m3 in
Elizabeth). Seasonally, median AER was higher in summer compared to winter in Los
Angeles (summer: 1.14/hour; winter: 0.61/hour). However, the opposite pattern occurred
in Houston (summer: 0.37/hour; winter: 0.63/hour) and to some degree in Elizabeth
(summer: 0.88/hour; winter: 1.07/hour). A similar pattern was reported by Jiao et al.
(2012). with lower mean AER in Houston during summer compared to winter, but higher
mean summer AER in New York city and central North Carolina. More prevalent use of
open windows in Los Angeles, where summertime tends to be less humid than in
Elizabeth or Houston, may promote greater air exchange. This difference may grow
smaller with the increased prevalence of air conditioning, because air conditioning usage
is an important factor in infiltration (Allen et al.. 2012). Low AER values in the mild
temperatures of autumn may be due to a diminished "stack effect," which refers to
airflow through the building due to indoor-outdoor temperature differential (Breen et al..
2014b).
Intra- and inter-home variability in AER was also tested in the RIOPA study (Yamamoto
et al.. 2010). Intra-home variability in AER indicated that individual homes' AER
changed considerably between seasons (32, 37, and 37% for Los Angeles, Elizabeth, and
Houston, respectively). Inter-home variability also differed substantially for all three
cities, with the interquartile range of AER exceeding the median AER consistently across
seasons and cities.
Vehicle AERs can be substantially higher than residential AERs, leading to rapid
infiltration of on-road pollutants. While on-road SO2 emissions have declined due to
reductions in fuel sulfur content (Section 2.2.3). high vehicle AER would increase
exposure in areas with high ambient SO2 concentrations. Window position has been
3-27
-------
shown to be the most important determinant of vehicle AER, even when windows are
only partially open (Ott et al.. 2008). When windows are closed, several other factors can
affect vehicle AER, including vehicle make and model, vehicle age, driving speed, and
fan/recirculation setting on the vehicle ventilation system. The combined effect of these
factors result in AERs that vary by more than two orders of magnitude, from less than
1/hour (approximately equivalent to atypical residential AER) to more than 100/hour
(Huddaet al.. 2011). In a model fit to AER measurements on 59 vehicles driven at three
different speeds under recirculation conditions, the most important variables were vehicle
age, mileage (which is related to age), and speed, plus an adjustment for manufacturer
(Train et al.. 2011). Fan speed and vehicle shape were not influential variables.
Penetration Factor
Limited information was identified regarding the penetration factor/' (Equation 3-4).
Lopez-Aparicio et al. (2011) measured SO2 concentrations indoors and outdoors at the
National Library in Prague, Czech Republic from July 2009 to March 2010 and observed
SO2 penetration values ranging from P = 0.25 to 0.74. The authors described the library
as naturally ventilated, and it is unclear whether windows were open or closed during the
measurement periods. Measured outdoor SO2 concentrations were higher for the cold
months of January, February, and March compared with the remainder of the sampling
campaign, and penetration was lower during that period (P = 0.25 to 0.48). The literature
search only produced this one recent study of SO2 infiltration.
Deposition Factor
Indoor sources of SO2 are relatively scarce and SO2 rapidly reacts with indoor surfaces
[see Grontoft and Ravchaudhuri (2004) and references cited therein] or oxidizes rapidly
via indoor Criegee intermediates [see Section 2.3.1 for a description of Criegee chemistry
or Shallcross et al. (2014) for the role of indoor Criegee intermediates in SO2 losses]. The
main indoor source of SO2 is combustion of sulfur-containing fuels, such as kerosene,
which is generally used for an emergency or supplemental source of heat in the U.S.
Unlike fireplaces, woodstoves, or gas space heaters, kerosene heaters caused elevated
SO2 concentrations indoors in a study conducted in Connecticut and Virginia (Tricheet
al.. 2005). The median indoor SO2 concentration measured by passive sampler over
2 weeks in homes using kerosene heat was 6.4 ppb, compared with 0.22 ppb for homes
that did not use kerosene heat in the 2-week period. This relatively low concentration
when the kerosene heater was not in use is consistent with the rapid removal rate of
infiltrated ambient SO2. As discussed in Section 2.3. SO2 is removed from the
atmosphere by both dry and wet deposition to surfaces, represented by k in the conceptual
3-28
-------
model presented in Section 3.2.2. The deposition rate of SO2 in apartments in Athens,
Greece was found to range from 0.76-4.3/hour, similar to the rate observed for O3, but an
order of magnitude higher than the deposition rate measured for NO2 (Halios et al..
2009V
Understanding air exchange rate, penetration, and deposition inform characterization of
indoor exposure to ambient SO2. Given that people generally spend the majority of their
time indoors (Section 3.4.2.1). failure to account for these processes could result in
biasing SO2 exposure estimates and subsequent addition of bias and/or variability in the
health effect estimates obtained from epidemiologic studies.
3.4.1.2 Indoor-Outdoor Relationships
A number of studies from the U.S., Canada, Europe, and Asia summarized in the 2008
SOx ISA (U.S. EPA. 2008d). as well as a few new studies conducted outside the U.S.,
have characterized the relationship between outdoor and indoor SO2 concentrations.
Ratios and slopes of the indoor SO2 concentration versus the SO2 concentration
immediately outside the indoor microenvironment had an extremely wide range in the
studies described in the 2008 SOx ISA, from near zero to near unity. One of the most
detailed older studies of SO2 in a school was able to detect an indoor-outdoor slope of
0.02-0.03, with near-zero intercept and a correlation of 0.79-0.91, while measuring
indoor concentrations <1 ppb, obtained over 10-hour periods when school was in session
and 14-hour periods when the school was vacant (Patterson and Eatough. 2000). Studies
conducted since the 2008 SOx ISA have focused on public buildings and show generally
similar results to older studies. A historic library in Prague without heating or air
conditioning had indoor:outdoor ratios of 0.25-0.74 (mean = 0.49) for monthly average
outdoor SO2 concentrations of 1-7 ppb obtained with passive samplers (Lopez-Aparicio
et al.. 2011). In Brazil, ratios of average indoor and outdoor SO2 concentrations from
2-week passive samples were 0.7 and 1.0 for urban and suburban schools, respectively
(Godoi et al.. 2013).
Several factors could contribute to the differences observed among studies, including
building characteristics (e.g., forced ventilation, building age, and building type such as
residences or public buildings), behaviors affecting air exchange rates such as opening
windows, indoor deposition of SO2, and analytical capabilities. When reported,
correlations between indoor and outdoor ambient SO2 concentrations were relatively high
(>0.75), suggesting that variations in outdoor ambient SO2 concentration are driving
indoor SO2 concentrations. These high correlations were observed across seasons and
geographic locations. This is consistent with the relative lack of indoor sources of SO2
3-29
-------
(Section 3.4.1.1). For other criteria pollutants, nonambient sources can be an important
contributor to total personal exposure, but personal SO2 exposure is expected to be
dominated by ambient SO2 in outdoor microenvironments and in enclosed
microenvironments with high air exchange rates (e.g., buildings with open windows and
vehicles).
3.4.1.3 Personal-Ambient Relationships
As discussed in the 2008 SOx ISA (U.S. EPA. 2008d). personal monitoring studies for
SO2 exposure assessment have frequently found that most SO2 exposure concentrations
are below the detection limit of the personal samplers used in the study. Several studies
using passive samplers (Section 3.4.1.2) found that 95% or more of the personal SO2
exposure concentrations were less than the field detection limit of 2-6 ppb for 24-h avg
samples (Sarnat et al.. 2006; Sarnat et al.. 2005; Sarnat et al.. 2001; Sarnat et al.. 2000).
Thus, these data are not suitable for evaluating the relationship between personal
exposure and ambient concentration for SO2.
A study in Boston using a different type of sampler, a personal annular denuder
(Section 3.3.1.2) with a detection limit of 0.19 ppb, found that the slope between 24-hour
personal SO2 exposure concentration and ambient SO2 concentration was 0.13, with a
standard error of 0.02 and zero intercept (Brauer et al.. 1989). The 2008 SOx ISA
reported slopes of 0.03-0.13. Assuming that there are no nonambient sources of SO2
(Section 3.2.2). the slope serves as an estimate for a. The R2 value was 0.43 (r = 0.66) in
this analysis, which excluded values below the detection limit, indicating that personal
SO2 exposure concentration was moderately correlated with ambient SO2 concentration.
3.4.2 Factors Contributing to Error in Estimating Exposure to Ambient Sulfur
Dioxide
Ambient SO2 concentrations measured at fixed monitoring sites are commonly used for
exposure surrogates in community time-series (Sections 5.2.1. 5.3.1) and long-term
cohort (Sections 5.2.2. 5.3.2. 5.5.2) epidemiologic studies. As noted in Section 3.3.1.1.
use of a fixed-site SO2 monitor to capture a surrogate for true, likely unobserved ambient
SO2 exposure may lead to exposure error. Factors that may influence this type of error
include human activity patterns, spatial and temporal variation in ambient SO2
concentration, and indoor exposure to ambient SO2 including variation in infiltration
parameters (Brown et al.. 2009; Zeger et al.. 2000). Additionally, uncertainty in the
3-30
-------
metric used to represent exposure may be influenced by method detection limit, accuracy,
and precision of the instrument. These factors are discussed in the following section.
3.4.2.1 Activity Patterns
The activity pattern of individuals is an important determinant of their exposure.
Variation in SO2 exposure concentrations among microenvironments means that the
amount of time spent in each location will influence an individual's exposure to ambient
SO2. The effect of activity pattern on exposure is explicitly accounted for in Equation 3-3
by the fraction of time spent in different microenvironments. As discussed in the 2008
SOx ISA (U.S. EPA. 2008d). although activity patterns vary both among and within
individuals, resulting in corresponding variations in exposure across a population and
overtime, people generally spend more than 80% of their time indoors (Spalt et al.. 2015;
Klepeis et al.. 2001).
Time spent in different locations has been found to vary by age. Table 3-3 summarizes
National Human Activity Pattern Survey (NHAPS) data reported for four age groups,
termed very young (0-4 years), school age (5-17 years), working (18-64 years), and
retired (65+ years) (Klepeis et al.. 1996). Although this survey represents older data, an
uncertainty analysis conducted for the 2014 O3 REA suggests that historical activity
patterns can generally be used to represent current activity patterns (U.S. EPA. 2014a).
The working population spent the least time outdoors, while the school age population
spent the most time outdoors. NHAPS respondents aged 65 years and over spent
somewhat more time outdoors than adults aged 18-64 years, with a greater fraction of
time spent outdoors at a residence. Children aged 0-4 years also spent most of their
outdoor time in a residential outdoor location. On average, the fraction of time spent
outdoors by school age respondents was 2.62 percentage points higher than working
respondents, corresponding to approximately 38 minutes more time outdoors per day.
Moreover, in a survey comparing children (mostly less than age 8 years), their parents
who were mostly under age 55 years, and adults older than age 55 years, a larger
proportion of children reported spending over 30 minutes performing vigorous outdoor
physical activity (Wu et al.. 2011b).
3-31
-------
Table 3-3
Mean fraction of time spent in outdoor locations by various age
groups in the National Human Activity Pattern Survey study.
Age Group (yr)
Residential-Outdoor (%)
Other Outdoor (%)
Total Outdoors (%)
0-4
5.38
0.96
6.34
5-17
5.05
2.83
7.88
18-64
2.93
2.33
5.26
65+
4.48
1.27
5.75
Source: Data from (Klepeis et al.. 1996).
Longitudinal activity pattern information is also an important determinant of exposure, as
different people may exhibit different patterns of time spent outdoors over time due to
race/ethnicity, age, sex, employment, health status, and lifestyle-dependent factors. Spalt
et al. (2015) analyzed the relationship between time-activity patterns and demographic
patterns for the Multi-Ethnic Study of Atherosclerosis Air cohort. They found that time
spent outdoors was best predicted by employment status, and white, black, or Hispanic
study participants were more likely to spend time outdoors compared with participants of
Chinese ethnicity. These differences may manifest as higher mean SO2 exposures or
more frequent high-exposure episodes for some individuals. The extent to which
longitudinal variability in individuals contributes to the population variability in activity
and location can be quantified by the ratio of between-person variance to total variance in
time spent in different locations and activities [the intra-class correlation coefficient
(ICC)]. Netherv et al. (2009) studied time-activity patterns in a cohort of 62 pregnant
women in Vancouver, Canada observed that activity reduced over the course of
pregnancy. Xue et al. (2004) quantified ICC values in time-activity data collected by
Harvard University for 160 children aged 7-12 years in Southern California (Gevh et al..
2000). For time spent outdoors, the ICC was approximately 0.15, indicating that 15% of
the variance in outdoor time was due to between-person differences. The ICC value
might be different for other population groups. Schwab et al. (1992) studied time-activity
patterns among fourth- through sixth-grade children in a cohort of 50 children with
asthma or persistent wheeze and 50 children not reporting respiratory symptoms. Those
with asthma or wheeze reported more time outside on both school and nonschool days.
Several methods are available for sampling diary information, and the method chosen can
affect estimated personal SO2 exposures and related exposure errors. Che et al. (2014)
evaluated how diary sampling methods influenced estimates of children's exposure (in
this case, to ambient PM2 5). Random resampling, diversity and autocorrelation, and
3-32
-------
Markov-chain cluster methods of diary sampling were tested. The three sampling
methods provided similar results for total ambient exposure, outdoor ambient exposure,
and ambient exposure at homes and indoor locations not including home, school, or
vehicles.
The U.S. EPA's National Exposure Research Laboratory has consolidated many of the
most important human activity databases into one comprehensive database called CHAD.
The current version of CHAD contains data from 22 human activity pattern studies
(including NHAPS), which were conducted between 1982 and 2010 and evaluated to
obtain over 54,000 person-days of 24-hour human activities in CHAD (Isaacs. 2014;
McCurdv et al.. 2000). Five studies conducted between 1997 and 2010 comprising over
30,000 person-days have been added to CHAD since the previous SOx ISA (University
of Michigan. 2016; Isaacs et al.. 2013; Wu et al.. 2012; Hertz-Picciotto et al.. 2010;
Knowledge Networks. 2009; Williams et al.. 2009). The surveys include
probability-based recall studies conducted by U.S. EPA and the California Air Resources
Board, as well as real-time diary studies, telephone interviews, and internet-based surveys
conducted nationally and in individual U.S. metropolitan areas using both
probability-based and volunteer subject panels. All ages of both sexes are represented in
CHAD. The data for each subject typically consist of 1 to 3 days of activity diaries, in
which each activity is defined by start time, duration, activity type, and
microenvironmental classification (i.e., location). Activities vary from 1 minute to 1 hour
in duration, with longer activities being subdivided into clock-hour durations to facilitate
exposure modeling. CHAD also provides information on the level of exertion associated
with each activity, which can be used by exposure models, including the APEX model, to
estimate ventilation rate and pollutant dose (Section 3.3.2.6).
Time-activity patterns vary both across and within different populations and lifestages.
CHAD has a large set of activity diaries, but it cannot represent all of the variation
observed in study area(s) chosen for risk assessment. Previous uncertainty
characterizations conducted for the 2014 O3 REA (U.S. EPA. 2014a) suggested that
activity data for individuals with asthma is comparable to activity data for healthy
individuals and similar activity patterns are observed across different regions of the U.S.
Other factors that are not accounted for, but could be important influences on exposure,
include SES and intra-urban differences in outdoor and other activities (U.S. EPA.
2014a).
Algorithms for generating longitudinal activity patterns reduce uncertainty that would
result from repeatedly sampling CHAD for activities on successive days, given that
individuals often have similar day-to-day activity patterns. The method used in the 2014
O3 REA (U.S. EPA. 2014a) involves a within-person autocorrelation statistic and a
3-33
-------
population diversity statistic, which together can help represent the repeated nature of
individual activities while allowing for variability across the population.
To improve the characterization of activity patterns, mobile electronic devices, such as
smartphones with embedded global positioning system (GPS) receivers and dedicated
GPS data loggers, are increasingly used to collect time-location information. GPS
technology has the potential to provide increased resolution in recording activity patterns.
For example, Glasgow et al. (2014) analyzed the frequency of positional data collection
by Android-based smartphones among a panel of study participants and found that on
average 74% of the data were collected over intervals shorter than 5 minutes, which is a
marked improvement over many time-activity studies using diaries. However, manual
processing of GPS data to determine time spent in different microenvironments is limited
due to large (potentially thousands of samples per person per day) and multidimensional
(location, speed, time, signal quality) data sets, missing data due to loss of GPS signal
reception while inside certain buildings, and difficulty discriminating among certain
microenvironments (e.g., steel structures have substantial indoor/outdoor differences in
satellite signal strength). To address these limitations, automated microenvironmental
classification models have been developed (Brccn et al.. 2014a; Kim et al.. 2012; Wu et
al.. 2011a; Adams et al.. 2009; Elgethun et al.. 2007). For example, Breen et al. (2014a)
developed a classification model to estimate time of day and duration spent in eight
microenvironments (indoors and outdoors at home, work, school; inside vehicles; other
locations) from GPS data and geocoded building boundaries. The classification model
estimates were compared with diary data and correctly classified the microenvironment
for 99.5% of the daily time spent by the participants. In conjunction with accelerometers,
air pollutant monitors, and health monitors, GPS-based time-activity data and related
monitors have the potential to reduce error in exposure assessment (NRC. 2012).
Positional errors are a concern for geographic information systems (GIS) and GPS-based
technologies, although post-processing algorithms can compensate for loss of signal
(e.g., when inside a steel-frame building) to some degree. Lane et al. (2013) compared
three geocoding techniques with aerial photography and observed median positional
errors of 7-23 m. Glasgow et al. (2014) also compared smartphone positions with
geocoded diary-based locations to test the positional accuracy of the phones. For all data
combined, the smartphones had a median positional accuracy of 342.3 m, although this
includes any error due to the participant imprecisely recording location information in the
diary. When broken down by network, the median positional accuracy varied from 98.0
to 1,168.8 m. Wu et al. (2010) compared several portable GPS devices to aerial
photography. Median positional errors were 7.3-20.8 m for indoor measurements taken
3 m from a door or window. For outdoor measurements taken 6.1m from a window or
door, median positional errors were 4.1-16.3 m, and for on-road measurements, median
3-34
-------
positional errors were 3.5-5.5 m. Ganguly et al. (2015) compared two automated
(GIS-based) geocoding techniques with GPS positional data in Detroit, MI. Median
positional errors for two GIS methods were 26 m for both methods in comparison with
GPS.
3.4.2.2 Spatial Variability
Failure to account for spatial variability in ambient SO2 concentrations can contribute to
exposure error and error in the health effects estimates produced by epidemiologic
studies, whether the studies rely on fixed-site monitor data or model output as a surrogate
for exposure concentration. Low correlations between the monitor used to measure
concentration as an exposure surrogate and the true exposure concentrations at the
locations of the study population contribute to exposure error in time-series studies
(Goldman et al.. 2010).
The 2008 SOx ISA (U.S. EPA. 2008d) discussed spatial variability in ambient SO2
concentrations and the impact of this variability on effect estimates from community
time-series epidemiologic studies. Inter-monitor correlations within urban areas ranged
from very low to very high values, suggesting that ambient SO2 concentrations at some
monitors may not be highly correlated with the community average SO2 exposure
concentration. Of particular concern for SO2 is the predominance of point sources,
resulting in an uneven distribution of ambient SO2 concentrations across an urban area.
Factors contributing to differences among monitors include the presence of point sources,
proximity to point sources, terrain features, and uncertainty regarding the measurement of
low ambient SO2 concentrations. The 2008 SOx ISA (U.S. EPA. 2008d) concluded that
low correlation between a specific monitor and the community average ambient SO2
exposure concentration will tend to bias an effect estimate toward the null.
Because ambient SO2 concentrations can have high spatial variability, average SO2
exposure concentration estimates may have less error for populations living close to a
monitor in a community time-series epidemiologic model. Figures 3-1 and 3^2 illustrate
proximity of populations and SO2 monitors to multiple ambient SO2 sources in the
Cleveland and Pittsburgh CBSAs, respectively (discussed in Chapter 2). These CBSAs
were chosen for further discussion here, because they have both high population density
and numerous sources above 2,000 tpy. Figure 3-1 shows the location of fixed-site SO2
monitors and sources with respect to population density for the Cleveland, OH CBSA.
Four of the monitors are centrally located in the urban area, and are also within 10 km of
SO2 sources, while two other monitors are located much closer to point sources (<5 km).
While some densely populated areas are near fixed-site SO2 monitors, some of the
3-35
-------
highest density census block groups are located more than 10-15 km from fixed-site
monitors despite proximity to the sources. Table 3-4 indicates that approximately
two-thirds of the population in various age groups lives within 15 km from a fixed-site
SO2 monitor. For the Pittsburgh CBSA (Figure 3-2). only two of the monitors are located
near sources, with the other monitors distributed among population centers and less
densely populated areas. Here, approximately 40% of the population lives more than
15 km from a fixed-site SO2 monitor (Table 3-5). Such variability in the proximity of
populations to fixed-site monitors suggests that some portions of an urban area may be
subject to increased exposure error. While only minor differences were noted among age
groups in the portion of the population living at specific distances from monitors, the
potential exists for exposure error to differ among other potentially at-risk groups due to
monitor proximity.
Several recent studies have evaluated the impact of spatial variability in ambient SO2
concentration on community time-series health effect estimates. Strickland et al. (2011)
reported a 0-6% lower RR association [fixed-site monitor: RR= 1.009 (0.992, 1.027);
unweighted average: RR= 1.023 (1.006, 1.042); population-weighted average:
RR = 1.020 (1.001, 1.039)] per IQR increases in ambient 1-hour SO2 exposure
concentration (from a fixed-site monitor, unweighted average across monitors, and
population-weighted average) compared with other criteria pollutants in Atlanta, GA. The
authors attributed lower RR to spatial heterogeneity in ambient SO2 exposure
concentrations used as exposure surrogates and the inability of a fixed-site monitor to
capture ambient SO2 plume touchdowns in other parts of the city. The chi-squared
statistic moderately increased when average ambient SO2 exposure concentrations (both
population-weighted and unweighted) from monitors across the city were used. Effect
estimates were higher for the monitor average metrics than for the fixed-site monitor, and
this difference was magnified when effect estimates were based on a standardized
increment rather than the IQR. Because the IQR of the data covered the range of values
observed across the monitors in Atlanta for the Strickland et al. (2011) study, the IQR
partially accounted for spatial variability. The different exposure assignment approaches
only altered the magnitude, not direction, of observed associations.
3-36
-------
Distance to NEI Facilities
G NEI Facility
| 5 km
10 km
| 15 km
# Urban S02 Monitor
2011 ACS Population Per Sq Km Based
On Cleveland CBSA Block Groups
0-507
12.5
25
H
50 Kilometers
508 - 914
| 915-1387
| 1388-1977
| 1978-2895
I 2896-5194
ACS = American Cancer Society; CBSA = core-based statistical area; NEI = National Emissions Inventory; S02 = sulfur dioxide.
Note that the current map projection (GCS-WGS-1984) creates buffers that take on an elliptical shape instead of a circle. The map
projection was chosen to preserve the projection integrity across the data files and reduce error associated with merging data
projections.
Figure 3-1 Map of the Cleveland, OH core-based statistical area including
National Emissions Inventory facility locations, urban sulfur
dioxide monitor locations, and distance to each facility with
respect to core-based statistical area block group population
density estimates for 2011. National Emissions Inventory facility
emissions ranged from 1,942 tons/year to 48,300 tons/year.
3-37
-------
Table 3-4 2011 American Community Survey population estimates of people
living within a specified distance of an urban sulfur dioxide monitor
in the Cleveland, OH core-based statistical area. Population
estimates are based on census block group estimates.
Age Group (yr)
Total Population
Within 1 km
Within 5 km
Within 10 km
Within 15 km
Total
2,080,318
11,816
266,777
759,078
1,310,309
<4
121,820
781
17,608
46,551
75,947
5-17
364,740
1,872
44,719
129,432
222,401
18-64
1,280,478
7,793
178,439
482,808
822,787
>65
313,280
1,370
26,011
100,287
189,174
Source: Data from the 2011 American Community Survey (U.S. Census Bureau. 2011).
3-38
-------
Distance to NEI Facilities
0 NEI Facility |—^—j—
1 I 5 km 0 12.5
| 10 km
I I 15 km
• U rban SO? M on iter
2011 ACS Population Per Sq Km
Based on Pittsburgh CBSA Block Groups
0-391
392 - 958
¦l' 959 -1634
1685 - 2665
| 2666-4462
4463 - 8146
50 Kilometers
y
0 5 10 20 Kilometers
ACS = American Cancer Society; CBSA = core-based statistical area: NEI = National Emissions Inventory; S02 = sulfur dioxide.
Note that the current map projection (GCS-WGS-1984) creates buffers that take on an elliptical shape instead of a circle. The map
projection was chosen to preserve the projection integrity across the data files and reduce error associated with merging data
projections.
The inset map shows National Emissions Inventory facilities located to the southeast of the highly urbanized areas.
Figure 3-2 Map of the Pittsburgh, PA core-based statistical area including
National Emissions Inventory facility locations, urban sulfur
dioxide monitor locations, and distance to each facility with
respect to core-based statistical area block group population
density estimates for 2011. National Emissions Inventory facility
emissions ranged from 1,279 tons/year to 46,467 tons/year.
3-39
-------
Table 3-5 2011 American Community Survey population estimates of people
living within a specified distance of an urban sulfur dioxide monitor
in the Pittsburgh, PA core-based statistical area. Population
estimates are based on census block group estimates.
Total Population
Within 1 km
Within 5 km
Within 10 km
Within 15 km
Population (yr)
2,357,769
64,224
494,382
1,076,465
1,428,871
<4
121,101
2,646
24,748
56,178
73,853
5-17
358,500
8,641
65,882
152,858
211,204
18-64
1,471,310
41,989
325,041
683,445
897,459
>65
406,858
10,948
78,711
183,984
246,355
Source: Data from the 2011 American Community Survey (U.S. Census Bureau. 2011).
High spatial and temporal variability in ambient SO2 concentration leading to a
null-biased effect estimate was also observed in Atlanta by Goldman et al. (2010) when
using 1-h daily max SO2 concentration as an exposure surrogate in a community
time-series epidemiologic study. In this study, the authors used a semivariance analysis
incorporating both spatial and temporal variability to show that secondary pollutants such
as PM2 5 and O3 have lower exposure error (where ambient concentration is a surrogate
for exposure) than primary pollutants such as carbon monoxide (CO) and SO2, for which
concentrations tend to have higher spatial variability than those of secondary pollutants.
Goldman et al. (2010) simulated exposure error as the difference between concentration
measured at the fixed-site monitor and the concentration estimated at a receptor's
location. The study authors computed a semivariance term over distance to the fixed-site
monitor to concentration at a distance from the monitor. The estimated error for SO2 was
then added to a base case scenario, in which the authors assumed that the fixed-site
monitor would produce an accurate exposure. Both the fixed-site monitor estimate and
the estimate at the receptor location were used in epidemiologic models to estimate the
risk ratio for cardiovascular emergency department (ED) visits. The authors estimated
that the risk ratio was biased towards the null by approximately 60% when estimating
exposure using the fixed-site monitor in lieu of estimating exposure at the receptors'
locations. In a related study, Goldman et al. (2012) used different methods to obtain the
surrogate for exposure: fixed-site monitor, unweighted average across monitors,
population-weighted average across monitors, and area-weighted average across
monitors. The bias decreased for 1-h daily max SO2 when using unweighted,
population-weighted, and area-weighted averages of concentrations from multiple
3-40
-------
monitors for the exposure estimate compared with using concentration from a fixed-site
monitor for the exposure estimate. Similarly, epidemiologic studies in the U.S. (Kumar.
2012; Morello-Frosch et al.. 2010) and Australia (Jalaludin et al.. 2007) found higher
associations between ambient SO2 concentrations (used as exposure surrogates) and birth
outcomes when the analysis was restricted to mothers matched with an ambient SO2
monitor within 3-5 km of their residence, suggesting bias towards the null remained in
the spatial averages used in the base case (Section 5.4).
3.4.2.3 Temporal Variability
The influence of plume dynamics on human exposures is important for considering
results of time-series studies of ambient SO2 exposure. As described in Section 2.5.4.
peak concentrations within the ambient SO2 plume can exceed concentrations averaged
over an hour by up to a factor of five; for the observations made in this assessment, the
peak was observed to exceed the mean by up to a factor of 5.5. Hence, SO2 fixed-site
monitoring with averaging times of 1 hour or 1 day, commonly used in time-series
epidemiologic studies as an exposure metric (Chapter 5). may fail to characterize the
variability and peak SO2 exposure concentrations associated with a meandering plume,
resulting in exposure error. Moreover, controlled human exposure studies have
demonstrated health effects at 5-minute time scales (Chapter 5). The longer averaging
times used in epidemiologic studies may be misaligned with the critical time window of
the health effect corresponding to peak SO2 exposure.
Most of the community time-series epidemiologic studies on the health effects of ambient
SO2 exposure described in Chapter 5 use 24-h avg concentration as a surrogate for
exposure. Correlations among different temporal aggregations (1-h avg vs. 5-minute
hourly max, 24-h avg vs. 1-h daily max, and 24-h avg vs. 5-minute daily max) were
computed from the Air Quality System (AQS) data presented in Section 2.5.4 to provide
an indication of how well the 24-h avg represents the 1-h daily max and 5-minute daily
max measures that correspond to peak SO2 plume exposure (Figure 3-3). Approximately
75% of correlations between 1-h avg and 5-minute hourly max were above 0.9.
Correlations between 24-h avg and 1-h daily max were slightly lower, with roughly 75%
of the data having correlations above 0.75. A larger range of data was observed for the
correlations between 24-h avg and 5-minute daily max, with 75% of the data having
correlations above 0.60 and more than 50% of the data having correlations above 0.70.
These moderate-high correlations suggest that 24-h avg data used in many time-series
epidemiologic studies capture the peak exposure reasonably well, but exceptions may be
found for specific sites, as suggested by the lower outliers (r < 0.35) and lower whisker
(r < 0.6) of the correlation between 24-h avg and 5-minute daily max data.
3-41
-------
00
o
to
d
d
-------
areas. Guavetal. (2011) also found that coverage of the study population reduced from
40% for the fixed-time exposure assignments, to 31% when averaging fixed-time
exposure assignments with exposure assignments based on census subdivision, to 29%
when assigning exposures based only on census subdivision, suggesting that improved
spatial and temporal resolution in long-term studies may come at the expense of data
completeness. It should also be recognized that there may be differences in population
mobility between the U.S. and Canada that could influence these findings.
3.4.2.4 Method Detection Limit, Instrument Accuracy, and Instrument Precision
Personal SO2 exposure measurements with ambient SO2 concentration typically have
correlations of 0.4 < r < 0.9 when personal SO2 exposure measurements are above the
method detection limit (MDL). However, the magnitude of personal SO2 exposure
measurements is often much lower than the magnitude of ambient SO2 concentrations
[Section 3.4.1.3; (U.S. EPA. 2008dVI. Moderate to high correlation indicates that using
ambient concentration as a surrogate for personal exposure captures the variability
needed for epidemiologic studies, particularly for time-series and panel studies. Low
personal-ambient correlations reported in the literature are strongly influenced by low
personal exposures relative to the detection limits of personal samplers. When this
happens, personal samplers are unable to provide a signal to correlate with variations in
ambient concentration. Low correlations (r < 0.4) in situations with a high proportion of
samples below the detection limit should not be interpreted as evidence for the lack of a
relationship between personal exposure and ambient SO2 concentrations. Instead, a low
personal sample value likely represents a true low exposure and thus appropriately leads
to a low personal:ambient ratio. Low personal:ambient ratios may be due to low
penetration and high deposition of SO2 in indoor microenvironments where people spend
most of their time. In a study of personal: ambient exposure ratios by Brown et al. (2009).
the authors cited personal SO2 samples below MDL and extremely low SO2 levels to
rationalize not pursuing further analysis.
Instrument error occurs when the measured SO2 concentrations are subject to
interferences that cause biases or noise leading to error in estimating exposure. Ambient
SO2 concentrations measured by FRM or FEM are subject to positive bias from the
detection of interfering compounds. See Section 2.4.1.2 for details on errors that affect
FRMs and FEMs used for fixed-site monitoring. Inter-monitor comparison is often used
to estimate instrument precision. Goldman et al. (2010) used a simulation to investigate
the influence of instrument precision error at locations where ambient SO2 fixed-site
monitors were collocated. Instrument precision error increased with increasing ambient
concentration for the fixed-site monitors. When instrument error and ambient SO2
3-43
-------
concentration were correlated, error was larger in locations with more prevalent or
stronger sources or at times when SO2 emissions were higher for a given location. For
example, the magnitude of the instrument error was expected to be largest when SO2
emissions were highest, such as during peak energy usage times. Instrument error was
also observed to exhibit some autocorrelation at 1- and 2-day lags in the Goldman et al.
(2010) simulation. Hence, the diurnal variability in relative SO2 instrument error does not
change substantially from day to day. For epidemiologic studies of short-term SO2
exposure that use fixed-site-monitored ambient SO2 concentration as a surrogate for
exposure, instrument error would not be expected to influence the exposure surrogate on
a daily basis. When comparing health effect estimates among cities for an epidemiologic
study of long-term SO2 exposure, differences in instrument error among cities could lead
to biased exposure surrogates, given the reliance on differences in magnitude of the
exposure surrogate to study spatial contrasts. Section 3.4.4 describes the influence of
instrument error and high MDL on exposure error and health effect estimates for
community time-series (Section 3.4.4.1). long-term average (Section 3.4.4.2). and panel
(Section 3.4.4.3) epidemiologic studies.
3.4.3 Copollutant Relationships
Confounding is described in the Preamble to the IS As (U.S. EPA. 2015b). Briefly,
confounding occurs when the copollutant exposure concentrations are correlated with
those of the pollutant of interest and the health effect. Confounding can cause misleading
results for estimating the health effect of SO2 if the study does not account for the
copollutant (Rothman and Greenland. 1998). This differs from effect modification, where
the health effect estimate for SO2 is conditional upon the copollutant exposure
concentration via interaction of the SO2 and copollutant exposures.
Health effect estimates derived from both time-series and long-term average
epidemiologic studies are subject to confounding if the health effect model does not
account for copollutant correlation. Simulations by Zeger et al. (2000) indicate that
unaccounted correlation among exposure concentrations or exposure errors for
copollutants may lead to bias and uncertainty in the health effect estimates obtained in
community time-series epidemiologic studies using fixed-site monitoring data. Positive
correlation among copollutant exposure concentrations was shown to amplify the health
effect estimates in Zeger etal. (2000). while negative correlation attenuated the health
effect estimates. Correlation of the errors in measuring copollutant concentrations may
add bias to the health effect estimate, especially when one is measured with more error
than the other (Zeger et al.. 2000). Similarly, simulations of spatial confounding among
correlated copollutants for a long-term average epidemiologic study showed that the
3-44
-------
health effect estimate was attenuated when copollutants were omitted from the model
(Paciorck. 2010). In some cases, this could promote a false conclusion about the strength
of an association between a health effect and the copollutant exposure concentrations.
To assess the independent health effects of ambient SO2 exposure in an epidemiologic
study, it is necessary to identify (1) measurement error for copollutants; (2) which
copollutants [e.g., NO2, PM2.5, ultrafine particulate matter (UFP), BC] are potential
confounders of the health effect-S02 relationship so that their correlation and collinearity
with SO2 can be tested and, if needed, accounted for in the epidemiologic model;
(3) appropriate time lags for SO2 and copollutants; and (4) the spatial correlation
structure across multiple pollutants, if the epidemiologic study design is for long-term
exposure (Paciorck. 2010; Bateson et al.. 2007). Additionally, confounding can also vary
by the health endpoint studied.
In many cases, correlation of SO2 with copollutants, such as PM2 5 and O3, is sufficiently
low to assume that any health effects identified in SO2 models are independent of other
air pollutants. However, when SO2 and a copollutant are correlated (for example, SO2 and
NO2 are often moderately correlated, Section 5.2 and Figure 3-8). copollutant
epidemiologic models may be used to adjust the SO2 effect estimate for potential
confounding by the copollutant (Tolbcrt et al.. 2007). Two-pollutant models can help
identify which is the better predictor of the effect, particularly if the etiologically linked
pollutant is measured with more error than the other pollutant (Zcgcr et al.. 2000).
However, collinearity potentially affects the epidemiologic model's effect estimate when
highly correlated pollutants are modeled simultaneously, and differences in the spatial
distribution of ambient SO2 concentration and the copollutants' ambient concentrations
may also complicate model interpretation [Section 5.1.2.1 and Gryparis et al. (2007)1.
Because ambient SO2 exhibits a relatively high degree of exposure error compared with
other criteria pollutants [e.g., Section 3.4.4.l;Goldman et al. (2010)1. time-series studies
using two-pollutant models in which the SO2 effect estimate remains robust may provide
additional support for a health effect to be associated with SO2 exposure [e.g., Ito et al.
(2007)1.
This section considers temporal copollutant correlations and how relationships among
copollutants may change in space using AQS data and data reported in the epidemiologic
literature (Chapter 5). Temporal copollutant correlations are computed from the time
series of ambient concentrations for two copollutants measured with collocated AQS
monitors. Spatial relationships are computed from the set of ambient concentrations
across space for two copollutants at a point in time. The following sections review
coexposures that can potentially confound the relationship between a health effect and
ambient SO2 exposure over different temporal and spatial resolutions.
3-45
-------
3.4.3.1 Temporal Relationships among Ambient Sulfur Dioxide and Copollutant
Exposures
Short-Term Temporal Correlations
Short-term copollutant correlations were studied using collocated air quality data reported
within the U.S. EPA AQS repository system during 2013-2015. Of the sites reporting
SO2 data to AQS, 438 sites met the 75% data completeness criteria presented in
Section 2.5.1. Daily air quality metrics representing either 1-h daily max or 24-h avg
ambient SO2 concentration values were used. Pearson correlations were used to evaluate
temporal correlations among ambient SO2 concentrations and NAAQS copollutant
concentrations at the same location. In addition, correlations between ambient SO2 and
PM2 5-sulfur were examined because PM2 5-sulfur serves as a surrogate for SO2 oxidation
products (i.e., sulfate) and may have confounding effects on health outcomes associated
with ambient SO2 exposure. Figures 3-4 and 3^5 display the distribution of correlations
between NAAQS copollutants and SO2 daily metrics (24-h avg, 1-h daily max) for all
data combined, and Figures 3-6 and 3£7 display those copollutant correlations broken
down by season. Because epidemiologic studies may use either daily average or daily
maximum metrics, correlations are presented for both metrics, when available. 1-h daily
max concentrations are used for CO and NO2, while 8-h daily max concentrations are
used for O3.
3-46
-------
CO Max
CO Avg
N02 Max
N02 Avg
03 Max
PM25 S Avg
PM25 Avg
F'MIO Avg
~+
• ** *
+¦
¦+-
>-
*** *
-+-¦
-1.0
-0.8
-O.e -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
CO = carbon monoxide; N02 = nitrogen dioxide; 03 = ozone; PM25 = particulate matter with a nominal aerodynamic diameter less
than or equal to 2.5 |jm; PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm; S = sulfur.
Notes: Shown are the median (red line), mean (green star), and interquartile range (box), 5th and 95th percentile (whiskers) and
extremes (black circles).
Number of monitoring sites by pollutant: S02: 438, CO: 171, N02: 206, 03: 310, PM10: 110, PM2.5: 214, PM2 5 S: 137.
Figure 3-4 Distribution of Pearson correlation coefficients for comparison of
24-h avg sulfur dioxide concentration from the year-round data
set with collocated National Ambient Air Quality Standards
pollutants (and sulfur in PM2.5) from Air Quality System during
2013-2015.
3-47
-------
CO Max -
CO Avg
N02 Max-
N02 Avg ¦
03 Max
PM25 S Avg
PM25 Avg ¦
PM10 Avg
CO = carbon monoxide; N02 = nitrogen dioxide; 03 = ozone; PM25 = particulate matter with a nominal aerodynamic diameter less
than or equal to 2.5 |jm; PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm; S = sulfur.
Notes: Shown are the median (red line), mean (green star), and interquartile range (box), 5th and 95th percentile (whiskers) and
extremes (black circles).
Number of monitoring sites by pollutant: S02: 438, CO: 171, N02: 206, 03: 310, PM10: 110, PM2.5: 214, PM2 5 S: 137.
Figure 3-5 Distribution of Pearson correlation coefficients for comparison of
daily 1-h max sulfur dioxide concentration from the year-round
data set with collocated National Ambient Air Quality Standards
pollutants (and sulfur in PM2.5) from Air Quality System during
2013-2015.
3-48
-------
Winter
Spring
CO Max -
CO Avg
N02 Max -
N02 Avg
03 Max -
PM25 S Avg
PM25 Avg
PM10 Avg
— t
— *
4
*
— *
-+¦
±L
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 O.i
Summer
1.0
CO Max
CO Avg
N02 Max
N02 Avg
03 Max
PM25 S Avg
PM25 Avg
PM10 Avg
-¦¦h-
- u
f-
1 L.
1 IT
—-H ^
1 1"
••
11 L-..
• • •
• * • "1—1 *
ft
.. «|_
H b H- -
• hk:
*• «wj
j- ~
C H •
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
correlation with S02 hourly daily average
* "H—
• • *»|—|
....—
....—
+ -H- -
-J_h+ -
. —
....—
.. —
. ...^-—
* *
i —{¦ . •
» —• •
* —1 •• •
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
Fall
--
-
I>+-
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
correlation with S02 hourly daily average
CO = carbon monoxide; N02 = nitrogen dioxide; 03 = ozone; PM25 = particulate matter with a nominal aerodynamic diameter less
than or equal to 2.5 pm; PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 pm; S = sulfur;
S02 = sulfur dioxide.
Notes: Shown are the median (red line), mean (green star), and interquartile range (box), 5th and 95th percentile (whiskers) and
extremes (black circles).
Number of monitoring sites by pollutant: S02: 438, CO: 171, N02: 206, 03: 310, PM10: 110, PM25: 214, PM25 S: 137.
Figure 3-6 Distribution of Pearson correlation coefficients for comparison of
daily 24-h avg sulfur dioxide ambient concentration stratified by
season with collocated National Ambient Air Quality Standards
pollutants (and PM2.5) from Air Quality System during 2013-2015.
3-49
-------
Winter
Spring
CO Max -
CO Avg
N02 Max -
N02 Avg
03 Max -
PM25 S Avg
PM25 Avg
PM10 Avg
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 O.i
Summer
1.0
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
correlation with S02 hourly daily max
—14
* •
- -b
• "4
* -H-
- 4 -
m ••
¦t • •
* ™n
. —|—
' "b"
• •
• - — H 1* b-+-
• • •
• ••
H f 1—1
w —
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
Fall
CO Max -
CO Avg
** *4—I I* I—
"-HLlb
-j* • — ••
I
rp r
N02 Avg
*•«—)— |
^#*
j
03 Max -
-+»
PM25 S Avg
•HH T 1
—-f • •
PM25 Avg
nra:
•
j
PM10 Avg
- - U )[
: b" •
I
—H-Dl}
..-KI
" +
I
"b
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
correlation with S02 hourly daily max
CO = carbon monoxide; N02 = nitrogen dioxide; 03 = ozone; PM25 = particulate matter with a nominal aerodynamic diameter less
than or equal to 2.5 pm; PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 pm; S = sulfur;
S02 = sulfur dioxide.
Notes: Shown are the median (red line), mean (green star), and interquartile range (box), 5th and 95th percentile (whiskers) and
extremes (black circles).
Number of monitoring sites by pollutant: S02: 438, CO: 171, N02: 206, 03: 310, PM10: 110, PM25: 214, PM25 S: 137.
Figure 3-7 Distribution of Pearson correlation coefficients for comparison of
daily 1-h max sulfur dioxide ambient concentration stratified by
season with collocated National Ambient Air Quality Standards
pollutants (and PM2.5) from Air Quality System during 2013-2015.
While 24-h avg ambient SO2 concentration exhibits a wide range of correlations with
NAAQS copollutants, median correlations are all below 0.4 (Figure 3-4). The lowest
correlations are observed between ambient SO2 concentration and ambient O3
concentration, with median correlations below 0.1. Slightly higher correlations are
observed between ambient SO2 concentration and other primary NAAQS pollutant
concentrations (NO2 and CO), with median correlations between 0.3 and 0.4. Common
fuel combustion sources may be responsible for these correlations (Section 2.2). Lower
correlations with PM2.5 sulfur than PM2.5 mass may reflect the secondary formation of
3-50
-------
sulfate by oxidation of SO2, while PM2 5 mass also has a primary component.
Correlations close to 1 or below 0 are sometimes observed but only occur at a few outlier
monitoring sites. Comparatively, copollutant correlations of daily 1-h max ambient SO2
in Figure 3-5 are also slightly lower than the copollutant correlations based on ambient
SO2 24-h avg values in Figure 3-4. The medians of correlations between daily 1-h max
ambient SO2 concentrations and other NAAQS pollutants are below 0.3, with the
exception of NO2, which exhibits median correlations slightly above 0.3. Less than 5% of
the data had correlations above R = 0.7. Higher correlations may introduce a greater
degree of confounding into results of short-term epidemiologic studies. It is notable that
the nature of correlations between SO2 and copollutants is changing given rulemaking on
use of ultra-low sulfur diesel fuel that went into effect in 2006 (66 FR 5002). Some of the
epidemiologic studies cited in Chapter 5 included data obtained before 2006 and 2007,
when the new sulfur standards took effect for highway vehicles and heavy-duty vehicles,
respectively. Between 2004 and 2014, SO2 from highway emissions decreased by 86%
(Table 2-1). Because on-road vehicles are the largest source for ambient NOx (U.S. EPA.
2016d') and CO (U.S. EPA. 2010^ and contribute to ambient organic carbon (OC) and
elemental carbon (EC) (U.S. EPA. 2009a). the new sulfur standards may have contributed
to the wider variation observed in correlation between ambient SO2 and copollutant
concentrations. Note that potential for confounding also varies by health endpoint.
Correlations between ambient SO2 and NAAQS copollutant concentrations demonstrate
very little variability across seasons (Figures 3-6 and 3-7). All median and average
copollutant correlations are below 0.4 across every season. The only substantial seasonal
difference in correlations between ambient SO2 and copollutant concentrations occurs
during the winter, when ambient SO2 concentration exhibits lower negative correlations
with ambient O3 concentration (median winter correlations = -0.1). SO2-O3 correlations
are generally low year-round, potentially because the regional nature of O3 formation
contrasts with the local nature of SO2 plumes from point sources. In winter, the low
correlations could be directly linked to relatively low ambient O3 concentrations during
this time of year due to less photochemical O3 production and SO2 oxidation.
Daily and hourly ambient SO2 concentrations generally exhibit median correlations
around 0.2-0.4 with respect to other collocated NAAQS copollutants at AQS monitoring
sites. However, a small subset of sites report relatively higher copollutant correlations.
Tests to identify potential confounding in epidemiologic models may need to be
performed if high copollutant correlations are reported in the individual studies. High
copollutant correlations in the national distribution could be due either to consistently
similar concentrations for both SO2 and the copollutant or to consistent fluctuations in
concentrations of both pollutants due to source behavior and meteorology.
3-51
-------
Exposure studies have also examined correlations between ambient SO2 concentration
and ambient or personal copollutant exposure concentrations, generally reporting low or
moderate correlations. For SO2, within-hourly concentrations have median correlations
around 0.2 for most PM of different cutpoints and species. For gases, median correlations
of within-hourly data were lower for O3 than for CO and NO2, respectively, but median
correlations did not surpass 0.4. Correlations were mostly positive for all but O3, which
exhibits both negative and positive correlations. See Figure 3-8 and references cited
therein for copollutant correlation data reported in the literature cited in (U.S. EPA.
2017b).
More data were available for within-daily correlations of SO2 and copollutant exposure
concentrations. Median correlation around 0.5 were observed for sulfate, nitrate, BS, and
OC PM2 5 species, PM10, and NO2 for the within daily time-scale. Median correlation was
around 0.3 for particulate matter with a nominal aerodynamic diameter less than or equal
to 10 (mi and greater than a nominal 2.5 |im (PM10-2.5), around 0.4 for CO and PM2 5, and
around -0.2 for O3. Both data availability and inter-site variability were much greater for
the gases, PM2 5, and PM10 compared with the individual PM2 5 species or PM10-2.5.
Where data were available, a large degree of scatter was evident in the data. In studies
where within-daily correlations of SO2 exposure concentrations with NO2 and CO
exposure concentrations were observed to be high, it is possible the data were collected
before the rulemaking to reduce sulfur content in diesel fuel went into effect in 2006
(66 FR 5002) or when coal was in greater use in energy generation (Section 2.2). The
minority of sites with stronger correlations have the potential to reflect a greater degree of
confounding into the epidemiologic results if the copollutant correlations at those sites
are similar to the copollutant correlations experienced at the locations of exposure. It is
possible that the observed correlation at a single site may not reflect copollutant
correlations at the sites of exposure, particularly in areas with a large amount of spatial
heterogeneity of SO2.
Data for correlations between ambient SO2 concentrations and personal copollutant
exposures were reported in the 2008 SOx ISA (U.S. EPA. 2008d). and no studies have
been produced to substantiate or revise the observations reported at that time.
Between-subject correlations of daily ambient SO2 concentration with personal PM2 5
exposures were found to vary widely with positive and negative correlations in the Sarnat
et al. (2005) and Sarnat et al. (2001) studies. In the Sarnat et al. (2005) study, 95-97% of
the SO2 data were below the MDL, indicating high uncertainty. This evidence suggests
that correlations between personal copollutant exposures and ambient SO2 concentration
vary among individuals, and thus the potential for copollutant confounding cannot be
ruled in or out.
3-52
-------
Within-Hourly Correlations
Within-Daily Correlations
UFP
SO[4]
NO[3]
BS
EC
OC
PM[2.5]
PM[10-2.5]
PM[10]
Particulate Matter
^ I 0 f'
I 0
fr*
o
^0+
O
I
Within-Monthly Correlations
Particulate Matter
UFP -
Sulfate
Nitrate
PM 2.5
PM 10-2.5
PM 10
UFP
Sulfate
Nitrate
BS
EC
OC
PM[2.5]
PM[10-2.5]
PM[10]
0(3]
CO
NO[2]
Particulate Matter
•fit
4 I M
rn
:¦ ® • 'J.'i
C' *
rrtn | i*> f- - at
O
I
Long-Term Correlations
Particulate Matter
UFP -
Sulfate -
C I 3
Nitrate -
I
BS
~
0
EC
ili
OC -
I
PM[2.5] -
-e- cjo a> -o- - $
PM[10-2.5]
I
PM[10] -
Gases
0(3]
4** -©•<
BciiDcj^aoo oj OcjpM^ «B -3 -
O
I
BS = black smoke; CO = carbon monoxide; EC = elemental carbon; LUR = land use regression; NO[2] = nitrogen dioxide;
NO[3] = nitrate ion; 0[3] = ozone; OC = organic carbon; PM[2.5] = in general terms, particulate matter with a nominal aerodynamic
diameter less than or equal to 2.5 pm, a measure of fine particles; PM[10] = in general terms, particulate matter with a nominal
aerodynamic diameter less than or equal to 10 |jm, a measure of thoracic particles (i.e., that subset of inhalable particles thought
small enough to penetrate beyond the larynx into the thoracic region of the respiratory tract); PM[10-2.5] = in general terms,
particulate matter with a nominal aerodynamic diameter less than or equal to 10 pm and greater than 2.5 pm, a measure of thoracic
coarse particulate matter or the coarse fraction of PM,0; SO[2] = sulfur dioxide; SO[4] = sulfate; UFP = ultrafine particulate matter.
Notes: Each data point in each boxplot represents the correlation between S02 and copollutants. Boxes represent the interquartile
range of the data with the median line plotted, and 90th and 10th percentile of the data are plotted as the whiskers. Correlation data
computed from LUR studies are not included here. Correlations shown by open black circles either come from urban-regional scale
studies or do not specify the study's spatial scale. Within-monthly correlations include correlations obtained over 5 weeks or less for
S02. Long-term correlations refers to correlations obtained over a period longer than 5 weeks.
Sources of data for this figure are listed in (U.S. EPA. 2017b).
Figure 3-8 Summary of temporal sulfur dioxide-copollutant correlation
coefficients from measurements reported in the literature, sorted
by temporal averaging period.
3-53
-------
Long-Term Correlations
Long-term epidemiologic studies that have reported copollutant correlations are also
displayed in Figure 3-8 and references cited therein for correlations developed from data
averaged over periods longer than 5 weeks. Data were limited for many of the PM2 5
components. For exposure concentrations of PM2 5, PM10, O3, CO, and NO2, a wide range
of correlations has been estimated. Median correlation was lower for PM2 5 exposure
concentration (r = 0.2) compared with that of PM10 (r = 0.4), CO (r = 0.3), and NO2
(r = 0.3). Median correlation was negative (r = -0.3) for O3 exposure concentration. For
correlations between exposure concentrations of SO2 and PM2 5, most of the data were
clustered around the median, while variability in the correlations was larger for the other
copollutants. As for short-term copollutant relationships, no clear conclusion can be
drawn regarding the potential for confounding of long-term SO2 epidemiologic estimates
by copollutants. Wide variability in copollutant correlations was observed, with the
highest correlations around 0.7-0.8 for PM2 5, PM10, CO, and NO2. For the studies with
high copollutant correlations, health effect estimates may be inflated by copollutant
confounding (Zeger et al.. 2000). Copollutant correlations are reported for tabulated
epidemiologic studies throughout Chapter 5 to illustrate where copollutant confounding
may influence health effect estimates, and two-pollutant models designed to test for
copollutant confounding are presented where available (Section 5.2.2).
3.4.3.2 Spatial Relationships among Ambient Sulfur Dioxide and Copollutants
Spatial confounding can potentially bias health effect estimates in epidemiologic studies
of long-term SO2 exposure. Paciorek (2010) performed simulations to test the effect of
spatial confounding on health effect estimates in long-term exposure epidemiologic
studies. He identified unmeasured spatial confounding as a key driver in biasing health
effect estimates in a spatial regression towards the null. Paciorek (2010) maintained that
bias in the health effect estimate is lower when variation in the exposure metric occurs at
a smaller spatial scale than that of the unmeasured confounder compared with bias in the
health effect estimate when the spatial scale of the exposure metric is larger than that of
the unmeasured confounder.
3.4.4 Implications for Epidemiologic Studies of Different Designs
Exposure error is defined in Section 3.2.1. To summarize, exposure error refers to the
bias and uncertainty associated with using concentration to represent the actual exposure
of an individual or population. Exposure error has two components: (1) uncertainty in the
3-54
-------
metric used to represent exposure concentration and (2) the difference between the
exposure surrogate in the epidemiologic study and the true exposure (which may not be
observable) (Zcgcr et al.. 2000). Classical error is the component of exposure
measurement error derived from uncertainty in the metric being used to represent
exposure concentration. Classical error is defined as error scattered around the true
personal exposure and independent of the measured exposure concentration. Classical
error results in bias of the epidemiologic health effect estimate that is typically towards
the null (no effect of the exposure). Classical error can also cause inflation or reduction of
the standard error of the health effect estimate. Berkson error is the component of
exposure error related to the use of an exposure surrogate in the epidemiologic study in
lieu of the true exposure. Berkson error is defined as the unobserved portion of the true
exposure, and it is independent of the observed portion of the true exposure (Goldman et
al.. 2011; Armstrong. 2008; Reeves et al.. 1998). Pure Berkson error generally does not
bias the health effect estimate.
When investigators use statistical models to predict exposure concentrations, the
exposure error is no longer purely classical or purely Berkson and may have
characteristics of each error type. Measurement error for modeled exposure
concentrations has been decomposed into Berkson-like and classical-like components,
sharing some characteristics with pure Berkson and classical errors, respectively, but with
key differences (Szpiro etal.. 2011). Berkson-like errors tend to occur when the modeled
exposure concentration does not capture all of the variability in the true exposure. Under
ideal conditions, Berkson-like error increases the variability around the health effect
estimate in a manner similar to pure Berkson error and does not induce bias, but
Berkson-like error is spatially correlated and not independent of predicted exposure
concentrations, so it results in underestimation of standard errors (Goldman et al.. 2011;
Szpiro etal.. 2011). Szpiro and Paciorek (2013) analyzed the impact of Berkson-like
error under more general conditions and found that it can bias health effect estimates
either toward the null or away from the null. For example, in one simulation study in
which the spatial distributions of monitor and subject locations were dramatically
different, the health effect estimates were biased away from the null. In another example,
where spatially structured covariates were included in the health model but not in the
exposure model, the health effect estimates were biased toward the null. Hence,
Berkson-like error can lead to bias of the health effect estimate in either direction and
should not be ignored. Classical-like errors result from uncertainty in estimating exposure
model parameters. It can add variability to predicted exposure concentrations and can
bias health effect estimates in a manner similar to pure classical error, but it differs from
pure classical error in that the additional variability in estimated exposure concentrations
is also not independent across space. Exposure error can bias epidemiologic associations
between ambient pollutant concentrations and health outcomes, compared with the effect
3-55
-------
estimate obtained using the true exposure, and it tends to widen confidence intervals
around those estimates so that nominal coverage of the confidence intervals is below 95%
for exposure effect estimates conditional on mismeasured covariates (Sheppard et al..
2005; Zeger et al.. 2000).
Exposure error can be an important contributor to uncertainty and variability in
epidemiologic study results. The importance of exposure error varies with study design
and is dependent on the spatial and temporal aspects of the design. Factors that could
influence estimates of exposure to ambient SO2 include time-activity patterns of the study
population, topography of the natural and built environment, meteorology, instrument
errors, use of ambient SO2 concentration as a surrogate for exposure to ambient SO2, and
the presence of SO2 in a mixture of pollutants. The following sections will consider
various sources of error and how they affect the interpretation of results from
epidemiologic studies of different designs.
3.4.4.1 Community Time-Series Studies
In most short-term exposure epidemiologic studies of the health effects of SO2, the health
effect endpoint is modeled as a function of ambient exposure, A',. which is defined as the
product of Ca, and a, a term encompassing time-weighted averaging of exposure
concentration and infiltration of SO2 (Section 3.2.2). Community time-series
epidemiologic studies capturing the exposures and health outcomes of a large cohort
frequently use the ambient concentration at a fixed-site monitor (Ca,f) as a surrogate for A',
in an epidemiologic model (Wilson et al.. 2000). At times, an average of
fixed-site-monitored concentrations is used for the A', surrogate. For studies involving
thousands of participants, it is not feasible to measure personal exposure concentrations
or time-activity patterns. Moreover, for community time-series epidemiology studies of
short-term exposure, the temporal variability in ambient SO2 concentration is of primary
importance to relate to variability in the health effect estimate (Zeger et al.. 2000). Ca,f
can be an acceptable surrogate if the fixed-site monitor captures the temporal variability
of the actual SO2 exposure concentration. Spatial variability in ambient SO2
concentrations across the study area could attenuate an epidemiologic health effect
estimate if the exposures are not correlated in time with Ca f when fixed-site monitoring is
used to represent exposure in the epidemiologic model. If exposure assessment methods
that more accurately capture spatial variability in the concentration distribution over a
study area are employed, then the confidence intervals around the health effect estimate
may decrease. Ca, f may be an acceptable surrogate for A', if the concentration time series
at the fixed-site monitor is correlated in time with the exposures.
3-56
-------
In a time-series study of ED visits for cardiovascular disease, Goldman et al. (2011)
simulated the effect of classical and Berkson errors due to spatiotemporal variability
among ambient air pollutant concentrations over a large urban area. For 1-h daily max
SO2, the RR per ppm was negatively biased in the case of classical error (-1.3%) and
negligibly positively biased in the case of Berkson error (0.0042%). The 95% confidence
interval range for RR per ppm was wider for Berkson error (0.028) compared with
classical error (0.0025).
Recent studies have explored the effect of spatial exposure error on health effect
estimates to test the appropriateness of using fixed-site monitoring for time-series studies.
Goldman et al. (2010) simulated spatial exposure error based on a semivariogram
function across monitor sites with and without temporal autocorrelation at 1- and 2-day
lags to analyze the influence of spatiotemporal variability among ambient concentrations
over a large urban area on a time-series study of ED visits for cardiovascular disease. A
semivariogram estimated the change in spatial variability in exposure concentration with
distance from the monitoring site. The average of the calculated random term was added
to an ambient fixed-site monitoring SO2 concentration time series (designated in this
study to be the base case) to estimate SO2 population exposure concentration subject to
spatial error. For the analysis with temporal autocorrelation accounted for, RR per ppm
for 1-h daily max SO2 dropped slightly to 1.0045 (95% CI: 1.0023, 1.0065) when it was
compared with the fixed-site monitor RR per ppm = 1.0139 (for all air pollutants).1 When
the model did not account for temporal autocorrelation, RR per ppm dropped very
slightly to 1.0042 for 1-h daily max SO2. The results of Goldman et al. (2010) suggest
that spatial exposure error from the use of ambient fixed-site SO2 concentration
monitoring data results in biasing the health effect estimate towards the null, but the
magnitude of the change in effect was small.
In another simulation study analyzing the influence of spatiotemporal variability among
ambient concentrations over a large urban area on health effect estimates, Goldman et al.
(2012) evaluated the effect of different types of spatial averaging on bias in the health
effect risk ratio and the effect of correlation between measured and reference ambient
concentrations of SO2 and other air pollutants. Ambient concentrations were simulated at
alternate monitoring locations using the geostatistical approach described above
(Goldman et al.. 2010) for the 20-county Atlanta metropolitan area for comparison with
ambient concentration measurements obtained directly from monitors at those sites.
Geostatistical-simulated ambient exposure concentrations were designated as the
reference in this study, and other exposure assessment methods were assumed to have
1 Note that 95% CIs were not reported for the fixed-site monitor RR or for the cases where temporal autocorrelation
was not modeled.
3-57
-------
some error. Five different exposure assessment approaches were tested: (1) using a single
fixed-site monitor, (2) averaging the simulated exposures across all monitoring sites,
(3) performing a population-weighted average across all monitoring sites, (4) performing
an area-weighted average across all monitoring sites, and (5) performing
population-weighted averaging of the geostatistical simulation. Goldman et al. (2012)
observed that the exposure error was somewhat correlated with both the measured
exposure concentration and the reference ambient concentrations, reflecting both Berkson
and classical error components. For the fixed-site monitor, the exposure errors were
somewhat inversely correlated with the exposure concentration reference value but had
relatively higher positive correlation with the measured ambient concentration. For the
other exposure estimation methods, the exposure errors were inversely correlated with the
reference exposure concentration, while having positive but lower magnitude correlation
with the measured ambient concentration. Additionally, the exposure bias, given by the
ratio of the exposure error to the measured value, was much higher in magnitude at the
fixed-site monitor than for the spatial averaging techniques for SO2. Hence, compared
with other exposure assessment methods, the health effect estimate would likely have
greater bias towards the null with reduced precision when a fixed-site monitor is used to
measure ambient SO2 concentration as a surrogate for exposure. However, exposure error
is likely to cause some bias and imprecision for other exposure surrogate methods as
well.
In addition to the effect of the correlations and ratios themselves, spatial variation across
urban areas also impacts time-series epidemiologic results. The Goldman et al. (2010)
and Goldman et al. (2012) findings suggest more Berkson error in the spatially resolved
exposure concentration metrics compared with the fixed-site monitor ambient
concentration and more classical error for the fixed-site monitor ambient concentration
estimate compared with the other exposure concentration measurement techniques.
Hence, more bias would be expected for the health effect estimate calculated from the
fixed-site monitor ambient concentration, and more variability would be expected for the
health effect estimate calculated from exposure concentrations estimated by the more
spatially resolved methods. Differences in the magnitude of exposure concentration
estimates are not likely to cause substantial bias, but they tend more to widen confidence
intervals and thus reduce the precision of the effect estimate beyond the nominal
coverage of the confidence intervals that would be obtained if using the true, albeit
unknown, exposure concentration (Zeger et al.. 2000). The more spatially variable air
pollutants studied in Goldman et al. (2012) also had more bias in the health effect
estimates. This occurred across exposure assignment methods but was more pronounced
for the fixed-site measurement ambient concentration data. Note that the Goldman et al.
(2010). Goldman et al. (2011). and Goldman et al. (2012) studies were performed only in
3-58
-------
Atlanta, GA. These simulation studies are informative, but similar simulation studies in
additional cities would aid generalization of these study results.
Section 3.4.2.4 describes the influence of high MDL on the relationship between
measured ambient SO2 concentrations and personal SO2 exposures. When measurements
are above MDL, then the amount of correlation between personal SO2 exposure and
ambient SO2 concentrations may influence the extent of bias in a time-series study. If the
reported values of personal exposure measurements are below MDL, correlation between
personal SO2 exposure measurements and ambient SO2 concentrations will likely be low
due to random noise in the signal. To the extent that true correlations are less than one,
epidemiologic effect estimates based on ambient concentration will be biased toward the
null, based on simulations by Zeger et al. (2000). Time-series epidemiologic studies
employing data below MDL may demonstrate attenuated effect, but this scenario cannot
be used to reject the hypothesis of a health effect.
Section 3.4.2.4 also describes the influence of instrument accuracy and precision on the
relationship between ambient SO2 concentrations and personal SO2 exposures. Exposure
measurement error related to instrument precision has a smaller influence on health effect
estimates in time-series studies compared with error related to spatial gradients in the
ambient SO2 concentration because instrument precision would not be expected to
modify the ability of the instruments to respond to changes in ambient concentration over
time. Goldman et al. (2010) investigated the influence of instrument error on health effect
estimates in a time-series epidemiology study by studying differences in exposure
concentration estimates and health effect estimates obtained using collocated monitors.
Very small changes in the risk ratios were observed for 1-h daily max SO2 ambient
concentrations. For 1-h daily max SO2 ambient concentration, the RR per ppm of SO2
ambient concentration with simulated instrument precision error was 1.0132 compared
with RRper ppm = 1.0139 for the fixed-site monitor. The amount of bias in the health
effect estimate related to instrument precision was very small.
As described in Section 3.4.1. nonambient sources of SO2 are rare. Even in
microenvironments where nonambient SO2 exposure is substantial, such as in a room
with a kerosene heater, such nonambient exposure concentrations are unlikely to be
temporally correlated with ambient SO2 exposure concentrations (Wilson and Suh. 1997).
and therefore, would not affect epidemiologic associations between ambient SO2
exposure concentrations and a health effect in a time-series study. Sheppard et al. (2005)
concluded that nonambient exposure does not influence the health outcome effect
estimate if ambient and nonambient exposure concentrations are independent. Personal
exposure to ambient SO2 is some fraction of the ambient concentration. Therefore, effect
estimates based on personal SO2 exposure rather than ambient SO2 concentration will be
3-59
-------
positively biased in proportion to the ratio of ambient SO2 concentration to ambient SO2
exposure concentration. Daily fluctuations in this ratio can widen the confidence intervals
in the ambient SO2 concentration effect estimate beyond the nominal coverage of the
confidence intervals obtained using the true exposure. Uncorrelated nonambient exposure
concentration will not bias the effect estimate but may also widen the confidence
intervals (Shcppard et al.. 2005; Wilson and Suh. 1997).
3.4.4.2 Long-Term Cohort Studies
For long-term epidemiologic studies, the lack of personal exposure data means that
investigators must rely on fixed-site ambient SO2 concentration data or model estimates
of SO2 exposure concentration. Ambient SO2 concentration data may be used directly,
averaged across counties or other geographic areas, or used to construct geospatial or
regression models to estimate exposure concentrations at unmonitored locations
(Sections 3.3.2.1 through 3.3.2.3V For cohort epidemiologic studies of long-term human
exposure to SO2 where the spatial difference in the magnitude of the ambient SO2
exposure is often of most interest and if Ca,f is used as a surrogate for A'„ then a can
represent the exposure measurement error related to uncertainties in the time-activity data
and air exchange rate. Spatial variability in ambient SO2 exposure concentrations across
the study area could lead to bias in the health effect estimate if Ca,f is not representative of
E&. This could occur, for example, if the study participants were clustered in a location
where their SO2 exposure concentration is higher or lower than the exposure
concentration estimated at a model or measurement site. Ca,f may be an acceptable
surrogate for A', if the fixed-site monitor is located close to the study participants and the
ambient SO2 source (e.g., near the plume touchdown of a power plant) and spatial
variability of the ambient SO2 concentration across the study area where the study
participants are located is minimal in the vicinity of each sample group.
Model surfaces can also be constructed from physics-based models (Sections 3.3.2.4 and
3.3.2.5). but the amount of exposure error depends on model specifications. For example,
Lipfert et al. (2009) estimated ambient SO2 concentration based on the CMAQ model for
use as an exposure surrogate. Annual average ambient SO2 concentrations were estimated
with a 36 by 36-km grid across the contiguous U.S. The modeled ambient SO2
concentrations were used as exposure surrogates to determine their association with
county-level mortality data for the Washington University-Electric Power Research
Institute Veterans Cohort Mortality Study. To assign exposures at the county level, the
CMAQ grid that included the largest city within each county was determined, and the
associated CMAQ ambient SO2 concentration was used as a surrogate for exposure
concentration for the entire county. Lipfert et al. (2009) did not provide data to validate
3-60
-------
the model results with measurements, but they found that the magnitude of the mortality
risk coefficients was lower for modeling results compared with that obtained by
measurements (risk coefficient for measurement: -0.047, risk coefficient for model:
-0.012), suggesting that averaging over the 36-km grid may have caused attenuation of
the effect estimate.
The number of recent long-term studies of SO2 exposure that permit evaluation of the
relationship between long-term average ambient SO2 concentrations and personal or
population exposures is limited, and the value of short-term exposure concentration data
for evaluating long-term exposure concentration relationships is uncertain. If the longer
averaging time (annual vs. daily or hourly) smoothes out short-term fluctuations,
long-term ambient SO2 concentrations employed in long-term epidemiologic studies may
be well correlated with true long-term exposures. For example, Guav et al. (2011)
observed high correlation between single-year/single-location ambient SO2
concentrations used for an exposure surrogate with ambient SO2 concentrations averaged
over a 22-year period when the annual SO2 concentrations were assigned based on the
study participants' census subdivision. However, lower correlation between long-term
exposure and ambient SO2 concentration could occur if important exposure determinants
change over a period of several years, including activity pattern and residential air
exchange rate.
Minimization of error in the exposure concentration prediction does not always minimize
error in the health effect estimate. Szpiro etal. (2011) used simulation studies to evaluate
the bias and uncertainty of the health effect estimate obtained when using correctly
specified and misspecified long-term exposure concentration models. The correct
exposure concentration model was designated by the authors to be an LUR with three
covariates while the misspecified model included only two of these three covariates. The
study authors estimated the exposure concentration model parameters using monitor data
and predicted exposure concentrations at subject locations. They studied two conditions:
where the variation in the third covariate was identical in the monitor and subject data
versus where it was much smaller in the monitor data than in the subject data. Szpiro et
al. (2011) showed that prediction accuracy of the exposure concentration estimate was
always higher for the correctly specified model compared with the misspecified model.
The health effect estimate had lower RMSE for the correct model when the third
covariate had identical variability in the monitor and subject data. However, when the
third covariate was much less variable in the monitor data, then the health effect estimate
had lower RMSE for the misspecified model. The results of the Szpiro et al. (2011)
simulations demonstrate one situation where use of a more accurately defined exposure
concentration metric does not improve the health effect estimate.
3-61
-------
Error correction is a relatively new approach to estimate the correct standard error and to
potentially correct for bias in air pollution cohort studies. Szpiro and Paciorek (2013)
established that two conditions must hold for the health effect estimate to be predicted
correctly: (1) the exposure concentration estimates from monitors must come from the
same underlying distribution as the true exposure concentrations and (2) the health effect
model includes all covariates relevant to the population. Szpiro and Paciorek (2013) and
Bergen and Szpiro (2015) developed methods to correct for bias from classical-like
measurement error. Valid standard error estimates are obtained by means of the
nonparametric bootstrap. Methods have also been proposed to correct for bias from
Berkson-like error, but these require stronger conditions, including compatibility between
subject and monitor locations and inclusion of spatially structured health model
covariates in the exposure concentration model.
In the Szpiro and Paciorek (2013) study, when the assigned exposure concentration
measurements were set to be uniform across space, the health effect estimate was biased
away from the null with different standard error compared with the case when the
exposure subjects were collocated with the study participants. When an additional spatial
covariate was omitted, the health effect estimate was biased towards the null with
different standard errors compared with the correctly specified model. Bias correction
and bootstrap calculation of the standard errors reduced bias in the model prediction.
Furthermore, bias correction with bootstrapped simulation of standard error improved the
confidence interval obtained from the simulation. With no correction or bootstrapping,
the standard errors and confidence interval were underestimated. Bias correction with
bootstrapped simulation of standard errors produced standard errors and confidence
intervals close to the true standard errors and confidence intervals. These findings imply
that without bias correction, effect estimates would be biased with standard errors that
underestimate the true standard error. None of the epidemiologic studies cited in
Chapter 5 applied bias correction or standard error bootstrapping. The findings of Szpiro
and Paciorek (2013) suggest that bias away from the null may occur when spatial
resolution of the exposure model is insufficient, while bias towards the null may occur
when the model is misspecified. Without bootstrapped simulation of the standard error, it
is possible that the confidence intervals around the effect estimates seem more precise
than they actually are. Spiegelman (2013) noted that the new measurement error
correction methods developed by Szpiro and Paciorek (2013) are a version of regression
calibration. This study illustrated the influence of classical-like and Berkson-like errors
on long-term exposure cohort study health effect estimates through these simulations.
Instrumentation bias could be expected to influence health effect estimates from
epidemiologic studies of long-term SO2 exposures in some situations. Section 2.4.1
describes how the presence of copollutants can cause ambient SO2 concentrations
3-62
-------
measured using fixed-site monitors to be overestimated and how high relative humidity
can cause ambient SO2 concentration measurements to be underestimated. Relative
humidity would not be expected to vary greatly within a city. However, local ambient
copollutant concentrations may be spatially variable such that failure to account for
differences in measurement errors could lead to some differential bias in health effect
estimates across a city related to instrument error. Because climate and ambient sources
are more likely to differ among cities, instrumentation error could have a larger influence
on the comparison of health effect estimates among cities when fixed-site monitors are
used to estimate exposure concentrations.
3.4.4.3 Panel Studies
Panel or small-scale cohort studies involving dozens of individuals (including some
studies cited in Sections 5.2.2.2 and 5.2.2.3) may use more individualized exposure
concentration measurements, including personal exposures, residential indoor or outdoor
concentration measurements, or concentration data from local study-specific monitors.
Modeled concentrations are typically not used as exposure surrogates in panel
epidemiologic studies. A main disadvantage of the modeling approach is that the results
of modeling exposure concentration must be compared to an independent set of measured
exposure concentration levels (Klepeis. 1999). In addition, a modeling approach requires
resource-intensive development of validated and representative model inputs, such as
human activity patterns, distributions of AER, and deposition rate. Therefore, modeled
exposure concentrations are used infrequently in panel epidemiologic studies.
Section 3.4.2.4 describes the influence of high MDL on the relationship between
measured ambient SO2 concentrations and personal exposures for ambient SO2. Personal
exposure measurements below MDL will likely cause the correlation between personal
exposure measurements and ambient SO2 concentrations to be low due to random noise
in the signal. Noise in the exposure signal would add noise to the health effect estimate in
a panel epidemiologic study as well. Below MDL measurements would be unlikely to
bias the effect estimate, however, because the magnitude of exposure would be low
whether measured with a high-precision or low-precision device.
It is also possible that the ratio of personal SO2 exposure to ambient SO2 concentration in
panel studies is low due to the compound's low penetration and high reactivity. This
results in attenuation of the magnitude of the exposure concentration-based effect
estimate relative to the ambient concentration-based effect estimate (see Equation 3-6).
However, if the ratio is approximately constant over time, the strength of the statistical
3-63
-------
association would be similar for ambient concentration- and exposure
concentration-based effect estimates (Shcppard. 2005; Sheppard et al.. 2005).
3.5 Summary and Conclusions
The 2008 SOx ISA (U.S. EPA. 2008d') evaluated studies of ambient SO2 concentrations
and exposures in multiple microenvironments, discussed methods for estimating personal
and population exposure concentrations via monitoring and modeling, analyzed
relationships between personal exposure and ambient concentrations, and discussed the
implications of using ambient SO2 concentrations as estimates of exposure concentration
in epidemiologic studies. Key findings were that indoor SO2 concentrations and personal
SO2 exposure concentrations tended to be below the detection limit of personal SO2
samplers for averaging times of 24 hours or less, making it difficult to evaluate the
relationship between ambient SO2 concentrations and indoor or personal SO2 exposure
concentrations. However, in studies with the bulk of personal samples above the
detection limit, personal measurements of SO2 exposure were moderately correlated with
ambient SO2 concentrations. Regarding the influence of exposure concentration estimates
on epidemiologic study results, high spatial variability of ambient SO2 concentrations
across an urban area results in highly variable correlations among urban SO2 monitors.
Low correlations between individual monitored ambient SO2 concentrations and the
community average ambient SO2 concentration tend to bias effect estimates toward the
null, while variations in individual personal-ambient relationships across a community
will tend to widen confidence intervals around the effect estimates compared with what
would be obtained if the true exposure were used in the epidemiologic model. All of
these findings are supported by the recent evidence available since the previous ISA.
In the current ISA, increased focus has been placed on the use of exposure surrogates in
epidemiologic studies. Multiple techniques can be used to assign SO2 exposure
concentrations for epidemiologic studies, including the use of fixed-site monitor ambient
SO2 concentrations, personal SO2 monitors, and various types of models. Each has
strengths and limitations, as summarized in Table 3-2. Fixed-site monitors provide a
continuous record of ambient SO2 concentrations at their locations over many days or
years, but they do not capture the relatively high spatial variability in ambient SO2
concentration across an urban area and may not capture temporal variability of the plume
when the monitor is not in the plume's path. These features tend to attenuate health effect
estimates in time-series epidemiologic studies. For long-term studies, bias may occur in
either direction depending on whether the monitor is over- or underestimating ambient
SO2 exposure concentration for the population of interest. In all study types, use of
fixed-site monitor ambient SO2 concentrations in lieu of the true SO2 exposure
3-64
-------
concentrations is expected to widen confidence intervals compared with confidence
intervals that would be obtained if the true exposure were used. Personal SO2 monitors
directly measure exposure, but low ambient SO2 concentrations often result in a
substantial fraction of the samples falling below the MDL for averaging times of 24 hours
or less. Personal monitors also provide a relatively limited data set, making them more
suitable for panel epidemiologic studies. For microenvironmental sampling, FRM and
FEMs, described in Section 2.4. have also been deployed for panel studies as well.
Computational models can be used to develop exposure concentration surrogates for
individuals and large populations when personal exposure measurements are unavailable.
Modeling approaches may include LUR models, IDW models, dispersion models, and
CTMs. Strengths and limitations of each method are discussed in Table 3-2. Briefly, LUR
and IDW do not take into account atmospheric chemistry and physics. IDW is a weighted
average of ambient SO2 concentrations measured at several monitors. Other spatial
interpolation techniques, such as kriging, also require ambient SO2 concentrations from
several monitors and apply more complex mathematical functions to interpolate among
monitors. LUR regresses measured ambient SO2 concentrations on local variables and
then uses the resulting model to predict ambient SO2 concentrations across a study area or
at the locations of specific receptors. As such, LUR enables higher spatial resolution of
predicted ambient SO2 concentrations and requires more detailed input data compared
with IDW. Mechanistic models, such as dispersion models and CTMs, simulate the
transport and dispersion of ambient SO2, and in the case of CTMs, the atmospheric
chemistry. The strength of mechanistic models is increased accuracy of the ambient SO2
concentration field overtime and space. However, they are much more computationally
intensive. Microenvironmental models require personal sensor data for input and are
resource intensive. The strength of these models is that they account for time the exposed
population spend in different microenvironments. Microenvironmental models
(Section 3.3.2.6) incorporate time-activity data to overcome some limitations of spatial
smoothing in grid-based models, but they are rarely used in epidemiologic models. With
the exception of microenvironmental models, these methods tend to be used in
epidemiologic studies of long-term ambient SO2 exposure. Depending on the modeling
approach, there is the potential for bias and reduced precision due to model
misspecification, missing sources, smoothing of concentration gradients, and complex
topography. Evaluation of model results helps demonstrate the suitability of that
approach for particular applications.
The current ISA also reviews the newly available literature regarding indoor and personal
exposures to SO2. New studies of the relationship between indoor and outdoor SO2
concentrations have focused on public buildings and are consistent with previous studies
showing that indoor:outdoor ratios and slopes cover an extremely wide range, from near
3-65
-------
zero to near one. Differences in results among studies are due to the lack of indoor
sources of SO2, indoor deposition of ambient SO2, building characteristics (e.g., forced
ventilation, building age, and building type such as residences or public buildings),
personal activities, and analytical approaches. When reported, correlations between
indoor and outdoor SO2 concentrations were relatively high (>0.75), suggesting that
variations in outdoor SO2 concentrations are driving indoor SO2 concentrations. Several
studies of personal-ambient SO2 relationships available at the time of the previous ISA
showed a large fraction of samples below the MDL, making them unsuitable for
determining personal-ambient correlations. In a study with all personal samples above the
MDL, personal exposure was moderately correlated with ambient concentration.
Additional factors that could contribute to error in estimating exposure to ambient SO2
include time-location-activity patterns, spatial and temporal variability in SO2
concentrations, and proximity of populations to sources and monitor or model receptor
sites. Activity patterns vary both among and within individuals, resulting in
corresponding variations in exposure across a population and over time. Ambient SO2
concentrations among different microenvironments and the amount of time spent in each
location will jointly influence an individual's exposure to ambient SO2 (see
Equation 3-3). Time spent in different locations has also been found to vary by age, with
younger and older age groups spending a greater percentage of time outdoors than adults
of typical working age (18-64 years). These variations in activity pattern contribute to
differences in exposure and introduce error into population-averaged SO2 exposure
estimates.
Failure to account for spatial and temporal variability in ambient SO2 concentrations can
contribute to exposure error in epidemiologic studies, whether the study relies on
fixed-site monitor data or concentration modeling for exposure assessment. Ambient SO2
concentrations have low to moderate spatial correlations between ambient monitors
across urban geographic scales. This implies that a finer geographic scale is needed to
measure exposure concentration. Thus, using ambient SO2 concentration data measured
at fixed-site monitors as exposure surrogates in epidemiologic studies may introduce
exposure error into the resulting health effect estimate. Spatial variability in the
magnitude of ambient SO2 concentrations can affect cross-sectional and large-scale
cohort studies by undermining the assumption that intra-urban ambient SO2 exposure
differences across space are less important than inter-urban differences. This issue may
be less important for time-series studies, which rely on day-to-day temporal variability in
ambient SO2 exposure concentrations to evaluate health effects. Modeling techniques to
capture spatial variability can reduce exposure error in long-term average epidemiologic
models.
3-66
-------
Proximity of populations to ambient SO2 monitors may influence how well human
exposure to ambient SO2 is represented by measurements at the monitors, although
factors other than distance also play an important role. Many ambient SO2 monitoring
sites are located near dense population centers, but other near-source sites may not be
near population centers. Use of monitoring sites in epidemiologic studies introduces
exposure error into health effect estimates. The literature has shown that exposure error
and related bias in the health effect estimate is reduced by using averaging schemes in
lieu of a single fixed-site monitor (Section 3.4.2.2V
Exposure to copollutants may result in confounding of health effect estimates. For
ambient SO2, daily concentrations generally exhibit low to moderate correlations with
daily NAAQS copollutant concentrations at collocated monitors (Figure 3-4). However, a
wide range of copollutant correlations is observed at different monitoring sites, from
moderately negative to moderately positive. In studies where daily correlations of
ambient SO2 concentrations with ambient NO2 and CO concentrations were observed to
be high, it is possible the data were collected before rulemaking to reduce sulfur content
in diesel fuel went into effect in 2006 (66 FR 5002). The minority of sites with stronger
correlations have the potential to reflect a greater degree of confounding into the
epidemiologic results if the copollutant correlations at those sites are similar to the
copollutant correlations experienced at the locations of exposure. It is possible that the
observed correlation at a single site may not reflect copollutant correlations at the sites of
exposure, particularly in areas with a large amount of spatial heterogeneity of SO2. A
similar impact is expected for epidemiologic studies of long-term ambient SO2 exposure,
because a wide range of copollutant correlations have also been reported over time
periods of months to years.
Exposure error can contribute to variability in epidemiologic study results by biasing
effect estimates toward or away from the null and changing the size of the confidence
intervals compared with the confidence intervals around the effect estimate if the true
ambient SO2 exposure could have been used. The importance of exposure error varies
according to the study design, especially regarding the study's spatial and temporal
aspects. For example, in time-series and panel studies, low personal-ambient correlations
tend to bias the effect estimate toward the null, while spatial variation in
personal-ambient correlations across an urban area contributes to widening of the
confidence interval around the effect estimate compared with the confidence interval
produced using the true ambient SO2 exposure. For long-term studies, bias of the health
effect estimate may occur in either direction depending on whether the monitor is
over- or underestimating true ambient SO2 exposure concentration for the population of
interest. In all study types, use of fixed-site monitors in lieu of the true ambient SO2
exposure concentration is expected to decrease precision of the health effect estimate
3-67
-------
because spatial variation in personal-ambient correlations across an urban area can alter
the confidence interval around the effect estimate compared with the confidence interval
that would be obtained if the true ambient SO2 exposure concentration were used. Choice
of exposure concentration estimation method also influences the impact of exposure error
on epidemiologic study results. Fixed-site monitors offer a convenient source of
time-series data. However, because they are in a fixed location, ambient SO2
concentration measurements obtained from a fixed-site monitor do not account for the
effects of spatial variation in ambient SO2 concentration, ambient and nonambient
concentration differences, and varying activity patterns on personal exposure to ambient
SO2. Personal exposure measurements, such as those made in panel epidemiologic
studies, provide specific exposure estimates that may more accurately reflect spatial and
temporal variability, but sample size is often small and only a limited set of health
outcomes can be studied. Modeled ambient SO2 concentration or exposure concentration
estimates offer alternatives or supplementation to measurements, with the advantage of
estimating ambient SO2 exposure concentrations over a wide range of scales, populations,
and scenarios, particularly for locations lacking monitoring data. Model estimates are
most useful when compared to an independent set of measured ambient SO2
concentrations or exposure concentrations. The various sources of exposure error and
their potential impact on epidemiologic study results are evaluated in this ISA.
3-68
-------
CHAPTER 4 DOSIMETRY AND MODE OF
ACTION
4.1 Introduction
Chapter 4 begins by providing background information on the structure and function of
the respiratory tract (Section 4.1.1) and breathing rates and habits (Section 4.1.2). The
subsequent discussion of dosimetry of inhaled sulfur dioxide (SO2) (Section 4.2)
considers the chemical properties of SO2 and the processes of absorption, distribution,
metabolism, and elimination, as well as sources and levels of exogenous and endogenous
sulfite. The biological pathways that potentially underlie health effects are described in
"Modes of Action of Inhaled Sulfur Dioxide" (Section 4.3). This section includes a
description of processes by which inhaled SO2 initiates a cascade of molecular and
cellular responses and the organ-level responses that follow. Together, these sections
provide the foundation for understanding how exposure to inhaled SO2 may lead to health
effects. This understanding may provide biological plausibility for effects observed in the
epidemiologic studies.
4.1.1 Structure and Function of the Respiratory Tract
The basic structure of the human respiratory tract is illustrated in Figure 4-1. In the
literature, the terms extrathoracic (ET) region and upper airways or upper respiratory
tract are used synonymously. The terms lower airways and lower respiratory tract refer to
the intra-thoracic airways [i.e., the tracheobronchial (TB) and alveolar regions of the
lung].
4-1
-------
Posterior
Nasal Passage
Nasal Part
Oral Part
Extrathoracic
Region
Pharynx
Larynx
Trachea
Main Bronchi
Tracheobronchial
Region
Bronchi
Bronchioles
Bronchioles
— Terminal Bronchioles
Respiratory Bronchioles
Alveolar
Region
Alveolar Duct +
Alveoli
Source: Based on !CRP (1994).
Figure 4-1 Diagrammatic representation of respiratory tract regions in
humans.
4-2
-------
4.1.2 Breathing Rates and Breathing Habit
4.1.2.1 Breathing Rates
Breathing rates vary across the day and are generally a function of an individual's age,
sex, and activity level. Table 4-1 provides median ventilation rates extracted from
Tables 6-17 (males) and 6-19 (females) of the Exposure Factors Handbook (U.S. EPA.
2011). The original source of these ventilation rates is Table C4 of U.S. EPA (200%)
which describes their derivation. The median ventilation rates in Table 4-1 represent
central tendency estimates across a distribution of body weights for each age group and a
distribution of metabolic equivalents of work for each age group. Additional information
for other ages and percentiles of the ventilation rate distribution are available from those
tables. Except for the oldest age range, ventilation rates (volume/time) increase with
activity level and age and are greater in men than women.
Table 4-1 Ventilation rates in humans as a function of activity.
Median Ventilation Rate (L/min)
Sex
Age (Yr)
Sleep
Light Activity
Moderate Activity
Strenuous Activity
Male
3 to <6
4.29
11.1
20.6
37.8
6 to <11
4.46
11.3
21.6
41.9
21 to <61
5.17
13.6
29.7
52.9
>81
5.90
13.8
28.2
50.9
Female
3 to <6
4.1
10.7
19.8
33.3
6 to <11
4.24
10.8
20.4
38.0
21 to <61
4.06
11.1
23.0
44.2
>81
4.39
10.3
20.6
41.4
Source: U.S. EPA (2011) and U.S. EPA (2009b).
Ventilation rates are also higher in overweight individuals compared with those of normal
weight (Brochu et al.. 2014). For example, median daily ventilation rates (m3/day) are
about 1.2 times greater in overweight [>85th percentile body mass index (BMI)] than
normal-weight children (5-10 years of age). In 35-45-year-old adult males and females,
4-3
-------
ventilation rates are 1.4 times greater in overweight (BMI > 25 kg/m2) than
normal-weight (18.5 to <25 kg/m2 BMI) individuals. Across all ages, overweight/obese
individuals respire greater amounts of air and associated pollutants than age-matched
normal-weight individuals.
One way to consider differences in ventilation rates between adults and children is to
normalize to body weight. This metric is relevant especially for SO2 absorbed in the nasal
airways and the fraction of absorbed SO2 that distributes systemically (Section 4.2.3).
Normalized to body mass, median daily ventilation rates (m3/kg-day) decrease over the
course of life (Brochu et al.. 201IV This decrease in ventilation relative to body mass is
rapid and nearly linear from infancy through early adulthood. Relative to normal-weight
male and female adults (25-45 years of age; 0.271 m3/kg-day), ventilation rates
normalized to body mass are increased 1.5 times in normal-weight children (7-10 years
of age; 0.402 m3/kg-day) and doubled in normal-weight infants (0.22-0.5 years of age;
0.538 m3/kg-day). Although adults have greater absolute ventilation rates than children in
terms of inhaled volume per unit time, normalized to body size, children intake greater
volumes of air and associated pollutants than adults.
A potentially useful metric for assessing SO2 dose to the bronchi and differences between
children and adults in bronchial effects of SO2 is SO2 absorbed dose per bronchial surface
area (Section 4.2.2). Ventilation per tracheobronchial surface area is also used to
approximate absorbed dose per bronchial surface area for inter-species extrapolation
[Appendix A of U.S. EPA (2009d)l.
4.1.2.2 Breathing Habit
As humans, we breathe oronasally (i.e., through both our nose and mouth). In general, we
breathe through our nose when at rest and increasingly through the mouth with increasing
activity level. Few people breathe solely through their mouth. In contrast to the oronasal
breathing of humans, rodents are obligate nasal breathers. Described in Section 4.2.2. the
nasal passages more efficiently remove SO2 from inhaled air than the oral passage. As the
fraction of inhaled air passing through the mouth increases so too does the amount of
inhaled SO2 reaching the tracheobronchial airways where SO2 may cause
bronchoconstriction. Thus, route of breathing (namely, the fraction of inhaled air passing
through the mouth) is a critical determinate of dose to the lower airways and the potential
respiratory effects of SO2. This section describes how route of breathing, also referred to
as "respiratory mode" or "breathing habit" in the literature, is affected by age, sex,
obesity, activity level, and upper respiratory tract anomalies.
4-4
-------
One of the more commonly referenced studies in dosimetric papers is Niinimaa et al.
(1981). These investigators found that most people, 87% (26 of 30) in the study, breathed
through their nose until an activity level was reached where they switched to oronasal
breathing. Thirteen percent (4 of 30) of the subjects, however, were oronasal breathers
even at rest. These two subject groups are commonly referred to in the literature (e.g..
ICRP. 1994) as "normal augmenters" and "mouth breathers," respectively. Bennett et al.
(2003) reported a more gradual increase in oronasal breathing, with males (n = 11;
22 ± 4 years) tending to have a greater oral contribution than females (n = 11; 22 ± 2
years) at rest (87 vs. 100% nasal, respectively) and during exercise (45 vs. 63% nasal at
60% max workload, respectively).
Consistent with this trend for women to have a greater nasal contribution (Bennett et al..
2003); in a large study of children (63 M, 57 F; 4-19 years), Leiberman et al. (1990)
reported a statistically greater nasal fraction during inspiration in girls relative to boys (77
and 62%, respectively; p = 0.03) and a marginally significant difference during expiration
(78 and 66%, respectively;p = 0.052). Another large study (88 M, 109 F; 5-73 years)
also reported females as having a significantly greater fraction of nasal breathing than
males (Vig and Zaiac. 1993). This effect was largest in children (5-12 years) with an
inspiratory nasal fraction, under resting conditions, of 66% in males and 86% in females.
This study also reported that the partitioning between the nose and mouth was almost
identical between inspiration and expiration. In children and adults, sex explains some
inter-individual variability in route of breathing, with females breathing more through the
nose than males.
A few studies have attempted to measure oronasal breathing in children as compared to
adults (Bennett et al.. 2008; Becquemin et al.. 1999; James et al.. 1997; Vig and Zaiac.
1993). James et al. (1997) found that children (n = 10; 7-16 years) displayed more
variability than older age groups (n = 27; 17-72 years) with respect to their oronasal
pattern of breathing with exercise. Becquemin et al. (1999) found that children (n = 10;
8-16 years) tended to display more oral breathing both at rest and during exercise than
adults (n = 10; 27-56 years). The highest oral fractions were also found in the youngest
children. Similarly, Bennett et al. (2008) reported children (n = 12; 6-10 years) tended to
have a greater oral contribution than adults (n = 11; 18-27 years) at rest (68 vs. 88%
nasal, respectively) and during exercise (47 vs. 59% nasal at 40% max workload,
respectively). Vig and Zaiac (1993) reported a statistically significant effect of age on
route of breathing which was most apparent in males with the fraction of nasal breathing
increasing from 67% in children (5-12 year olds) to 82% in teens (13-19 year olds), and
86% in adults (20-73 years). Females had a nasal fraction of 86% in children and teens
and 93% in adults. Based on these studies, the nasal fraction increases with age until
adulthood.
4-5
-------
Several large studies have reported an inverse correlation (r= -0.3 to -0.6) between nasal
resistance and nasal breathing fraction (Vig and Zaiac. 1993; Leiberman et al.. 1990;
Leiter and Baker. 1989). However, neither pharmaceutical constriction nor dilation of the
nasal passages affected the nasal fraction (Leiberman et al.. 1990; Leiter and Baker.
1989). Nasal resistance decreases with age and is lower in females and may account for
larger nasal fractions in adults and females (Vig and Zaiac. 1993). Smaller studies
(n = 37) have not found a significant correlation between nasal resistance and nasal
fraction, but have noted that those having high resistance breathe less through the nose
(James et al.. 1997). Bennett et al. (2003) reported a tendency of lower nasal resistance in
African-American blacks (5 M, 6 F; 22 ± 4 years) relative to Caucasians (6 M, 5 F;
22 ± 3 years). The nasal fraction in blacks tended to be greater at rest and 40% max
workload and achieved statistical significance relative to Caucasians at 20 and 60% max
workload. Leiter and Baker (1989) reported that of the 15 mouth-breathing children as
identified by a dentist, pediatrician, or otolaryngologist in their study, the 3 having
greatest nasal resistance breathed 100% through the mouth. These investigators also
reported that the nasal fraction was negatively correlated (p < 0.004) with nasal resistance
during both inspiration and expiration; however, the correlation appears driven by the
three individuals with 100% mouth breathing. Overall, breathing habit is related to nasal
resistance, which may explain some of the effects of age and sex on breathing habit.
Diseases affecting nasal resistance may also affect breathing route. Chadha et al. (1987)
found that the majority (11 of 12) of patients with asthma or allergic rhinitis breathe
oronasally (i.e., they breathe partially through the mouth) even at rest. James et al. (1997)
also reported the subjects (n = 37; 7-72 years) having hay fever, sinus disease, or recent
upper respiratory tract symptoms tended to the have a greater oral contribution relative to
those absent upper respiratory tract symptoms. James et al. (1997) additionally observed
that two subjects (5.4%) breathed purely through the mouth, but provided no other
characteristics of these individuals. Greater oral breathing may occur due to upper
respiratory tract infection and inflammation.
Some studies of children suggest obesity also affects breathing habit. Using MRI,
Schwab et al. (2015) examined anatomic risk factors of obstructive sleep apnea in
children (n = 49 obese with sleep apnea, 38 obese control, 50 lean controls; 11-16 years
of age). In obese children with sleep apnea, adenoid size was increased relative to both
obese and lean controls not having sleep apnea. The size of the adenoid was also
increased in male obese controls (n = 24) relative to male lean controls (n = 35), whereas
adenoid size was similar between female obese controls (n = 14) and female lean controls
(n = 15). Both nasopharyngeal cross-sectional area and minimum area were similar
between lean and obese controls, but decreased in obese children with obstructive sleep
apnea. In a longitudinal study of children (n = 47 F, 35 M) assessed annually from 9 to
4-6
-------
13 years of age, Crouse et al. (1999) found nasal cross section was minimal at 10 years of
age. The authors speculated this may be due to prepubertal enlargement of the adenoids.
In a 5-year longitudinal study of children (n = 17 M, 9 F) following adenoidectomy, Kerr
et al. (1989) reported a change in mode of breathing from oral to nasal. These studies
suggest that obese children, especially boys, may have increased oral breathing relative to
normal-weight children.
In summary, breathing habit is affected by age, sex, nasal resistance, and perhaps by
obesity. Numerous studies show children to inhale a larger fraction of air through their
mouth than adults. Across all ages, males also inhale a larger fraction of air through their
mouth than females. Other factors that increase nasal resistance such as allergies or acute
upper respiratory infections can also increase the fraction of oral breathing. Obesity,
especially in boys, may also contribute to increased nasal resistance and an increased oral
fraction of breathing relative to normal-weight children.
4.2 Dosimetry of Inhaled Sulfur Dioxide
This section provides a brief overview of SO2 dosimetry and updates information
provided in the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d). Dosimetry of SO2 refers
to the measurement or estimation of the amount of SO2 and its reaction products reaching
and/or persisting at specific sites within the respiratory tract or systemically after
exposure. One principal effect of inhaled SO2 is to stimulate bronchial epithelial irritant
receptors and initiate a reflexive contraction of smooth muscles in the bronchial airways.
Health effects may be due to the inhaled SO2 or its chemical reaction products. Complete
identification of the causative agents and their integration into SO2 dosimetry is a
complex issue that has not been thoroughly evaluated. The major factors affecting the
transport and fate of gases and aerosols in the respiratory tract are the morphology of the
respiratory tract; the physiochemical properties of the epithelial lining fluid (ELF);
respiratory functional parameters, such as tidal volume, flow rate, and route of breathing;
the physicochemical properties of the gas; and the physical processes that govern gas
transport. Few studies have investigated SO2 dosimetry since the 1982 AQCD for
Particulate Matter and Sulfur Oxides (U.S. EPA. 1982a) and the 1986 Second Addendum
(U.S. EPA. 1986).
The following sections address SO2 chemistry and the processes of absorption,
distribution, metabolism, and elimination that pertain to the dosimetry of inhaled SO2.
Studies investigating the dosimetry of SO2 generally are for concentrations of SO2 that
are higher than those present in ambient air. However, these studies are included here
because they provide the foundation for understanding SO2 toxicokinetics and
4-7
-------
toxicodynamics. The discussion of dosimetry concludes with a consideration of other
sources of SCh-derived products in the body.
4.2.1 Chemistry
Physicochemical properties of SO2 most relevant to respiratory tract uptake include its
solubility in the ELF and its chemical transformations and reactions that occur there.
Henry's law relates the gas-phase and liquid-phase interfacial concentrations at
equilibrium and is a function of temperature and pressure. The Henry's law constant,
defined as the ratio of partial pressure or concentration of SO2 in the gas phase to SO2
dissolved in the liquid phase, is an inverse measure of solubility. Although the solubility
of SO2 in the ELF is not known, the effective Henry's law constant or solubility factor is
known for SO2 in water and is 0.047 (mol/L)air per (mol/L)water at 37°C and 1 atmosphere
(Hales and Sutter. 1973). For comparison, Henry's law constant for ozone (O3) is 6.4
(mol/L)air per (mol/L)„ater under the same conditions (Kimbcll and Miller. 1999). Thus,
SO2 is nearly 140 times more soluble than O3 in water. In general, the more soluble a gas
is in biological fluids, the more rapid and proximal its absorption will be in the
respiratory tract. In addition to the Henry's law constant, it is also necessary to consider
the transport of SO2 from the lumen to the ELF of the tracheobronchial airways
(Section 4.2.2). When the partial pressure of SO2 on mucosal surfaces exceeds that of the
gas phase, such as during expiration, some desorption of SO2 from the ELF may be
expected (Section 4.2.5).
Once SO2 contacts the fluids lining the airways, it dissolves into the aqueous
compartment and rapidly hydrates to form sulfurous acid (H2SO3), which forms hydrogen
(H+) ions, bisulfite (HSO3 ) anions, and sulfite (SO32 ) anions (Gunnison et al.. 1987a;
Gunnison. 1981).
-H+ -H+
S02 + H20^H2S03 ^ hso3 ^ SOf"
+H+ +H+
Equation 4-1
The prevalence of these sulfur species in solution is determined primarily by pH, and to a
lesser extent, by temperature and ionic strength. In the human respiratory tract (pH of 7.4
and 37°C), dissolved SO2 exists exclusively as a mixture of bisulfite and sulfite, with the
latter predominating (Gunnison. 1981). Subsequent reactions of bisulfite and sulfite such
as sulfitolysis, enzymatic detoxification, and auto-oxidation are described below.
Hydrogen ions may impact airway physiology via acidification reactions.
4-8
-------
4.2.2
Absorption
Because SO2 is highly soluble in water, it is expected to be almost completely absorbed
in the nasal passages of both humans and laboratory animals under resting conditions.
The dosimetry of SO2 can be contrasted with the lower solubility gas, O3, for which the
predicted tissue doses (O3 flux to the liquid-tissue interface) are very low in the trachea
and increase to a maximum in the terminal bronchioles or first airway generation in the
pulmonary region [Chapter 5 of U.S. EPA (2013bYI.
Melville (1970) measured the absorption of SO2 (1.5 to 3.4 ppm) during nasal and oral
breathing in 12 healthy volunteers. Total respiratory tract absorption of SO2 (expressed as
a percentage of the amount inhaled) was significantly greater (p < 0.01) during nasal than
oral breathing (85 vs. 70%, respectively) and was independent of the inspired
concentration. Respired flows were not reported. Andersen et al. (1974) measured the
nasal absorption of SO2 (25.5 ppm) in seven volunteers at an average inspired flow of
23 L/minute [i.e., eucapnic hyperpnea (presumably to simulate light exertion)]. These
investigators reported that the oropharyngeal SO2 concentration was below their limit of
detection (0.25 ppm), implying that at least 99% of SO2 had been absorbed in the nose of
subjects during inspiration. Speizer and Frank (1966) also measured the absorption of
SO2 (16.1 ppm) in seven healthy subjects at an average ventilation of 8.5 L/minute
(i.e., at rest). They reported that 14% of the inhaled SO2 was absorbed within the first
2 cm into the nose. The concentration of SO2 reaching the pharynx was below the limit of
detection, suggesting that at least 99% was absorbed in the nose during inspiration.
Frank et al. (1969) and Brain (1970) investigated the oral and nasal absorption of SO2 in
the surgically isolated upper respiratory tract of anesthetized dogs. Radiolabeled SO2
(35S02) at concentrations of 1, 10, 25, or 50 ppm was passed separately through the nose
and mouth at steady flows of 3.5 and 35 L/minute for 5 minutes by Brain (1970V The
nasal absorption of SO2 (1 ppm) was effectively 100% at 3.5 L/minute and 96.8% at
35 L/minute. The effect of SO2 concentration on nasal absorption was negligible, with
nasal absorption increasing from 99.9% at 1 ppm to 99.99% at 10 ppm and 99.999% at
50 ppm at 3.5 L/minute. The oral absorption of SO2 (1 ppm) was 99.56% at 3.5 L/minute,
but only 34% at 35 L/minute. There was a slight decrease in oral SO2 absorption from
99.56 to 96.3% when the concentration was increased from 1 to 10 ppm at 3.5 L/minute,
whereas nasal absorption was unaffected by changes in concentration (1-50 ppm). In an
earlier experiment, Frank et al. (1967) showed that nasal absorption of 2.2 ppm 35SC>2 at
3.5 L/minute was 100% throughout the first 20 minutes of exposure. On average, there
was a small reduction in 35SC>2 absorption to 94% approaching 30 minutes of exposure.
Frank et al. (1969) noted that the aperture of the mouth may vary considerably, and that
this variation may affect SO2 uptake in the mouth. Although there was a minor effect of
4-9
-------
inhaled concentration on SO2 absorption, the route of breathing and rate of flow were the
main factors affecting the magnitude of SO2 absorption in the upper airways of dogs.
Modeling shows that virtually all SO2 reaching the lower airways in young adults, as well
as in dogs and rats, is absorbed in the bronchi and does not penetrate into the bronchioles
or alveolar region (Tsuiino et al.. 2005). Thus, ventilation per bronchial surface area can
serve as a surrogate for inhaled SO2 dose per unit bronchial surface area. Considering the
effect of age on SO2 dose to the airways of humans, dose as ventilation per bronchial
surface area can be estimated using bronchial morphology from Phalen et al. (1985) and
50th-percentile ventilation rates from Brochu etal. (2011). This approximation shows a
gradual reduction in bronchial surface dose with decreasing age from young adults to
infants. Using this approximation, an infant (4-months old) would have approximately
80% of the bronchial surface dose of a young adult (18-years old). However, as described
in Section 4.1.2.2. children breathe more through the mouth than adults, which is
associated with greater SO2 penetration to the lower respiratory tract. SO2 penetrating
through the upper airways is rapidly removed in the trachea and the first several
generations of bronchi, possibly resulting in somewhat greater airway surface doses of
SO2 of children than adults in proximal bronchi due to the greater oral breathing
contribution of children than adults.
In summary, inhaled SO2 is readily absorbed in the upper airways of both humans and
laboratory animals. During nasal breathing, the majority of available data suggests 95%
or greater SO2 absorption occurs in the nasal passages, even under ventilation levels
comparable to those during exercise. Somewhat less SO2 is absorbed in the oral passage
than in the nasal passages. The difference in SO2 absorption between the mouth and the
nose is highly dependent on respired flow rates. With an increase in flow from 3.5 to
35 L/minute, nasal absorption is relatively unaffected, whereas oral absorption is reduced
from 100 to 34%. Inhaled SO2 concentration has a negligible effect on nasal absorption,
whereas oral absorption may decrease slightly with increasing concentration from 1 ppm
to 10 ppm SO2. Thus, the rate of breathing (namely, for oral breathing) and the route of
breathing (i.e., the contribution through the nose vs. mouth) have a great effect on the
magnitude of SO2 absorption in the upper airways and on the penetration of SO2 to the
lower airways. Overall, the available data clearly show a pattern of SO2 absorption that
shifts from the upper airways to the tracheobronchial airways in conjunction with a shift
from nasal to oronasal breathing and associated increased ventilatory rates in exercising
humans. Due to their increased amount of oral breathing, children (particularly boys and
perhaps the obese) and individuals with allergies or upper airway infections may be
expected to have greater SO2 penetration into the lower respiratory tract than healthy
adults (Section 4.1.2).
4-10
-------
4.2.3
Distribution
Once inhaled, SO2 is absorbed in the respiratory tract, and SC>2-derived products are
widely distributed throughout the body, as demonstrated in studies using radiolabeled
35S02. Although rapid extra-pulmonary distribution of SCh-derived products occurs, the
highest tissue concentrations of the 35S retained in the body at any given time are found
primarily in the respiratory tract (upper and lower) and may be detected there for up to a
week following inhalation (Balchum et al.. 1960. 1959). Frank etal. (1967) observed 35S
in the blood and urine of dogs within 5 minutes, the first time point, after starting 22 ppm
35S02 exposures of the surgically isolated nasal airways. At the end of 30-60-minute
exposures, the authors estimated that 5-18% of the administered 35S was in the blood.
Balchum et al. (1959) investigated the tissue distribution of 35S in dogs exposed for
20-40 minutes to 35SC>2 ranging in concentration from 1.1 to 141 ppm via tracheostomy
or by nose/mouth breathing. At approximately 1-hour post-exposure, regardless of the
exposure route or the 35SC>2 exposure concentration, about 6% of the retained 35S was
found in the liver, with lesser amounts found in the heart, spleen, kidney, brain, and other
tissues. However, the percent of retained 35S was, on average, 13 times greater in the
trachea and lungs of the tracheostomized group than in the nose/mouth breathing group,
demonstrating the protection of the lower respiratory tract provided by SO2 removal in
the upper airways. Comparison of dogs retaining similar total amounts of35S
(i.e., controlling for retained dose), showed that the blood concentrations of 35S were
higher in the tracheostomized dogs than in the nose/mouth breathing dogs. Given very
high 35S concentrations in the tongues of the nose/mouth breathing dogs and that blood
concentrations had not decreased in two-thirds of these dogs by 1 -hour post-exposure, the
authors postulated that a substantial portion of the 35SC>2 products may have been retained
within the upper airways with only slow absorption into the blood. Studies in rabbits and
rats also show that there can be an accumulation and retention of SCh-derived products
within proximal regions of the respiratory tract (discussed below).
The distribution and clearance of inhaled SO2 from the respiratory tract may involve
several intermediate chemical reactions and transformations. In particular, hydrated SO2
transforms to sulfite/bisulfite at physiologic pH. Sulfite can diffuse across cell
membranes, and bisulfite can react with disulfide bonds (R1-S-S-R2) to form thiols
(Ri-SH) and S-sulfonates (R2-S-SO3 ) by a process termed sulfitolysis (Gunnison and
Benton. 1971). Disulfide bonds are important determinants of protein structure and
function in biological systems. Secreted airway mucins contain many disulfide bonds,
and breaking these bonds might alter their function and thereby alter mucociliary
clearance.
4-11
-------
Studies in rabbits and rats found measurable levels of sulfite and S-sulfonates in the
upper respiratory tract following inhalation of 10-30 ppm SO2. Levels of sulfite and
S-sulfonates were increased in tracheal washings of rabbits exposed to 10 ppm SO2 for up
to 72 hours (Gunnison et al.. 1981). implying reaction of sulfite with disulfide groups in
mucus proteins in the ELF. In addition, tracheal tissue contained elevated levels of
S-sulfonates, implicating reaction of sulfite with disulfide groups in tissue proteins.
Bronchial tissue from rats had increased levels of sulfites and S-sulfonates when higher
concentrations (30 ppm) of SO2 were employed (Gunnison et al.. 1987b). Under these
conditions, no S-sulfonates were found in lung parenchyma, and neither sulfites nor
S-sulfonates were found in the plasma. The lack of sulfites and S-sulfonates in the plasma
of rats may have been due to these animals' high levels of sulfite oxidase and rapid
metabolism of sulfite (Section 4.2.4). Consistent with 35S rapidly appearing in the blood
of 35SC>2-exposed dogs, S-sulfonates were found in plasma of rabbits following 10 ppm
SO2 exposure, providing evidence for absorption of sulfite into the blood of rabbits
(Gunnison et al.. 1981; Gunnison and Palmes. 1973). Experiments examining in vivo and
ex vivo plasma have shown that sulfite reacts with disulfide bonds in albumin and
fibronectin to produce S-sulfonates (Gregory and Gunnison. 1984).
Exposure of humans to SO2 also resulted in measurable S-sulfonates in plasma (Gunnison
and Palmes. 1974). In this study, humans were exposed continuously to 0.3-6 ppm SO2
for up to 120 hours, and plasma levels of S-sulfonates were positively correlated with
concentrations of SO2 inhaled. The regression line for this relationship had a correlation
coefficient of 0.61 and the slope was 1.1 nmol/mL of plasma S-sulfonate for each 1-ppm
increment in SO2 concentration. A subacute study measured sulfite plus S-sulfonate
content of the lung, liver, and brain of mice exposed to 5, 10, or 20 ppm SO2, 4 hours/day
for 7 days (Meng et al.. 2005a). A concentration-dependent increase in sulfite and
S-sulfonate levels was observed. Thus, in humans and mice, the amount of SCh-derived
species in blood and other tissues increases with the concentration in inhaled air. It
should also be noted that measurable amounts of sulfite/S-sulfonate were found in tissues
of humans and mice inhaling filtered air instead of SO2 (Meng et al.. 2005a; Gunnison
and Palmes. 1974). Besides inhaled SO2, sulfite is derived from other exogenous, as well
as endogenous sources (Section 4.2.6).
Inhaled SO2 need not reach the lower airways for SC>2-derived species to be found in the
blood. During the 5 full days of SO2 exposure in the Gunnison and Palmes (1974) study,
volunteers were likely at rest or sleeping for much of their exposures. Given that
ventilation rates would be relatively low and breathing would be largely nasal
(Section 4.1.2). most inhaled SO2 would likely be absorbed in the extrathoracic airways
(Section 4.2.2). A number of studies also exposed the surgically isolated upper airways of
dogs to 35S02 and observed35S to rapidly appear in the blood and for the concentration in
4-12
-------
blood to continuously increase during exposure (e.g., Yokoyama et al.. 1971; Frank et al..
1967). Frank et al. (1969) proposed the majority of SC>2-derived products found in the
blood originated from SO2 absorbed in the upper airways.
In summary, inhaled SO2 is readily dissolved in the ELF where it exists as a mixture of
bisulfite and sulfite with the latter predominating. Bisulfite reacts with disulfide groups
forming S-sulfonates; sulfite can diffuse across cell membranes and reach the circulation.
Following absorption in the respiratory tract, SC>2-derived products (e.g., sulfite and/or
S-sulfonates) are widely distributed throughout the body and have been observed in the
blood and urine within 5 minutes of starting an SO2 exposure of surgically isolated nasal
airways. Measurable levels of S-sulfonates have been observed in plasma following
inhalation of SO2 in humans, dogs, mice, and rabbits. Perhaps due to higher levels of
hepatic sulfite oxidase relative to other species, sulfites, and S-sulfonates are not found in
the plasma of rats. Although the majority of SCh-derived products remain in the
respiratory tract following exposure, extrapulmonary SC>2-derived products are found in
the liver, with lesser amounts found in the heart, spleen, kidney, brain, and other tissues.
The amount of SCh-derived species in blood and other tissues increases with the
concentration of SO2 in inhaled air, while the distribution within the body is generally
unaffected. A substantial portion of SC>2-derived products appear to be retained within the
upper airways, particularly during nasal breathing, with only slow absorption into the
blood.
4.2.4 Metabolism
The primary route of sulfite metabolism is by sulfite oxidase-catalyzed enzymatic
oxidation to sulfate (Gunnison. 1981). Because of this pathway, intra-cellular steady-state
concentrations of sulfite are low in normal individuals (Gunnison etal.. 1987a). Sulfite
oxidase is a molybdenum-containing enzyme that is found in mitochondria. Its
distribution varies widely across tissues. While lung tissue has very low sulfite oxidase
activity, liver has high sulfite oxidase activity and plays a major role in detoxification of
circulating sulfite. Maieretal. (1999) examined the distribution of sulfite oxidase activity
in the respiratory tract and liver of four beagle dogs. Sulfite oxidase activity was highest
in the liver. The median sulfite oxidase activity in the nose was about 30% of the liver.
Median activity levels in the trachea and bronchi were about 20% of the liver and the
median activity levels in the lung parenchyma were only 10% of those in the liver. The
1982 AQCD (U.S. EPA. 1982a) noted that depleting the activity of sulfite oxidase in an
animal model through a low-molybdenum diet supplemented with the competitive
inhibitor tungsten resulted in a substantial lowering of the lethal dose for
intra-peritoneally injected bisulfite. A deficiency in sulfite oxidase activity may lead to
4-13
-------
toxicity even in the absence of exogenous sulfite or bisulfite exposures. For example,
humans and mice with homozygous genetic defects in the sulfite oxidase protein or in the
enzymes required to synthesize the essential molybdenum cofactor develop ultimately
lethal neurologic disease attributable to accumulation of endogenous sulfite postnatally
(i.e., following loss of maternal protection in utero) (Johnson-Winters et al.. 2010; Reiss
et al.. 2005).
Sulfite oxidase activity is highly variable among species. Liver sulfite oxidase activity in
the rat is 10-20 times that in humans. Rapid metabolism of circulating sulfite to sulfate
may explain the lack of sulfite/S-sulfonates found in blood of rats exposed by inhalation
to 30 ppm SO2, whereas these products were found in other species (Gunnison et al..
1987a). In sulfite oxidase-deficient rats, plasma sulfite levels increase with the severity of
the deficiency (Gunnison et al.. 1987b).
Gunnison and Benton (1971) also identified S-sulfonate in blood as a reaction product of
inhaled SO2. S-sulfonates, which are produced by the reaction of bisulfite with disulfide
bonds, may be metabolized back to disulfides. Although the enzymatic pathways and
cofactors are not clearly established for this repair process, it requires reducing
equivalents, and thus, has a metabolic cost.
In summary, the primary route of sulfite metabolism is by sulfite oxidase-catalyzed
oxidation into sulfate. The sulfite oxidase levels vary widely among tissues with very low
levels found in the lung and high levels found in the liver, which plays a major role in the
detoxification of circulating sulfite. Sulfite oxidase activity is also highly variable among
species with liver sulfite oxidase activity in rats being 10-20 times greater than in
humans.
4.2.5 Elimination
Mechanisms involved in elimination include both desorption of SO2 from the respiratory
tract and the clearance of reaction products from the body.
When the partial pressure of SO2 on mucosal surfaces exceeds that of the gas phase, such
as during expiration, some desorption of SO2 from respiratory tract lining fluids may be
expected. Speizer and Frank (1966) found that on expiration, 12% of the SO2 absorbed
during inspiration was desorbed into the expired air. During the first 15 minutes after the
25- to 30-minute SO2 exposure, another 3% was desorbed. In total, 15% of the amount of
originally inspired and absorbed SO2 was desorbed from the nasal mucosa. Frank et al.
(1969) reported that up to 18% of the SO2 was desorbed within -10 minutes after
exposure.
4-14
-------
SO2 that does not desorb is transformed to bisulfite/sulfite (Section 4.2.1). Because the
lung tissue has a low activity of sulfite oxidase, diffusion into the circulation may be a
more important route of sulfite clearance from the lung than enzyme-catalyzed
transformation to sulfates. Within a period of minutes after starting 35SC>2 inhalation
exposures, 35S was observed in the blood and urine of dogs and distributed about the
body (Frank etal.. 1967; Balchum et al.. 1959V At the end of 30-60-minute exposures,
5-18% of the administered 35S was in the blood, and 1-6% had been excreted in the
urine by 3 hours post-exposure (Yokovama et al.. 1971; Frank etal.. 1967). The rate of
urinary excretion was proportional to the blood concentration, and 92% of the urinary 35 S
was in the form of sulfate (Yokovama etal.. 1971). In contrast, S-sulfonates formed in
the circulation were reported to have a clearance half-time of 3.2 days (Gunnison and
Palmes. 1973).
In summary, when the partial pressure of SO2 on mucosal surfaces exceeds that of the gas
phase, such as during expiration or following exposure, some desorption of SO2 from the
respiratory tract lining fluids may be expected. SO2 that does not desorb is transformed to
bisulfite/sulfite. Given the low activity of sulfite oxidase in the respiratory tract, sulfite is
more likely to diffuse into the circulation or react with tissue constituents than be
metabolized to sulfate. Circulating sulfite may subsequently react with constituents of the
blood to form S-sulfonates or other species. Sulfite reaching the liver (Section 4.2.3) is
efficiently metabolized to sulfate (Section 4.2.4). Urinary excretion of sulfate is rapid and
proportional to the concentration of SO2 products in the blood. S-sulfonates are cleared
more slowly from the circulation with a clearance half-time of days. The portion of
SCh-derived products that are retained within the respiratory tract are only slowly
absorbed into the blood (Section 4.2.3).
4.2.6 Sources and Levels of Exogenous and Endogenous Sulfite
The primary endogenous contribution of sulfite is from the catabolism of
sulfur-containing amino acids (namely, cysteine and methionine). Sulfite may
subsequently be metabolized to sulfate in a reaction catalyzed by sulfite oxidase in most
tissues, but especially in the liver (Section 4.2.4). Mean daily sulfate produced following
ingestion of cysteine and methionine in the U.S. increases from 70 mg/kg-day in infants
(ages 2-6 months) to 100 mg/kg-day in young children (1-3 years old) and then
decreases to 30 and 40 mg/kg-day in adult (19-50 years old) females and males,
respectively (IOM. 2005). To facilitate comparison with exogenous sources, a mole of
SO2 can produce a mole of sulfate, but the SO2 mass is only two-thirds of the sulfate
mass.
4-15
-------
Sulfite is also added to foods because it has antioxidant and antimicrobial properties
(Vandeviivere et al.. 2010; Gunnison. 1981). In a study considering actual food
consumption of Belgian adults and measured sulfite levels in food, Vandeviivere et al.
(2010) observed a wide distribution in exogenous sulfite from ingestion. Expressed in
terms of SO2 equivalents, rates of exogenous sulfite ingestion may be described by a
log-normal distribution with a median intake of 0.14 SO2 mg/kg-day and a geometric
standard deviation of 2.15. Individuals at the 5th and 95th percentiles of this distribution
are estimated to consume 0.04 and 0.49 SO2 mg/kg-day. In a comparison of theoretical
food-consumption data with maximum permissible SCh/sulfites to foods, the Belgian
adults in the Vandeviivere et al. (2010) study had a similar potential sulfite intake to U.S.
adults. The estimated intake for children could be in the range of that for adults or less
due to the likely minimal consumption of sulfite sources such as wine. Endogenous
sulfite from catabolism of ingested sulfur-containing amino acids far exceeds exogenous
sulfite from ingestion of food additives [by 140 and 180 times in adult (19-50 years old)
females and males, respectively, and by 500 times or more in young children (1-3 years
old)].
Exogenous sulfite may also be derived from SO2 inhalation. For the purposes of
comparisons herein, all inhaled SO2 is assumed to contribute to systemic sulfite levels. In
reality, as discussed in Section 4.2.3. the majority of S02-derived products from SO2
inhalation are retained in the respiratory tract and may be detected there for up to a week
following inhalation. The potential contribution of inhaled SO2 to systemic sulfite levels
varies with age, activity level, and SO2 concentration. Using median and 97.5th percentile
daily ventilation rates from Brochuetal. (2011). normal-weight adults (25-45 years of
age) are estimated to receive 0.004 and 0.006 mg SO2 per kg body mass, respectively,
from a full day exposure to 5 ppb SO2. As an upper-bound estimate for ambient exposure
in most locations, a full-day exposure to 75 ppb SO2 (the level of the current National
Ambient Air Quality Standard for SOx) would result in 0.053 SO2 mg/kg-day and
0.085 SO2 mg/kg-day for adults having median and 97.5th percentile ventilation rates,
respectively. The estimated daily SO2 intake (mg/kg-day) would be roughly 1.5 times
greater in children (7-10 years of age) and doubled in infants (0.22-0.5 years of age) due
to the greater ventilation rate per body mass of children compared to adults (25-45 years
of age). Even upper-bound sulfite levels from inhalation (75 ppb SO2, 24 hours, 97.5th
percentile ventilation) are far less than those derived from catabolism of sulfur-containing
amino acids, by 230 to 300 times in adults (25-45 years) and nearly 500 times in young
children (1-3 years).
Comparison of sulfite derived from SO2 inhalation with that from ingestion of food
additives is more complicated. In adults (25-45 years), sulfite intake (mg/kg-day) from
inhalation (75 ppb SO2, 24 hours, 97.5th percentile ventilation) is 1.6 times lower than
4-16
-------
median sulfite intake from ingestion of food additives. In children (<10 years), assuming
similar levels of sulfite intake as adults, sulfite intake from inhalation (75 ppb SO2,
24 hours, 97.5th percentile ventilation) is approximately the same as median sulfite intake
from ingestion of food additives. However, ingested sulfite absorbed into the blood goes
directly to the liver where much of it is metabolized into sulfate. The majority of sulfite
derived from inhalation that enters the blood is rapidly distributed [as either sulfite or
S-sulfonate (Yokovama et al.. 1971; Balchum et al.. 1959)1 about the body with around a
quarter of total blood flow going to the liver (ICRP. 2002) where there is a high activity
of sulfite oxidase compared to other tissues. For lower exposure concentrations and
durations than considered above, sulfite (and/or S-sulfonate) levels in the blood following
SO2 inhalation could exceed those from ingestion of food additives, particularly in
children.
In summary, exogenous sources contribute hundreds of times lower amounts of sulfite
than the catabolism of sulfur-containing amino acids, when averaged across the entire
body. Sulfite and sulfate derived from the catabolism of sulfur-containing amino acids
are distributed broadly and do not accumulate in respiratory tract tissues. Following
ingestion of sulfite-containing food additives, sulfite enters the circulation and is subject
to first pass clearance in the liver where it is metabolized to sulfate. Following inhalation,
a substantial portion of SCh-derived products accumulate and are retained within the
respiratory tract. SCh-derived products that enter the circulation are rapidly distributed
throughout the body, appear primarily in the liver, and are excreted via the urine
(Section 4.2.5).
4.3 Mode of Action of Inhaled Sulfur Dioxide
This section describes the biological pathways that potentially underlie health effects
resulting from short-term and long-term exposure to SO2. Extensive research carried out
over several decades in humans and in laboratory animals has yielded much information
about these pathways. This section is not intended to be a comprehensive overview, but
rather, it updates the basic concepts derived from the SO2 literature presented in the 1982
AQCD (U.S. EPA. 1982a) and the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d) and
introduces the recent relevant literature. While this section highlights findings of studies
published since the 2008 SOx ISA (U.S. EPA. 2008d). earlier studies that represent the
current state of the science are also discussed. Studies conducted at more environmentally
relevant concentrations of SO2 (i.e., <2 ppm, see Section 1.2) are of greater interest
because biological pathways responsible for effects at higher concentrations may not be
identical to those occurring at lower concentrations. Some studies at higher
concentrations are included if they were early demonstrations of key biological pathways
4-17
-------
or if they are recent demonstrations of potentially important new pathways. This
information is used to develop a mode of action framework for inhaled SO2 that serves as
a guide to interpreting health effects evidence presented in Chapter 5.
Mode of action refers to a sequence of key events, endpoints, and outcomes that result in
a given toxic effect (U.S. EPA. 2005a'). Elucidation of mechanism of action provides a
more detailed understanding of key events, usually at the molecular level (U.S. EPA.
2005a'). The framework developed in this chapter will include some mechanistic
information on initiating events at the molecular level, but will mainly focus on the
effects of SO2 at the cellular, tissue, and organism level.
SO2 is a highly reactive gas. At physiologic pH, its hydrated forms include sulfurous
acid, bisulfite, and sulfite, with the latter species predominating. Sulfite is a strong
nucleophilic anion that readily reacts with nucleic acids, proteins, lipids, and other classes
of biomolecules. It participates in many important types of reactions including
sulfonation (sulfitolysis) and auto-oxidation with the generation of free radicals. This
latter reaction may be responsible for the induction of oxidative stress that occurs as a
result of exposure to SO2.
As described in Section 4.2. SO2 is a water-soluble gas that is absorbed almost entirely in
the upper respiratory tract. However, under conditions of mouth breathing and exercise,
some SO2 may penetrate to the tracheobronchial region. The main effects of SO2
inhalation are seen at the sites of absorption (i.e., the respiratory tract) and include
(1) activation of sensory nerves in the respiratory tract resulting in neural reflex
responses, (2) injury to airway mucosa, and (3) increased airway responsiveness and
allergic inflammation. Effects outside the respiratory tract may occur at very high
concentrations of inhaled SO2. Biologic pathways involved in mediating these responses
to inhaled SO2 are discussed below. In addition, a brief synopsis of pathways involved in
mediating the effects of endogenous SCh/sulfite is presented.
4.3.1 Activation of Sensory Nerves in the Respiratory Tract
SO2 is classified as a sensory (or nasal) irritant in mice, guinea pigs, rats, and humans
(Alarie. 1973). As such, it may stimulate trigeminal nerve endings when inhaled by the
nose, which results in an inhibition of respiration. It may also stimulate trigeminal nerves
in the larynx, which results in coughing, and in the cornea, which induces tearing. Other
reflexes stimulated by trigeminal nerve endings include decreased heart rate, peripheral
vasoconstriction, closure of the glottis, closure of the nares, and increased nasal flow
resistance. These responses are variable among species. Increased nasal flow resistance
has been demonstrated in humans breathing SO2 gas through the nose. Furthermore,
4-18
-------
de sensitization of the respiratory rate response occurs with repeated exposure. Most
sensory (or nasal) irritants, including SO2, also cause bronchoconstriction, but at
concentrations higher than those stimulating nerve endings in the nose.
SO2 is also classified as a pulmonary (or bronchial) irritant that evokes reflex reactions
through effects on pulmonary nerve endings (Alaric. 1973). These reactions usually
include an increase in respiratory rate accompanied by a decrease in tidal volume,
sometimes preceded by coughing and brief apnea, and sometimes accompanied by
bronchoconstriction. These responses have been observed in guinea pigs and cats
breathing via a tracheal cannula, which bypasses the nose. In the cat, SO2 exposure
increased the activity of vagal afferent fibers by either stimulating or sensitizing
tracheobronchial receptors on the nerve endings. SO2 also increased airway resistance in
guinea pigs and humans breathing through the nose, mouth, and/or tracheal cannula.
Increased airway resistance may occur via a variety of mechanisms including
accumulation of secretions, inflammatory changes of the airway walls, collapsing
airways, and constrictions of the central and peripheral airways. Constriction may be due
to direct action on the smooth muscle, axonal reflexes, vagal nerve stimulation, and
release of mediators such as histamine.
Continuous or repeated exposure to inhaled SO2 has a different pattern of responses in
different species (Alaric. 1973). In guinea pigs, the increase in airway resistance rose to a
plateau upon exposure and decreased to baseline with cessation of exposure. In humans
and dogs, resistance increased with exposure but decreased after 10 minutes (humans) or
3 minutes (dogs) despite the continuous presence of the gas. Studies in adults with
asthma demonstrated a different pattern. When exposure to SO2 occurred during a
30-minute period with continuous exercise, the response to SO2 developed rapidly and
was maintained throughout the 30-minute exposure (Kchrl et al.. 1987; Linn et al.. 1987;
Linn et al.. 1984c). Sequential exposures in nonasthmatic human subjects and in cats
resulted in a decreased response to SO2 in the second exposure compared with the first,
indicative of desensitization.
Early experiments demonstrated that SC>2-induced reflexes were mediated by cholinergic
parasympathetic pathways involving the vagus nerve and inhibited by atropine (G run stein
et al.. 1977; Nadel et al.. 1965a. b). Bronchoconstriction was found to involve smooth
muscle contraction because (3-adrenergic agonists such as isoproterenol reversed the
effects. Rapid shallow breathing was observed in SC>2-exposed tracheotomized cats
(bypassing the nose). Histamine was proposed to play a role in SCh-induced
bronchoconstriction (U.S. EPA. 1982a). but this hypothesis remains unsupported.
Hydrogen ions, sulfurous acid, sulfite, and bisulfite are all putative mediators of the
reflex responses (Gunnison et al.. 1987a). In particular, sulfite-mediated sulfitolysis of
4-19
-------
disulfides present in receptor proteins on sensory nerve fibers has been postulated
because S-sulfonate formation may potentially disrupt protein structure or function
(Alarie. 1973).
More recent experiments in animal models conducted since 1982 have demonstrated that
both cholinergic and noncholinergic mechanisms may be involved in SC>2-induced
effects. In two studies using bilateral vagotomy, vagal afferents were found to mediate
the immediate ventilatory responses to SO2 (Wang et al.. 1996). but not the prolonged
bronchoconstrictor response (Barthelemv et al.. 1988). Other studies showed that atropine
failed to block SC>2-induced bronchoconstriction, and that a local axon reflex resulting in
C-fiber secretion of neuropeptides (i.e., neurogenic inflammation) was responsible for the
effect (Haii et al.. 1996; Atzori et al.. 1992). Neurogenic inflammation has been shown to
play a key role in animal models of airway inflammatory disease (Groncbcrg et al..
2004). Furthermore, in isolated perfused and ventilated guinea pig lungs,
bronchoconstriction to SO2 was biphasic. The initial phase was mediated by a local axon
reflex involving the release of the neuropeptide calcitonin gene-related peptide from
sensory nerves, while the later phase involved other mechanisms (Banncnbcrg et al..
1994).
In humans, the mechanisms responsible for SC>2-induced bronchoconstriction are not
entirely understood. In nonasthmatic subjects, near complete attenuation of
bronchoconstriction has been demonstrated using the anticholinergic agents atropine and
ipratropium bromide (Yildirim et al.. 2005; Snashall and Baldwin. 1982; Tan et al..
1982). However, in asthmatic subjects, these same anticholinergic agents (Field et al..
1996; Myers et al.. 1986a). as well as short- and long-acting (32-adrenergic agonists
(Gong et al.. 1996; Linnetal.. 1988). theophylline (Koenig et al.. 1992). cromolyn
sodium (Myers et al.. 1986a). neodocromil sodium (Bigbv and Boushev. 1993). and
leukotriene receptor antagonists (Gong et al.. 2001; Lazarus et al.. 1997) only partially
blocked SC>2-induced bronchoconstriction. That none of these therapies have been shown
to completely attenuate the effects of SO2 implies the involvement of both
parasympathetic pathways and inflammatory mediators in asthmatic individuals. Strong
evidence of this was borne out in a study by Myers et al. (1986a) in which asthmatic
adults were exposed to SO2 following pretreatment with cromolyn sodium (a mast cell
stabilizer), atropine (a muscarinic receptor antagonist), and the two medications together.
While both treatments individually provided some protection against the
bronchoconstrictive effects of SO2, there was a much stronger and statistically significant
effect following concurrent administration of the two medications. Besides mast cell
stabilization, cromolyn sodium may also reduce the activity of lung irritant receptors
(Harries et al.. 1981). providing an alternative mechanism for the reduction in
SCh-induced bronchoconstriction observed.
4-20
-------
It has been proposed that inflammation contributes to the enhanced sensitivity to SO2
seen in asthmatic human subjects by altering autonomic responses (Tunnicliffc ct al..
2001). enhancing mediator release (Tan et al.. 1982). and/or sensitizing C-fibers and
rapidly adapting receptors (Lee and Widdicombe. 2001). Whether local axon reflexes
also play a role in SCh-induced bronchoconstriction in asthmatic individuals is not known
(Groneberg et al.. 2004; Widdicombe. 2003; Lee and Widdicombe. 2001). However,
differences in respiratory tract innervation between rodents and humans suggest that
C-fiber-mediated neurogenic inflammation may be unimportant in humans (Groneberg et
al.. 2004; Widdicombe. 2003; Widdicombe and Lee. 2001). Furthermore, enhanced
sensitivity to SO2 in asthmatic individuals may be related to genetic polymorphisms of
inflammatory mediators, such as TNF-a (Winterton et al.. 2001).
Studies conducted in vitro provide support for SO2 exposure-mediated effects that
involve inflammatory cells. It is known that sulfite exposure of cultured rat basophil
leukemia cells, a mast cell analog, causes immunoglobulin E (IgE)-independent
degranulation, release of histamine, serotonin and other mediators, and intra-cellular
production of reactive oxygen species (Collaco et al.. 2006). In addition, peroxidases,
such as neutrophil myeloperoxidase, oxidize bisulfite anion to several radical species that
in turn attack proteins (Ranguelova et al.. 2013; Ranguelova et al.. 2012). This represents
a potentially important new toxicological pathway for sulfite, especially in the presence
of neutrophilic and/or eosinophilic inflammation.
Irritant responses are indicative of a chemical's ability to damage the respiratory tract
(Alarie and Luo. 1986; Alarie. 1981). In the case of sensory irritation, there is a
characteristic decrease in respiratory rate, which is often used to set standards for
occupational exposures. Chemicals that are pulmonary irritants often lead to rapid
shallow breathing. They typically induce pulmonary edema or congestion if inhaled for a
long enough period of time. Some chemicals are both sensory and pulmonary irritants
and pulmonary irritation may occur at concentrations below which sensory irritation
occurs. In the case of SO2, a concentration-dependent hierarchy of effects has been noted
in humans (Kane et al.. 1979). Lethal or extremely severe injury to the respiratory tract
has been reported at and above 190 ppm. Intolerable sensory irritation and respiratory
tract injury that may occur with extended exposure has been associated with
10-15-minute exposures to 30-100 ppm SO2, and tolerable sensory irritation has been
associated with 10-minute exposures to 5-11.5 ppm SO2. Minimal sensory irritation has
been associated with exposures at and below 1 ppm. Increased airway resistance, likely
due to pulmonary irritation and reflex bronchoconstriction, has been observed in adults
without asthma at 5 ppm while at rest and at 1 ppm SO2 while exercising (Arts et al..
2006). However, lung function changes have been observed at concentrations of SO2
lower than 1 ppm in exercising adults with asthma. Thus, pulmonary irritation may occur
4-21
-------
at levels of SO2 below those that cause sensory irritation, especially in exercising adults
with asthma.
In summary, SO2 acts as both a sensory and a pulmonary irritant through activation of
sensory nerves in the respiratory tract resulting in neural reflex responses. This occurs in
a variety of species, including humans. Pulmonary irritant responses due to SO2 exposure
result in reflex bronchoconstriction, especially in adults with asthma. Both cholinergic
parasympathetic pathways involving the vagus nerve and inflammation contribute to
reflex bronchoconstriction in asthmatic individuals.
4.3.2 Injury to Airway Mucosa
A common feature of irritant gases, including SO2, is the capacity to injure airway
mucosa, resulting in decreased epithelial barrier function, inflammation, and
compromised ciliary function (Carson et al.. 2013). Despite being the initial site of SO2
absorption and having low activity of sulfite oxidase, the respiratory tract of healthy
humans is thought to be capable of detoxifying 5 ppm inhaled SO2 (Gunnison et al..
1987a). In fact, exposure to 0.5-2 ppm SO2 for 4 hours did not result in any measurable
changes in biomarkers of oxidative stress or inflammation in exhaled breath condensate
(EBC) or nasal lavage fluid (NALF) from healthy adults subjected to two periods of
moderate exercise (Raulf-Hcimsoth et al.. 2010). In addition, no changes in nasal lining
fluid ascorbic acid or uric acid levels were observed following 1-hour exposure of adults
with asthma to 0.2 ppm SO2 (Tunnicliffc et al.. 2003).
However, respiratory tract injury has been observed in humans exposed for extended
periods to SO2 concentrations of 30 ppm and greater. In animal models, airway injury and
histopathological changes, such as mucous cell metaplasia and intra-mural fibrosis, have
generally been observed following chronic exposure to SO2 concentrations of 10 ppm and
higher (U.S. EPA. 2008d). Rats exposed to 20 ppm SO2 for several weeks exhibit fibrotic
remodeling of airway epithelium and mucus hypersecretion, key features of chronic
obstructive pulmonary disease and chronic asthma in humans (Wagner et al.. 2006).
Inflammatory changes have been noted in some animal models following subacute
exposure to 5-100 ppm SO2 (U.S. EPA. 2008d). Impaired mucociliary clearance has also
been demonstrated at high concentrations of SO2. In humans, nasal mucus flow was
decreased during a 5-hour exposure to 5 and 25 ppm SO2 (Gunnison et al.. 1981).
Impaired mucus flow in the trachea has been observed in rats exposed subacutely to
11.4 ppm SO2 and in dogs exposed chronically to 1 ppm SO2 (Gunnison et al.. 1981;
Hirsch et al.. 1975). Whether these effects were due to compromised ciliary function or
4-22
-------
altered properties of the mucus due to sulfite-mediated sulfitolysis of disulfide bonds in
mucus was not investigated.
Recent studies provide additional insight. An ultrastructural examination of nasal biopsy
tissue by freeze fracture microscopy was conducted in healthy humans exposed to
0.75 ppm SO2 for 2 h (Carson et al.. 2013). Evidence of fragmentation of the tight
junctional complex and polymorphonuclear infiltrate was reported although no effects on
ciliary membranes were observed. These subtle responses suggest a slight decrease in
barrier function due to acute SO2 exposure at this level. Furthermore, a subacute exposure
of rats to 2.67 ppm SO2 (6 hours/day, 7 days) resulted in altered lung mRNA levels for
inducible nitric oxide synthase (involved in inflammation) and for bax (or B-cell
lymphoma 2-like protein 4; involved in regulating apoptosis) (Sang et al.. 2010). In this
study, gene expression changes were also found in the heart and they were more
pronounced than in the lung. These results suggest that, despite low sulfite oxidase
activity, the respiratory tract may be more resistant than the heart to the effects of inhaled
S02.
In summary, exposure to SO2 results in injury to airway mucosa, especially at higher
concentrations and following extended periods of exposure. There is little evidence of
injury or inflammation in response to acute exposures to concentrations of 2 ppm SO2 or
less in healthy human subjects. However, one new study found subtle histopathological
changes at the ultrastructural level following a 2-hour exposure to 0.75 ppm SO2. New
evidence also suggests subtle changes in the lung related to inflammation and apoptosis
in rats exposed over several days to 2.67 ppm SO2.
4.3.3 Modulation of Airway Responsiveness and Allergic Inflammation
Respiratory irritants, including SO2, are thought to be a major cause of occupational
asthma (Baur et al.. 2012; Andersson et al.. 2006). Both peak high-level exposures and
low-level persistent exposures to respiratory irritants have been associated with the
development of asthma. The propensity for airways to narrow following inhalation of
some stimuli is termed airway responsiveness. The term airway hyperresponsiveness
(AHR) is generally used in cases where airway responsiveness to methacholine or
histamine is assessed and the provocative concentration is determined to be sufficiently
low to classify the subjects as having AHR based on criteria such as ATS (2000a'). Along
with symptoms, variable airway obstruction, and airway inflammation, AHR is a primary
feature in the clinical definition and characterization of asthma severity (Reddel et al..
2009).
4-23
-------
Studies in several different animal species have shown that a single exposure to SO2 at a
concentration of 10 ppm or less failed to increase airway responsiveness following a
challenge agent (U.S. EPA. 2008d). However, in an animal model of allergic airway
disease, SO2 exposure enhanced airway responsiveness. In this study, sheep previously
sensitized and challenged with Ascaris suum extract were exposed to 5 ppm SO2 for
4 hours (Abraham et al.. 1981). Airway responsiveness to carbachol was increased at
24 hours, but not immediately, after SO2 exposure. This response was not observed in
sheep that had not been sensitized and challenged with Ascaris suum extract. The
mechanism underlying the S02-mediated increases in airway responsiveness was not
investigated in this study. However, this response could have resulted from sensitization
of vagal irritant receptors, greater sensitivity of smooth muscle to bronchoconstriction
agents, or enhanced concentrations of bronchoconstriction agents reaching the receptors
or bronchial smooth muscle. The delayed nature of the response points to a possible role
of inflammation in increasing airway responsiveness.
Two controlled human exposure studies in adults with asthma provide further evidence of
increased responsiveness to an allergen when exposure to SO2 was in combination with
nitrogen dioxide (NO2). In one of these studies, exposure to 0.2 ppm SO2 or 0.4 ppm NO2
alone did not affect airway responsiveness to house dust mite allergen immediately after
a 6-hour exposure at rest (Devalia et al.. 1994). However, following exposure to the two
pollutants in combination, subjects demonstrated an increased response to the inhaled
allergen. Rusznak et al. (1996) confirmed these results in a similar study and found that
increased responsiveness to dust mites persisted up to 48 hours post-exposure. These
results provide evidence that exposure to SO2 and NO2 in combination elicits an increase
in airway responsiveness to an allergen. This effect is longer in duration than other
effects typically associated with exposure to SO2.
Several other studies have examined the effects of SO2 exposure on allergic
inflammation. One of these was a controlled human exposure study of adults with
asthma. Subjects were exposed for 10 minutes to 0.75 ppm SO2 while exercising at a
moderate level (Gong et al.. 2001). In addition to changes in lung function and
symptoms, there was a statistically significant increase in eosinophil count in induced
sputum 2 hours post-exposure. Pretreatment with a leukotriene receptor antagonist
dampened these responses, implicating a role for leukotrienes in mediating SO2
exposure-induced effects.
The other studies investigated the effects of repeated exposure to SO2 on inflammatory
and immune responses in an animal model of allergic airways disease. Li et al. (2007)
demonstrated that exposure of ovalbumin-sensitized rats to 2 ppm SO2 for 1 hour
followed by challenge with ovalbumin each day for 7 days resulted in an increased
4-24
-------
number of inflammatory cells in bronchoalveolar lavage fluid (BALF) and an enhanced
histopathological response compared with rats treated with SO2 or ovalbumin alone.
Similarly, inter-cellular adhesion molecule 1 (ICAM-1), a protein involved in regulating
inflammation, and mucin 5AC glycoprotein (MUC5AC), a mucin protein, were
upregulated in lungs and trachea to a greater extent in rats treated both with SO2 and
ovalbumin. A follow-up study involving the same exposure regimen (2 ppm SO2 for
1 hour) in the same allergic animal model (rats sensitized and challenged with
ovalbumin) also found that repeated SO2 exposure enhanced inflammatory and allergic
responses to ovalbumin (Li et al.. 2014). Numbers of eosinophils, lymphocytes, and
macrophages were greater in the BALF of SC>2-exposed and ovalbumin-treated animals
than in animals treated only with ovalbumin. In addition, SO2 exposure enhanced
upregulation and activation of nuclear factor kappa-light-chain-enhancer of activated
B cells (NFkB), a transcription factor involved in inflammation, and upregulation of the
cytokines interleukin-6 (IL-6) and interleukin-4 (IL-4) in lung tissue. Furthermore, BALF
levels of IL-6 and IL-4 were increased to a greater extent in SC>2-exposed and
ovalbumin-treated animals compared with ovalbumin treatment alone. These results
indicate that repeated SO2 exposure enhanced activation of the NFkB inflammatory
pathway and upregulation of inflammatory cytokines in ovalbumin-treated animals.
Furthermore, SO2 exposure enhanced the effects of ovalbumin on levels of inter-feron
gamma (IFN-y) (decreased) and IL-4 (increased) in BALF and on IgE levels in serum
(increased). Because levels of IL-4 are often indicative of T helper 2 (Th2) status and
levels of IFN-y are indicative of a T helper 1 (Thl) status, these results suggest a shift in
Thl/Th2 balance towards Th2 in rats made allergic to ovalbumin, an effect that was
exacerbated by SO2 exposure. These Th2-related changes are consistent with the
observed increases in serum IgE and BALF eosinophils in ovalbumin-treated animals,
effects that were also enhanced by SO2 exposure. Taken together, these results indicate
that repeated exposure to SO2 exacerbated inflammatory and allergic responses in this
animal model. It should be noted, however, that Group 2 innate lymphoid cells can
mediate Type 2 immunity, as has been described for 03-mediated responses in mice (Ong
et al.. 2016). Whether Group 2 innate lymphoid cells mediate effects of inhalation of SO2,
which like O3 is an irritant gas, is unexplored.
Two other follow-up studies by the same laboratory examined the effects of inhaled SO2
on the asthma-related genes encoding epidermal growth factor (EGF), epidermal growth
factor receptor (EGFR), and cyclooxygenase-2 (COX-2), and on apoptosis-related genes
and proteins in this same model based on sensitization with ovalbumin (Xie et al.. 2009;
Li et al.. 2008). While EGF and EGFR are related to mucus production and airway
remodeling, COX-2 is related to apoptosis and may play a role in regulating airway
inflammation. SO2 exposure enhanced the effects of ovalbumin in this model, resulting in
greater increases in mRNA and protein levels of EGF, EGFR, and COX-2 in the trachea
4-25
-------
compared with ovalbumin treatment alone. SO2 exposure enhanced other effects of
ovalbumin in this model, resulting in a greater decline in mRNA and protein levels of
tumor protein p53 (p53) and bax and a greater increase in mRNA and protein levels of
B-cell lymphoma 2 (bcl-2) in the lungs compared with ovalbumin challenge alone. The
increased ratio of bcl-2:bax, an indicator of susceptibility to apoptosis, observed
following ovalbumin challenge, was similarly enhanced by SO2. Thus, repeated exposure
to SO2 may impact numerous processes involved in inflammation and/or airway
remodeling in allergic airway disease.
The effects of repeated SO2 exposure on the development of an allergic phenotype and
altered physiologic responses in naive animals was examined in two studies in which SO2
exposure preceded allergen sensitization. Repeated exposure of guinea pigs to SO2
promoted allergic sensitization and subsequently enhanced allergen-induced bronchial
obstruction, as reported by U.S. EPA (2008d). Riedel et al. (1988) examined the effect of
SO2 exposure on local bronchial sensitization to inhaled antigen. Guinea pigs were
exposed by inhalation to 0.1, 4.3, and 16.6 ppm SO2 for 8 hours/day for 5 days. During
the last 3 days, SO2 exposure was followed by exposure to nebulized ovalbumin for
45 minutes. Following bronchial provocation with inhaled ovalbumin (0.1%) 1 week
later, bronchial obstruction was measured by examining the respiratory loop obtained by
whole-body plethysmography. In addition, specific antibodies against ovalbumin were
measured in serum and BALF. Results showed significantly higher bronchial obstruction
in animals exposed to SO2 (at all concentration levels) and ovalbumin, compared with
animals exposed only to ovalbumin. In addition, significant increases in antiovalbumin
immunoglobulin G (IgG) antibodies were detected in BALF lavage fluid of animals
exposed to 0.1, 4.3, and 16.6 ppm SO2 and in serum from animals exposed to 4.3 and
16.6 ppm SO2 compared with controls exposed only to ovalbumin. These results
demonstrate that repeated exposure to SO2 enhanced allergic sensitization in the guinea
pig at a concentration as low as 0.1 ppm. In a second study, guinea pigs were exposed to
0.1 ppm SO2 for 5 hours/day for 5 days and sensitized with 0.1% ovalbumin aerosols for
45 minutes on Days 4 to 5 (Park et al.. 2001). One week later, animals were subjected to
bronchial challenge with 0.1% ovalbumin and lung function was evaluated 24 hours later
by whole-body plethysmography. Results demonstrated a significant increase in
enhanced pause, a measure of airway obstruction, in animals exposed to SO2 and
ovalbumin but not in animals treated with ovalbumin or SO2 alone. Results also
demonstrated increased numbers of eosinophils in lavage fluid and an infiltration of
inflammatory cells, bronchiolar epithelial cell damage, and plugging of the airway lumen
with mucus and cells in the bronchial tissues of animals treated with both SO2 and
ovalbumin, but not in animals treated with ovalbumin or SO2 alone. These experiments
indicate that repeated exposure to near ambient levels of SO2 plays a role in allergic
sensitization and also exacerbates allergic inflammatory responses in the guinea pig.
4-26
-------
Furthermore, increases in bronchial obstruction observed in both studies suggest that
repeated SO2 exposure increased airway responsiveness.
Longer term exposure of naive newborn rats to SO2 (2 ppm, 4 hours/day for 28 days)
resulted in altered cytokine levels that suggest a shift in Thl/Th2 balance towards Th2
(Song et al.. 2012). Th2 polarization is one of the steps involved in allergic sensitization.
It should be noted, however, that Group 2 innate lymphoid cells can mediate Type 2
immunity, as has been described for 03-mediated responses in mice (Ong et al.. 2016).
Whether Group 2 innate lymphoid cells mediate effects of inhalation of SO2, which like
O3 is an irritant gas, is unexplored. In naive animals exposed to SO2, levels of IL-4,
which is indicative of a Th2 response, were increased and levels of IFN-y. indicative of a
Thl response, were decreased in BALF. In ovalbumin-sensitized newborn rats, SO2
exposure resulted in a greater enhancement of lavage fluid IL-4 and an increase in serum
IL-4 levels compared with ovalbumin-sensitization alone. In addition, SO2 exposure led
to increased bronchial obstruction and airway remodeling, as indicated by increased
content of airway smooth muscle, in the ovalbumin-sensitized animals. Stiffness and
contractility of airway smooth muscle was assessed in vitro using cells from
experimentally treated animals. In allergic rats, both stiffness and contractility were
increased as a result of SO2 exposure, suggesting an effect on the biomechanics of airway
smooth muscle. This study provides evidence for allergic sensitization by SO2 in naive
newborn rats and for enhanced allergic inflammation, increased airway responsiveness,
and airway remodeling in SCh-exposed allergic newborn rats.
Supportive evidence that SO2 may promote allergic sensitization is provided by a study in
mice that were first treated with sodium sulfite and then sensitized and challenged with
house dust mite allergen (Lin etal.. 2011a). Sulfite is formed in ELF following inhalation
of SO2 (Section 4.2.1). Repeated intra-nasal treatment with 10 (iL of a 5-mM solution of
sodium sulfite aggravated inflammation (measured by histopathology) and allergic
sensitization in this model. Specific IgE levels were higher in sulfite-treated and
allergen-challenged animals compared with either sulfite treatment or allergen challenge
alone. Specific IgG2a levels, indicative of a Thl response, were decreased as a result of
sulfite treatment in house dust mite-challenged mice. In addition, inter-leukin-5 (IL-5)
levels, indicative of a Th2 response, and the ratio of Il-5:IFN-y, a marker of Th2
polarization, were higher in lung tissue from sulfite-treated and allergen-challenged mice
compared with either sulfite treatment or allergen challenge alone.
Mixtures of SO2 and other criteria pollutants have also been shown to modulate airway
responsiveness and/or allergic inflammation. As discussed above, increased airway
responsiveness to house dust mite allergen occurred in human subjects with mild allergy
and asthma immediately following 6 hours of concurrent exposure to 0.2 ppm SO2 and
4-27
-------
0.4 ppm NO2, but not to either pollutant alone (Rusznak et al.. 1996; Dcvalia ct al..
1994). This effect persisted for 48 hours. Recently, the effects of simulated downwind
coal combustion emissions (SDCCE), which contains SO2, on allergic airway responses
was investigated in mice (Barrett et al.. 2011). Mice were sensitized and challenged with
ovalbumin and exposed for 6 hours/day for 3 days to several concentrations of SDCCE
with and without a particle filter. Concentration of SO2 in the highest exposure was
0.2 ppm. Other gases present in this exposure were NO2 (0.29 ppm), NO (0.59 ppm), and
carbon monoxide (0.02 ppm). SDCCE exposure was followed by another challenge with
ovalbumin in some animals. Results demonstrated that both the particulate and the
gaseous phases of SDCCE exacerbated allergic airways responses. Airway
responsiveness to methacholine (measured by the forced oscillation technique) was
enhanced by the gaseous phase of SDCCE in mice that were challenged with ovalbumin
after SDCCE exposure. While results of this study are consistent with SO2 playing a role
in enhancing allergic responses and increases in airway responsiveness, a role for other
components in the mixture cannot be ruled out.
In summary, a growing body of evidence supports a role for SO2 in increasing airway
responsiveness and/or allergic inflammation in animal models of allergic airway disease,
as well as in asthmatic individuals. Some responses in asthmatic individuals were
observed only when exposure to SO2 occurred in combination with NO2. In animal
studies, repeated or prolonged exposure to SO2 promoted allergic sensitization. One study
in newborn allergic rats suggested that airway remodeling may contribute to increases in
airway responsiveness following prolonged exposure to SO2.
4.3.4 Induction of Systemic Effects
As described in the 2008 SOx ISA (U.S. EPA. 2008d). two controlled human exposure
studies reported that acute exposure to 0.2 ppm SO2 resulted in changes in heart rate
variability in healthy adults and in asthmatic adults (Routledge et al.. 2006; Tunnicliffe et
al.. 2001). More recently, altered parasympathetic regulation of heart rate was reported in
rats exposed to 5 ppm SO2 during the peri-natal and post-natal period (Wocrman and
Mendelowitz. 2013a. b). Whether these responses were due to activation of sensory
nerves in the respiratory tract resulting in a neural reflex response and altered autonomic
function or some other mechanism is not known.
Numerous studies over several decades have reported other extrapulmonary effects of
inhaled SO2 (U.S. EPA. 2008d'). Most of these occur at concentrations far higher than
those measured in ambient air. As discussed in Section 4.2.3. studies in mice and humans
demonstrating the presence of sulfite and S-sulfonates in blood and tissues outside of the
4-28
-------
respiratory tract point to the likely role of circulating sulfite in mediating these responses.
A subacute study measured sulfite plus S-sulfonate content of the lung, liver, and brain of
mice exposed to 5, 10, or 20 ppm SO2 for 4 hours/day for 7 days (Meng et al.. 2005a) and
found a concentration-dependent increase. Similarly, exposure of human subjects to
0.3-6 ppm SO2 for up to 120 hours resulted in the appearance in the plasma of sulfite
plus S-sulfonates (Gunnison and Palmes. 1974). The relationship between
sulfite/sulfonate concentration and chamber SO2 concentration was linear (regression
coefficient of 0.61) with a slope of 1.1 nmol/mL of plasma S-sulfonate for each 1-ppm
increment in SO2 concentration. These results indicate that prolonged (i.e., hours to days)
exposure to as low as 0.3 ppm SO2 results in measurable amounts of circulating sulfite in
humans. The relationship between circulating sulfite/S-sulfonate and extrapulmonary
effects of inhaled SO2 has not yet been explored in human subjects.
Because the activity of sulfite oxidase is variable among species, the degree of sensitivity
to SCh-mediated effects is likely to be variable among species. For example, sulfite
oxidase in rats is 10-20 times greater than in humans and 3-5 times greater than in
rabbits or rhesus monkeys (Gunnison et al.. 1987a; Gunnison. 1981). Thus, the toxicity of
SO2 may be less in rats due to more rapid metabolism of sulfite to sulfate.
Systemic effects are likely due to oxidative stress, possibly from sulfite auto-oxidation.
Alternatively, sulfite-mediated S-sulfonate formation may disrupt protein function, and
metabolic reduction of S-sulfonates may alter reduction-oxidation (redox) status.
Moreover, sulfite may serve as a substrate for peroxidases, such as myeloperoxidase and
eosinophil peroxidase, to produce free radicals, as has been demonstrated in neutrophils
and eosinophils (Ranguelova et al.. 2013; Ranguelova et al.. 2012; Ranguelova et al..
2010). These sulfur-based free radical species may then initiate protein or lipid oxidation.
Baskurt (1988) found that exposure of rats to 0.87 ppm SO2 for 24 hours resulted in
increased hematocrit, sulfhemoglobin, and osmotic fragility, as well as decreased whole
blood and packed cell viscosities. These results indicate a systemic effect of inhaled SO2
and are consistent with an oxidative injury to red blood cells. Other studies have reported
lipid peroxidation in erythrocytes and tissues of animals exposed to SO2 (Qin et al.. 2012;
Ziemann et al.. 2010; Haider etal.. 1982). Supplementation with ascorbate and
a-tocopherol decreased SCh-induced lipid peroxidation in erythrocytes (Etlik et al..
1995). Additionally, recent studies report mitochondrial changes in the hearts and brains
of rats exposed to 1.34 ppm (4 hours/day) SO2 for several weeks (Qin et al.. 2016; Qin et
al.. 2012). Demonstration of mitochondrial biogenesis in rat brain suggests that SO2
exposure induces an adaptive response to oxidative stress (Qin et al.. 2012). Changes in
cardiac function were observed at higher concentrations (2.7 ppm SO2); pretreatment
with antioxidants blocked this effect (Qin et al.. 2016). Other recent studies report altered
4-29
-------
markers of brain inflammation and synaptic plasticity following several weeks to months
of exposure to 1.34 ppm (4 hours/day) SO2 (Yao etal.. 2015; Yao et al.. 2014). Further
studies are required to confirm that inhalation exposures of SO2 at or near ambient levels
increase blood sulfite levels sufficiently for oxidative injury to occur in blood cells or
other tissues.
In summary, exposure to SO2 may result in effects outside the respiratory tract via
activation of sensory nerves in the respiratory tract resulting in a neural reflex response or
mediated by circulating sulfite. A few studies employing concentrations of 2 ppm SO2 or
less have demonstrated effects that are consistent with sulfite-mediated redox stress, such
as increased sulfhemoglobin in red blood cells and lipid peroxidation in the brain. Recent
studies also suggest possible inflammation and other effects in tissues distal to the
absorption site following several weeks to months of exposure to 1.34 ppm SO2.
4.3.5 Role of Endogenous Sulfur Dioxide/Sulfite
Endogenous SCh/sulfite is a product of normal metabolism of sulfur-containing amino
acids (e.g., cysteine and methionine) (Liu et al.. 2010). While SO2 gas is measured in the
head space gas of preparations of various tissues or bodily fluids (Balazv et al.. 2003).
sulfite/bisulfite is measured in soluble fractions. The distribution of SO2 and enzymes
responsible for SO2 generation has been reported in tissues of the rat (Luo etal.. 2011).
Chemical transformations between bisulfite/sulfite/SCh and the gasotransmitter H2S also
occur. H2S is similarly derived from sulfur-containing amino acids. Evidence has
accumulated that endogenous H2S acts as a biological signaling molecule (Filipovic et al..
2012) and plays important roles in the cardiovascular (Coletta et al.. 2012) and other
systems. Recent studies suggest that endogenous SChmay also be a gasotransmitter (Liu
et al.. 2010). Like the other gasotransmitters NO and CO, SO2 at physiologic levels may
activate guanylyl cyclase to generate cyclic guanosine monophosphate (cGMP), which
mediates effects through cGMP-dependent kinases (Li et al.. 2009). However, SO2 may
also act through non-cGMP-dependent pathways. Experimental studies in animal models
and in vitro systems demonstrate myriad effects of exogenous SO2 on the cardiovascular
system, including vasorelaxation, negative inotropic effects on cardiac function,
anti-inflammatory and antioxidant effects, and decreased blood pressure (BP) and
vascular remodeling in hypertensive animals (Liu et al.. 2010). Effects were
concentration dependent in many cases. In vivo studies generally were conducted using
5 ppm and higher concentrations of SO2 (or sulfite/bisulfite) (Liu et al.. 2010). In
summary, endogenous SO2 is a newly recognized gasotransmitter that may play
important roles in cardiovascular and other systems.
4-30
-------
4.3.6 Mode of Action Framework
This section describes the key events, endpoints, and outcomes that comprise the modes
of action of inhaled SO2. Here, key events are subclinical effects, endpoints are effects
that are generally measured in the clinic, and outcomes are health effects at the organism
level. Biological pathways discussed above that may contribute to health effects resulting
from short-term and long-term exposures to SO2 (Chapter 5) are summarized as a part of
this analysis. These proposed modes of action are based on the available evidence and
may not reflect all of the pathophysiology underlying health effects.
Figure 4-2 depicts the proposed mode of action for respiratory effects due to short-term
exposure to SO2.
so,
Legend
~ Pollutant
¦] Key Events
~ Endpoints
~ Outcomes
trigger
Bronchoconstriction
Asthma
exacerbation
/I" Inflammatory
mediators
Increased airway
responsiveness
Activation/
Sensitization of
neural reflexes
^Allergic
inflammation/
Allergic
sensitization
Formation of
sulfite In ELF/
Redox reactions
and formation of
sulfitolysis and/or
other products
ELF = epithelial lining fluid; redox = reduction-oxidation; S02 = sulfur dioxide.
Note: Pathways indicated by a dotted line are those for which evidence is limited to findings from experimental animal studies, while
evidence from controlled human exposure studies is available for pathways indicated by a solid line. Dashed lines indicate proposed
links to the outcomes of asthma exacerbation. Key events are subclinical effects, endpoints are effects that are generally measured
in the clinic, and outcomes are health effects at the organism level.
Source: National Center for Environmental Assessment,
Figure 4-2 Summary of evidence for the proposed mode of action linking
short-term exposure to sulfur dioxide and respiratory effects.
The propensity for airways to narrow following inhalation of some stimuli is termed
airway responsiveness. A characteristic feature of individuals with asthma is an increased
propensity of their airways to narrow in response to bronchoconstrictive stimuli relative
to nonatopic individuals without asthma. Different kinds of stimuli can elicit
4-31
-------
bronchoconstriction, but in general they act on airway smooth muscle receptors (direct
stimuli, e.g., methacholine) or act via the release of inflammatory mediators (indirect
stimuli, e.g., allergens) (O'Bvrne et al.. 2009). SO2 is a nonspecific bronchoconstrictive
stimuli that is not easily classified as a direct or indirect stimuli, as was discussed in
Section 4.3.1.
Because inhalation of SO2 results in chemical reactions in the ELF, the initiating event in
the development of respiratory effects is the formation of sulfite, sulfitolysis products,
hydrogen ion, and/or other products. Both sulfite and S-sulfonates have been measured in
tracheal and bronchial tissue as well as in tracheal washings of experimental animals
exposed to SO2. Reactive products formed as a result of SO2 inhalation are responsible
for a variety of downstream key events, which may include activation or sensitization of
sensory nerves in the respiratory tract resulting in neural reflex responses, release of
inflammatory mediators, and modulation of allergic inflammation or sensitization. These
key events may collectively lead to several endpoints, including bronchoconstriction and
increased airway responsiveness. Bronchoconstriction is characteristic of an asthma
attack. However, individuals who are not asthmatic may also experience
bronchoconstriction in response to SO2 inhalation; generally, this occurs at higher
concentrations than in an individual who is asthmatic (>1 ppm). Additionally, SO2
exposure may increase airway responsiveness to subsequent exposures of other stimuli
such as allergens or methacholine. These pathways may be linked to the epidemiologic
outcome of asthma exacerbation.
The strongest evidence for this mode of action comes from controlled human exposure
studies. SO2 exposure resulted in increased airway resistance due to bronchoconstriction
in healthy adults and in adults with asthma. In adults without asthma, this response
occurred primarily as a result of activation of sensory nerves in the respiratory tract
resulting in neural reflex responses mediated by cholinergic parasympathetic pathways
involving the vagus nerve. However, in adults with asthma, evidence indicates that the
response is only partially due to vagal pathways and that inflammatory mediators such as
histamine and leukotrienes also play an important role. Activation of sensory nerves in
the respiratory tract, which result in neural reflex responses, has been studied in humans
exposed to occupationally relevant concentrations of SO2 (up to 2 ppm). Responses
measured in these studies include increased respiratory rate and decreased tidal volume,
which involve the vagus nerve, and increased nasal air-flow resistance, which involves
the trigeminal nerve. These responses are not a part of the mode of action described here,
but are mentioned because they are known irritant effects of SO2. Studies in experimental
animals demonstrate that SO2 exposure activates reflexes that are mediated by cholinergic
parasympathetic pathways involving the vagus nerve. However, noncholinergic
mechanisms may also play a role because some studies demonstrate that a local axon
4-32
-------
reflex resulting in C-flber secretion of neuropeptides (i.e., neurogenic inflammation) is
responsible for the effects of SO2.
Evidence demonstrates that SO2 exposure modulates allergic inflammatory responses.
Enhancement of allergic inflammation was observed in adults with asthma who were
exposed for 10 minutes to 0.75 ppm SO2 (i.e., leukotriene-mediated increases in numbers
of sputum eosinophils). In an animal model of allergic airway disease, repeated exposure
to 2 ppm SO2 led to an enhanced inflammatory response, as measured by numbers of
BALF inflammatory cells, levels of BALF cytokines, histopathology, activation of the
NFkB pathway, and upregulation of intra-cellular adhesion molecules, mucin, and
cytokines, in lung tissue. Furthermore, repeated exposure to SO2 enhanced Th2
polarization (or Group 2 innate lymphoid cell-mediated Type 2 immunity), numbers of
BALF eosinophils, and serum IgE levels in this same model. Other studies demonstrated
that repeated exposure of naive animals to SO2 (as low as 0.1 ppm) over several days
promoted allergic sensitization (allergen-specific IgG levels) and enhanced
allergen-induced bronchial obstruction (an indicator of increased airway responsiveness)
and inflammation (airway fluid eosinophils and histopathology) when animals were
subsequently sensitized and challenged with an allergen. Similarly, intra-nasal treatment
with sulfite both aggravated allergic sensitization (Th2 cytokines and allergen specific
IgE levels) and exacerbated allergic inflammatory responses (histopathology) in animals
subsequently sensitized and challenged with allergen. These changes in allergic
inflammation may enhance airway responsiveness and promote bronchoconstriction in
response to a trigger. Thus, allergic inflammation and increased airway responsiveness
may link short-term SO2 exposure to asthma exacerbation.
Figure 4-3 depicts the proposed mode of action for respiratory effects due to long-term
exposure to SO2.
4-33
-------
so,
Legend
Pollutant
B Key Events
¦ Endpoints
EJ Outcomes
Recurrent
redox stress
Allergic
sensitization
Airway
inflammation
Airway
remodeling
•a
Increased airway
responsiveness
trigger
New onset asthma/
Asthma exacerbation
Redox = reduction-oxidation; S02 = sulfur dioxide.
Note: Pathways indicated by a dotted line are those for which evidence is limited to findings from experimental animal studies, while
evidence from controlled human exposure studies is available for pathways indicated by a solid line. Dashed lines indicate proposed
links to the outcomes of new onset asthma/asthma exacerbation. Key events are subclinical effects, endpoints are effects that are
generally measured in the clinic, and outcomes are health effects at the organism level.
Source: National Center for Environmental Assessment,
Figure 4-3 Summary of evidence for the proposed mode of action linking
long-term exposure to sulfur dioxide and respiratory effects.
The initiating event in the development of respiratory effects due to long-term SO2
exposure is the recurrent or prolonged redox stress due to the formation of reactive
products in the ELF. This is the driving factor for the potential downstream key events,
airway inflammation, allergic sensitization, and airway remodeling that may lead to the
endpoint increased airway responsiveness. The term AHR is generally used in cases
where airway responsiveness to methacholine or histamine is assessed and the
provocative concentration is determined to be sufficiently low to classify the subjects as
having AHR based on criteria such as ATS (2000a). Airway inflammation, airway
remodeling, and AHR are characteristic of asthma. The resulting outcome may be new
asthma onset, which presents as an asthma exacerbation that leads to physician-diagnosed
asthma.
Evidence for this mode of action comes from studies in both naive and allergic
experimental animals. Exposure of naive newborn animals to SO2 (2 ppm) for several
weeks resulted in hyperemia in lung parenchyma, inflammation in the airways, and Th2
polarization (or Group 2 innate lymphoid cell-mediated Type 2 immunity), the latter of
which is a key step involved in allergic sensitization. Support is also provided by
short-term studies in naive animals in which repeated exposure to SO2 (2 ppm) over
several days led to pathologic changes, including inflammatory cell influx. Th2
polarization (or other Type 2 immune responses) and airway inflammation may set the
4-34
-------
stage for increases in airway responsiveness. In addition, short-term SO2 exposure
(0.1 ppm) promoted allergic sensitization, enhanced other allergic inflammatory
responses, and increased airway responsiveness when animals were subsequently
sensitized with an allergen. Further, repeated exposure of allergic newborn animals to
SO2 (2 ppm) over several weeks enhanced allergic responses and resulted in morphologic
responses indicative of airway remodeling and increased airway responsiveness. Thus,
repeated exposure to SO2 in naive animals may lead to the development of allergic
airway disease, which shares many features with asthma. Furthermore, repeated exposure
of allergic animals to SChmay promote airway remodeling and increased airway
responsiveness. Increased airway responsiveness in animal models and the development
of AHR in humans may link long-term exposure to SO2 to the epidemiologic outcome of
new onset asthma.
Figure 4-4 depicts the proposed mode of action for extrapulmonary effects due to
short-term or long-term exposure to SO2.
Although experimental studies have shown extrapulmonary effects resulting from SO2
inhalation (see Section 4.3.4). there is uncertainty regarding the mode of action
underlying these responses. Evidence from controlled human exposure studies (0.2 ppm,
1 hour) points to SO2 exposure-induced activation/sensitization of neural reflex responses
as a key event leading to the endpoint of altered heart rate or heart rate variability.
Evidence also points to transport of sulfite into the circulation. Controlled human
exposure and experimental animal studies have demonstrated the presence of sulfite and
S-sulfonates in plasma, liver, or brain following SO2 exposure. This occurred at a
concentration as low as 0.3 ppm SO2 in humans exposed for up to 120 hours. Sulfite is
highly reactive and may be responsible for redox stress (possibly through auto-oxidation
or peroxidase-mediated reactions to produce free radicals) in the circulation and
extrapulmonary tissues. However, this is likely to occur only at very high concentrations
or during prolonged exposures because circulating sulfite is efficiently metabolized to
sulfate in a reaction catalyzed by hepatic sulfite oxidase.
4-35
-------
so,
Formation of
sulfite in ELF/
redox reactions
and formation of
sulfitolysis products
Legend
Pollutant
~ Key Events
~ Endpoints
Transport of
sulfite into
circulation
Activation/
sensitization of
neural reflexes
Systemic
redox stress
Altered heart rate
and/or heart rate
variability
ELF = epithelial lining fluid; redox = reduction-oxidation; S02 = sulfur dioxide.
Note: Pathways indicated by a dotted line are those for which evidence is limited to findings from experimental animal studies, while
evidence from controlled human exposure studies is available for pathways indicated by a solid line. No links to outcomes are
proposed. Key events are subclinical effects and endpoints are effects that are generally measured in the clinic.
Source: National Center for Environmental Assessment.
Figure 4-4 Summary of evidence for the proposed mode of action linking
exposure to sulfur dioxide and extrapulmonary effects.
Besides inhalation of SO2, the ingestion of food additives and the catabolism of
sulfur-containing amino acids also contribute to levels of sulfite in the body
(Section 4.3.5). In humans, the amount of sulfite derived from inhaled SO2 (assuming
100% absorption. 75 ppb and 24-hour exposure) is comparable to that derived from the
expected daily consumption of food additives. The amount of sulfite derived from the
breakdown of endogenous sulfur-containing amino acids is far greater. Sulfite derived
from inhaled SO2, unlike that derived from food additives, enters the circulation without
first passing through the liver, which efficiently metabolizes sulfite to sulfate. Thus, the
potential exists for inhaled SO2 to have a greater impact on circulating sulfite levels than
sulfite derived from food additives. While the amount of sulfite derived from the
breakdown of endogenous sulfur-containing amino acids is far greater, its metabolic
pathways and impact on circulating sulfite levels are not clear. Thus, the potential exists
for prolonged exposure to high concentrations of inhaled ambient SO2 to result in
extrapulmonary effects due to circulating sulfite.
In summary, this section provides a foundation for understanding how exposure to the
gaseous air pollutant SO2 may lead to health effects. This encompasses the many steps
between uptake into the respiratory tract and biological responses that ensue. The
4-36
-------
reaction of inhaled SO2 with components of the ELF initiates a cascade of events
occurring at the cellular, organ, and organism level. Biological responses discussed in
this section were organized in a mode of action framework that serves as a guide to
interpreting health effects evidence presented in Chapter 5.
4-37
-------
CHAPTER 5 INTEGRATED HEALTH EFFECTS
OF EXPOSURE TO
SULFUR OXIDES
5.1 Introduction
5.1.1 Scope of the Chapter
While the term "sulfur oxides" (SOx) refers to multiple gaseous oxidized sulfur
compounds [e.g., sulfur dioxide (SO2), sulfur trioxide], this chapter focuses on evaluating
the health effects associated with exposure to SO2. As discussed in Section 2.1. SO2 is the
most abundant SOx species in the atmosphere, and the available health evidence
examines SO2. The health effects of particulate sulfur-containing compounds
(e.g., sulfate) are considered in the current review of the National Ambient Air Quality
Standards (NAAQS) for particulate matter (PM) and were evaluated in the 2009
Integrated Science Assessment (ISA) for PM (U.S. EPA. 2009a') (see Section 1.1).
This chapter evaluates the epidemiologic, controlled human exposure, and animal
toxicological evidence of S02-related respiratory (Section 5.2). cardiovascular
(Section 5.3). reproductive and developmental (Section 5.4). total mortality (Section 5.5).
and cancer (Section 5.6) effects. Evidence from epidemiologic and animal toxicological
studies of other SC>2-related effects are included in Supplemental Tables 5S-1 (U.S. EPA.
2017c) and 5S-2 (U.S. EPA. 2017c). Sections for respiratory, cardiovascular, and
mortality effects are divided into subsections describing the evidence for short-term
(i.e., 1 month or less) and long-term (i.e., more than 1 month) exposures. The evidence
for reproductive and developmental and cancer effects is considered within one long-term
exposure section, with time windows of exposure addressed as appropriate. Causal
conclusions are determined for both short- and long-term exposures by evaluating the
evidence for each health effect and exposure category independently, using the causal
framework [described in the Preamble to the ISAs (U.S. EPA. 2015bVI.
Each chapter section begins with a summary of the conclusions from the 2008 ISA for
Sulfur Oxides, followed by an evaluation of recent health studies (i.e., those published
since the completion of the 2008 ISA for Sulfur Oxides) that build upon evidence from
previous reviews. The collective body of evidence, including recent studies and studies
included in previous assessments, is integrated across scientific disciplines to develop
conclusions and causality determinations. Within each of the sections focusing on
morbidity outcomes (e.g., respiratory morbidity, cardiovascular morbidity), the evidence
5-1
-------
is organized into more refined outcome groupings [e.g., asthma exacerbation, myocardial
infarction (MI)] that comprise a continuum of subclinical to clinical effects. The
discussion of specific health outcomes is then organized by scientific discipline
(i.e., epidemiology, controlled human exposure, toxicology). This structure helps in
evaluating coherence and biological plausibility of the effects observed in association
with exposure to SO2 and promotes the transparent characterization of the weight of
evidence in drawing the causal conclusions found at the end of each section (e.g., see
Section 5.2.1.9). Causal determinations for total mortality are based on the evidence for
nonaccidental causes of mortality and informed by the extent to which evidence for the
spectrum of cardiovascular and respiratory effects provides biological plausibility for
SCh-related total mortality. Findings for cause-specific mortality inform multiple causal
determinations. For example, studies of respiratory and cardiovascular mortality are used
to assess the continuum of effects and inform the causal determinations for respiratory
and cardiovascular morbidity. As described in Section 1.2. judgments regarding causality
are made by evaluating the evidence over the full range of exposures in animal
toxicological, controlled human exposure, and epidemiologic studies defined in this ISA
to be relevant to ambient air concentrations (i.e., <2,000 ppb).
5.1.2 Evidence Evaluation and Integration to Form Causal Determinations
5.1.2.1 Evaluation of Individual Studies
As described in the Preamble to the ISAs (U.S. EPA. 2015b) (Section 5.a), causal
determinations were informed by integrating evidence across scientific disciplines
(e.g., exposure, animal toxicology, controlled human exposure studies, epidemiology)
and related outcomes, as well as by judgments on the strength of inference from
individual studies. These judgments were based on evaluating strengths, as well as
various sources of bias and uncertainty related to study design, study population
characterization, exposure assessment, outcome assessment, consideration of
confounding, statistical methodology, and other factors. This evaluation was applied to
controlled human exposure, animal toxicological, and epidemiologic studies included in
this ISA, comprising studies from previous assessments as well as those studies published
since the 2008 ISA for Sulfur Oxides. The major considerations in evaluating individual
studies are described in the Preamble and are consistent with current best practices
employed in other approaches for reporting or evaluating health science data.1
1 For example, National Toxicology Program Office of Health Assessment and Translation approach (Rooncv el al..
20141. Integrated Risk Information System Preamble (U.S. EPA. 2013d). ToxRTool (Klimischetal.. 19971.
5-2
-------
Additionally, these considerations are compatible with published U.S. EPA
(Environmental Protection Agency) guidelines related to cancer, neurotoxicity,
reproductive toxicity, and developmental toxicity (U.S. EPA. 2005a. 1998. 1996a. 1991).
The evaluation factors described in the Preamble were used as a guideline rather than a
checklist or criteria to define the quality of a study. The presence or absence of a
particular feature did not necessarily define a less informative study or preclude a study
from consideration in the ISA. Further, these aspects were not criteria for a particular
determination of causality in the five-level hierarchy. As described in the Preamble to the
ISAs (U.S. EPA. 2015b). causal determinations were based on judgments of the overall
strengths and limitations of the collective body of available studies and the coherence of
evidence across scientific disciplines and related outcomes. Where possible,
considerations such as exposure assessment and confounding (i.e., bias due to a
relationship with the outcome and correlation with exposures to SO2), were framed to be
specific to sulfur oxides. Thus, judgments of the strength of inference from a study can
vary depending on the specific pollutant being assessed.
Evaluation of the extent to which the available scientific information informs the
understanding of uncertainties related to the independent effect of sulfur oxides is of
particular relevance in the review process. Because examination of copollutant
confounding is based largely on copollutant models, the inherent limitations of such
models are considered in drawing inferences about independent associations for SO2. For
example, collinearity potentially affects model performance when highly correlated
pollutants are modeled simultaneously, and inference can be limited if there are
differences in the spatial distributions of SO2 and the copollutant such that model
assumptions of equal measurement error or constant correlations for SO2 and the
copollutant are not satisfied (Section 3.4.3V Correlations of short-term SO2
concentrations with other NAAQS pollutants are generally low to moderate, but may
vary by location (Section 3.4.3V Thus, the interpretation of copollutant model results
reported in epidemiologic studies depends on a variety of factors, which are discussed
throughout the chapter, generally in the context of a specific study and/or health
endpoint.
5.1.2.2 Integration of Scientific Evidence
Causal determinations are made by considering the strength of inference from individual
studies and on integrating multiple lines of evidence. As detailed in the Preamble to the
STROBE guidelines (von Elm et al.. 2007). Animals in Research: Reporting In Vivo Experiments guidelines
(Kilkenny et al.. 20101.
5-3
-------
ISAs (U.S. EPA. 2015b). evidence integration involved evaluating the consistency and
coherence of findings within and across disciplines, as well as within and across related
outcomes. Cross-disciplinary integration often addresses uncertainties within a particular
discipline. Controlled human exposure and animal toxicological studies can provide
direct evidence for health effects related to SO2 exposures. Coherence of experimental
evidence with epidemiologic findings can advance our understanding about whether
epidemiologic associations with health outcomes plausibly reflect an independent effect
of exposure to SO2 in ambient air. For example, the coherence of effects observed in
epidemiologic studies with human clinical studies demonstrating direct effects of SO2 on
lung function (Section 5.2.1.2). is drawn upon to reduce uncertainties in epidemiologic
studies. Thus, evidence across a spectrum of related outcomes and across scientific
disciplines (e.g., epidemiologic and controlled human exposure studies) was integrated
and used to clarify the understanding of uncertainties for a particular outcome or
discipline due to chance, publication bias, selection bias, and confounding by copollutant
exposures or other factors.
The integration of the scientific evidence is facilitated by presenting data from multiple
studies within and across disciplines. To increase comparability of results across
epidemiologic studies, the ISA presents effect estimates for associations with health
outcomes scaled to the same increment of SO2 concentration.1 The increments for
standardization vary by averaging time. For 24-h avg, effect estimates were scaled to a
10-ppb increase for SO2. For 1-h daily maximum (max), effect estimates were scaled to a
40-ppb increase for SO2. Effect estimates for long-term exposures to SO2 (i.e., annual or
multiyear averages) were scaled to a 5-ppb increase. These increments were derived by
calculating the U.S.-wide percentile distributions for a given averaging time and then
calculating the approximate difference between the median (i.e., 50th percentile, atypical
pollution day) and the 98th percentile (a more polluted day) for a given averaging time.
Units of dose in toxicological studies are typically presented in ppm; however, when
toxicological data are summarized in the context of epidemiologic findings, units are
converted to ppb for comparability.
5.1.3 Summary
The subsequent sections review and synthesize the evidence of S02-related health effects
from multiple disciplines (e.g., exposure, animal toxicology, and epidemiology).
Information on dosimetry and modes of action (Chapter 4) provides the foundation for
understanding how exposure to inhaled SO2 may lead to health effects, providing
1 This is as opposed to reporting effect estimates that are scaled to variable changes in concentration such as
interquartile range (1QR) for the study period or an arbitrary unit.
5-4
-------
biological plausibility for effects observed in the health studies. The science related to
sources, emissions, and atmospheric concentrations (Chapter 2). as well as the potential
for human exposure to ambient air sulfur oxides (Chapter 3). also informs the
interpretation of the health effects evidence. Integrative "Summary and Causal
Determination" sections for short- and long-term exposures follow the discussion of the
evidence for each health outcome category. These integrative summary sections include
assessments of the strength of inference from studies comprising the evidence base and
integrate multiple lines of evidence to characterize relationships between sulfur oxides
and various health effects.
5.2 Respiratory Effects
5.2.1 Short-Term Exposure
5.2.1.1 Introduction
The 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d) concluded that there is a causal
relationship between respiratory effects and short-term exposure to SO2. The rationale for
this causal determination was heavily based on evidence from multiple, high-quality
controlled human exposure studies demonstrating decreased lung function and increased
respiratory symptoms following SO2 exposures of 5-10 minutes in exercising adults with
asthma.
Epidemiologic evidence also indicated associations between short-term increases in
ambient SO2 concentration and respiratory effects in populations living in locations with
ambient concentrations below the previous 24-h avg NAAQS level of 140 ppb. Evidence
was strongest for increased respiratory symptoms and respiratory-related hospital
admissions and emergency department (ED) visits, especially in children. Due to
inadequate examination, a key uncertainty was potential confounding by copollutants,
particularly PM. However, controlled human exposure studies of individuals with asthma
clearly show that respiratory effects are caused by 5-10 minute SO2 exposures.
In contrast with asthma exacerbation, there was little information to assess whether
short-term SO2 exposure exacerbated allergy or chronic obstructive pulmonary disease
(COPD) or increased risk of respiratory infection. There was some experimental
evidence, however, for respiratory effects in healthy humans (>1,000 ppb) and animal
models (100 ppb) exposed to SO2. Epidemiologic evidence in healthy populations was
limited and inconsistent.
5-5
-------
As described in the following sections, evidence from recent studies is generally
consistent with that in the 2008 ISA and 1982 Air Quality Criteria Document (AQCD)
for Sulfur Oxides (U.S. EPA. 2008d. 1982a). To clearly characterize differences in the
weight of evidence and the extent of coherence among disciplines and related outcomes,
the sections are organized by respiratory outcome group [asthma exacerbation
(Section 5.2.1.2). allergy exacerbation (Section 5.2.1.3). COPD exacerbation
(Section 5.2.1.4). respiratory infection (Section 5.2.1.5). aggregated respiratory
conditions (Section 5.2.1.6). respiratory effects in the general population and healthy
individuals (Section 5.2.1.7). and respiratory mortality (Section 5.2.1.8)1. Epidemiologic
studies comprise most of the recent evidence base, and previous controlled human
exposure and animal toxicological studies form the basis for characterizing and
integrating evidence across disciplines. Recent epidemiologic evidence supports
associations between ambient SO2 concentrations and asthma-related symptoms, hospital
admissions, and ED visits, but uncertainties related to exposure measurement error and
copollutant confounding remain. Recent epidemiologic studies add information on
allergy and COPD exacerbation, respiratory infection, and respiratory effects in healthy
populations, but relationships of these outcomes with short-term SO2 exposure remain
unclear because of inconsistent evidence or limited coherence among disciplines.
5.2.1.2 Asthma Exacerbation
Asthma is a chronic inflammatory lung disease with a broad range of characteristics and
disease severity. SO2 exposure has been demonstrated to induce clinical features of
asthma exacerbation, including decreased lung function [e.g., decreased forced expiratory
volume in 1 sec (FEVi) or increased specific airway resistance (sRaw)], and increased
symptoms (e.g., wheezing, cough, shortness of breath), as well as some subclinical
effects such as inflammation. This section describes evidence for S02-associated lung
function changes and respiratory symptoms in people with asthma, hospital admissions
and emergency department visits for asthma and related respiratory conditions, and
subclinical effects underlying asthma such as pulmonary inflammation and oxidative
stress.
As detailed in the previous 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d). controlled
human exposure studies reported increased respiratory symptoms and decreased lung
function after short-term exposures of 5-10 minutes to 0.2-0.6 ppm SO2 during exercise
or eucapnic hyperpnea (a rapid and deep breathing technique through a mouthpiece that
prevents an imbalance of carbon dioxide due to hyperventilation) in adults and
adolescents (12-18 years old) with asthma. In contrast, healthy adults demonstrated
increased airway resistance and decreased FEVi following exposure to higher
5-6
-------
concentrations (>1.0-5.0 ppm) in the majority of controlled human exposure studies
evaluating the respiratory effects of SO2 (Section 5.2.1.7). While children may be
especially susceptible to the respiratory effects of SO2 for dosimetric reasons
(Section 4.2.2). there are no available controlled human exposure studies in children
under 12, partly due to ethical concerns.
Coherent with controlled human exposure findings, epidemiologic evidence indicated
that short-term increases in ambient SO2 concentration were associated with
asthma-related hospital admissions, ED visits, and symptoms. The strongest evidence
was for children, which is consistent with their greater oral breathing and higher
ventilation rates relative to their size than adults and the consequent potential for them
receiving a higher SO2 dose to the tracheobronchial airways of the lower respiratory tract
(Sections 4.1.2. 4.2.2). Epidemiologic evidence for S02-related lung function decrements
was inconsistent among both children and adults with asthma. A key uncertainty in the
epidemiologic evidence was whether the findings reflected an independent association for
SO2 because the studies assigned exposure from a limited set of fixed-site monitors that
may not adequately reflect the spatial and temporal heterogenity of SO2 concentrations
(Section 3.3.1.1). Also, few of the studies examined potential confounding by particulate
matter with an aerodynamic diameter less than or equal to 2.5 (.un (PM2 5) or other
copollutants.
The 2008 SOx ISA (U.S. EPA. 2008d) also provided limited evidence for a relationship
between SO2 concentrations and allergic responses and inflammation in individuals with
asthma. Children and adults with atopy plus asthma were found to be at greater risk of
SCh-associated respiratory effects such as respiratory symptoms and lung function
decrements. In addition, animal toxicological studies demonstrated that repeated
exposure to SO2 enhanced inflammation and allergic responses in animal models of
allergic airway disease.
Both recent studies and the evidence presented in the 2008 SOx ISA link short-term SO2
exposure to asthma exacerbation. Most recent studies are epidemiologic, and they
continue to show ambient S02-associated increases in asthma symptoms, hospital
admissions, and ED visits among children. However, uncertainty regarding exposure
measurement error and copollutant confounding remains in the epidemiologic evidence.
A few recent animal toxicological studies add support for SC>2-induced allergic
inflammation. While there are no recent controlled human exposure studies in individuals
with asthma (see Section 5.2.1.7 for recent studies in healthy individuals), previous
evidence from controlled human exposure studies provides support for an independent
effect of SO2 exposure on asthma exacerbation.
5-7
-------
Lung Function Changes in Populations with Asthma
The 2008 SOx ISA (U.S. EPA. 2008cD reported strong evidence for the effects of SO2
exposure on decrements in lung function in controlled human exposure studies in adults
with asthma under increased ventilation conditions. Controlled human exposure studies,
none of which are new since the 2008 SOx ISA (U.S. EPA. 2008d). also demonstrated a
subset of individuals within this population who are particularly sensitive to the effects of
SO2 exposure. This finding is most evident in the recent analysis of several published
studies by Johns et al. (2010). Some additional data from the previous studies has also
become available since the 2008 SOx ISA and is summarized in Tables 5-2. 5-3. and 5-4.
Recent epidemiologic findings are inconsistent overall. A few recent epidemiologic
studies add evidence for SO2 measured at a children's school or in copollutant models
with PM, nitrogen dioxide (NO2), or ozone (O3), although their reliance on fixed site
monitors that may not capture the spatial and temporal variation of SO2 represents a
limitation. There is a paucity of evidence from animal toxicological studies. While some
animal toxicological studies of short-term exposure to SO2 have examined changes in
lung function, these experiments were conducted in naive animals rather than in models
of allergic airway disease, which share many phenotypic features with asthma in humans.
Controlled Human Exposure Studies
Bronchoconstriction in individuals with asthma is the most sensitive indicator of
S02-induced lung function effects. A characteristic feature of individuals with asthma is
an increased propensity of their airways to narrow in response to bronchoconstrictive
stimuli relative to nonatopic individuals without asthma. Different kinds of stimuli can
elicit bronchoconstriction, but in general, stimuli act on airway smooth muscle receptors
(direct stimuli, e.g., methacholine) or act via the release of inflammatory mediators
(indirect stimuli, e.g., allergens) (O'Bvrne et al.. 2009). SO2 is a nonspecific
bronchoconstrictive stimulus that is not easily classified as direct or indirect, as discussed
in Section 4.3.1.
The propensity for airways to narrow following inhalation of some stimuli is termed
bronchial or airway responsiveness. The term airway hyperresponsiveness (AHR) is
generally used in cases where airway responsiveness to methacholine or histamine is
assessed and the provocative concentration (PC) is determined to be sufficiently low to
classify the subjects as having AHR based on criteria such as ATS (2000a'). Along with
symptoms, variable airway obstruction, and airway inflammation, AHR is a primary
feature in the clinical definition and characterization of asthma severity (Reddel et al..
2009).
5-8
-------
Bronchoconstriction, evidenced by decrements in lung function, is observed in controlled
human exposure studies after approximately 5-10-minute exposures and can occur at
SO2 concentrations as low as 0.2 ppm in exercising individuals with asthma; more
consistent decrements are seen at concentrations of 0.4 ppm and greater (U.S. EPA.
2008d). In contrast, healthy adults are relatively insensitive to the respiratory effects of
SO2 below 1 ppm (Section 5.2.1.7). In all individuals, bronchoconstriction is mainly seen
during conditions of increased ventilation rates, such as exercise or eucapnic hyperpnea,
and is likely due to a shift from nasal breathing to oronasal breathing, which increases the
concentration of SO2 reaching the bronchial airways (Section 4.2.2). The majority of
controlled human exposures to SO2 were conducted with adult volunteers, although a
limited number were also conducted with adolescents (12-18 years). Characteristics of
controlled exposure studies in individuals with asthma are summarized in Table 5-1.
Controlled exposure studies in individuals without asthma are discussed in
Section 5.2.1.7.
Table 5-1 Study-specific details from controlled human exposure studies of
individuals with asthma.
Study
Disease Status; n;
Sex; (Agea)
Exposure Details
(Concentration; Duration)
Endpoints Examined
Balmes et al. (1987)
Asthma; n = 8; 6 M, 2 F
(23-39 yr)
0, 0.5, or 1 ppm SO2 for 1, 3, or 5 min
during eucapnic hyperpnea (60 L/min) via
mouthpiece
sRaw, symptoms
Bethel et al. (1983)
Asthma; n = 10; 8 M,
2 F
(22-36 yr)
0 or 0.5 ppm SO2 for 5 min with exercise
750 kg m/min (125 watts)
sRaw
Bethel etal. (1984)
Asthma; n = 7; 5 M, 2 F
(24-36 yr)
0 or 0.5 ppm SO2 for 3 min with humidified
room-temperature or cold dry air via
mouthpiece
sRaw
Bethel et al. (1985)
Asthma; n = 19; 16 M,
3 F (22-46 yr)
0 or 0.25 ppm SO2 for 5 min during heavy
exercise [bicycle, 750 (n = 19) or
1,000 (n = 9) kg m/min; 125 or 167 watts,
respectively]
sRaw
Gona et al. (1995)
Asthma; n = 14; 12 M,
2 F (18-50 yr)
0 or 0.5, 1.0 ppm SO2 with light, medium,
and heavy exercise (avg ventilation 30, 36,
and 43 L/min, respectively) for 10 min
sRaw, FEV1,
symptoms,
psychophysical
(stamina) changes
Gona et al. (1996)
Asthma; n = 10; 2 M,
8 F (19-49 yr)
0 or 0.75 ppm SO2 for 10 min with exercise
(29 L/min) at 1, 12, 18, and 24 h after
pretreatment with placebo or salmeterol
(long-acting p2-agonist)
FEV1, symptoms
5-9
-------
Table 5-1 (Continued): Study specific details from controlled human exposure
studies of individuals with asthma.
Disease Status; n;
Study Sex; (Agea)
Exposure Details
(Concentration; Duration)
Endpoints Examined
Gong et al. (2001) Asthma; n = 12; 2 M,
10 F (20-48 yr)
0 or 0.75 ppm SO2 for 10 min with exercise
(35 L/min) with or without pretreatment to
montelukast sodium (10 mg/day for 3 days)
sRaw, FEV1,
symptoms, eosinophil
counts in induced
sputum
Horstman et al.
(1986)
(1) Asthma; n = 27;
27 M with asthma and
sensitive to inhaled
methacholine
(19-33 yr)
(2) n = 4 from study
population above
(1)0, 0.25, 0.5, or 1.00 ppm SO2 for 10 min sRaw
with exercise (treadmill, 21 L/min per m2
body surface area)
(2) 2 ppm SO2 for 10 min with exercise
(treadmill, 21 L/min per m2 body surface
area)
Horstman et al.
(1988)
Asthma; n = 12; 12 M
(22-37 yr)
0 or 1.0 ppm SO2 for 0, 0.5, 1.0, 2.0, and
5.0 min with exercise (treadmill 40 L/min)
sRaw, symptoms
Jorres and
Maqnussen (1990)
Asthma; n = 14; 10 M,
4 F (21-55 yr,
34 ± 14 yr)
0 or 0.25 ppm NO2, or 0.5 ppm SO2 at rest
for 30 min via mouthpiece followed by
challenge with 0.75 ppm SO2 during
voluntary eucapnic hyperpnea via
mouthpiece. Ventilation increased in
15 L/min steps, each lasting 3 min
sRaw
Kehrl et al. (1987)
Asthma; n = 10; 10 M
(20-30 yr)
0 or 1 ppm SO2 for 1 h with exercise
(3x10 min at 41 L/min on a treadmill)
sRaw
Koenia et al. (1980)
Asthma; n = 9:
(14-18 yr)
7 M, 2 F 0 or 1 ppm SO2 with 1 mg/m3 of NaCI
droplet aerosol and filtered air (no SO2 or
NaCI) exposures for 60 min via facemask
with mouth breathing at rest, no exposure
to SO2 alone
FEV1, RT, FRC, Vmaxso,
V max75, symptoms
Koenia et al. (1981) Asthma; n =
(14-18 yr)
8; 6 M, 2 F 0 or 1 ppm SO2 with 1 mg/m3 of NaCI
droplet aerosol and filtered air (no SO2 or
NaCI) exposures for 30 min via mouthpiece
at rest followed by 10 min exercise on a
treadmill (sixfold increase in minute
ventilation), no exposure to SO2 alone
FEV1, RT, FRC, Vmaxso,
V max75, symptoms
Koenia et al. (1983)
(1) Asthma with EIB;
n = 9; 6 M, 3 F
(12-16 yr)
(2) Asthma with EIB;
n = 7 from study
population above
(1) 1 g/m3 of NaCI droplet aerosol, 1 ppm
SO2 + 1 mg/m3 NaCI, 0.5 ppm
SO2 + 1 mg/m3 NaCI for 30 min exposure
via mouthpiece at rest followed by 10 min
exercise on treadmill (five- to sixfold
increase in Ve), no exposure to SO2 alone
or filtered air
(2) 0.5 ppm SO2 + 1 mg/m3 NaCI via a face
mask without nose clip with exercise
conditions the same as above, no exposure
to SO2 alone or filtered air
FEV1, RT, FRC, Vmaxso,
V max75, symptoms
Koenia et al. (1987)
Allergic with EIB;
n = 10; 3 M 7 F
(13-17 yr)
0 or 0.75 ppm SO2 (mouthpiece) with FEV1, RT, FRC,
exercise (33.7 L/min) for 10 and 20 min symptoms
prior pretreatment (placebo or 180 |jg
albuterol)
5-10
-------
Table 5-1 (Continued): Study specific details from controlled human exposure
studies of individuals with asthma.
Disease Status; n;
Study Sex; (Agea)
Exposure Details
(Concentration; Duration)
Endpoints Examined
Koenia et al. (1988)
Asthma with EIB; n :
2 M, 6 F (13-17 yr)
1.0 ppm SO210 min (mouthpiece, treadmill,
35 L/min) with pretreatment (placebo 20,
40, 60 mg cromolyn) 20 min prior, no air
control exposure
FEVi.RT
Koeniq et al. (1990)
Asthma with EIB;
n = 13; 8 M, 5 F
(12-18 yr)
0.1 ppm SO2 for 15 min preceded by air or
0.12 ppm O3 for 45 min during intermittent
exercise (2><15 min at 30 L/min on a
treadmill), all exposures were via
mouthpiece, no air control exposure
FEV1, RT, FRC, Vmaxso,
symptoms
Koenia et al. (1992)
Asthma; n = 8; 2 M, 6 F 0 or 1 ppm SO2 for 10 min via mouthpiece
(18-46 yr; with exercise (13.4-31.3 L/min) with or
27.5 ± 9.6 yr) without pretreatment to theophylline
FEV1, RT
Lazarus et al.
(1997)
Asthma; n = 12; 7 M,
5 F (24-43 yr)
0, 0.25, 0.5, 1.0, 2.0, 4.0, or 8.0 ppm SO2
with eucapnic hyperpnea (20 L/min) via
mouthpiece for 4 min sequential exposures
with pretreatment with Zafirlukast (placebo
or 20 mg) 2 or 10 h earlier
sRaw
Linn et al. (1983b)
Asthma; n = 23; 13 M,
10 F (19-31 yr)
(1) 0, 0.2, 0.4, or 0.6 ppm SO2 with low
humidity or high humidity for 10 min with
exercise (bicycle, 5 min 50 L/min)
(2) 0 or 0.6 ppm SO2 with warm air or cold
air with exercise (bicycle, 50 L/min, ~5 min)
sRaw, sGaw, FVC,
FEV1, symptoms
Linn et al. (1983a)
Asthma; n = 23; 15 M,
8 F (18-30 yr,
23 ± 4 yr)
0 or 0.75 ppm SO2 with unencumbered
breathing and mouth only breathing (with
exercise 40 L/m, 10 min bicycle)
sRaw, thoracic gas
volume, symptoms,
FVC, FEV1, PEFR,
Vmax50, Vmax25
Linn et al. (1984c)
Asthma; n = 24; 13 M,
11 F (19-31 yr)
0, 0.3, or 0.6 ppm SO2 at 21°C, 7°C, and
-6°C, RH 80% (bicycle 50 L/min, ~5 min)
sRaw, sGaw,
symptoms
Linn et al. (1984a)
Asthma: n = 14; 12 M,
2 F (18-33 yr)
0 or 0.6 ppm SO2 for 6 h with exercise on
Day 1 and 2(2* 5-min exercise, bicycle,
50 L/min per exposure)
sRaw, sGaw,
symptoms
Linn et al. (1984b)
(1) Asthma; n = 8; 4 M,
4 F (19-29 yr)
(2) Asthma; n = 24;
17 M 7 F (18-30 yr)
(1) 0, 0.2, 0.4, or 0.6 ppm SO2 at 5°C, 50
and 85%RH with exercise (5 min, 50 L/min)
(2) 0 or 0.6 ppm SO2 at 5°C and 22°C, 85%
RH with exercise (5 min, 50 L/min)
sRaw, sGaw, FEV1,
symptoms
Linn et al. (1985b)
Asthma; n = 22; 13 M,
9 F (18-33 yr)
0 or 0.6 ppm SO2 at 21 °C and 38°C and 20
and 80% RH with exercise (~5 min,
50 L/min)
sRaw, sGaw,
symptoms
Linn et al. (1985a)
COPD; n =24; 15 M,
9 F (49-68 yr)
0, 0.4, or 0.8 ppm SO2 for 1 h with exercise
(2x15 min, bicycle, 18 L/min)
sRaw, FVC, FEV1,
MMFR, symptoms
5-11
-------
Table 5-1 (Continued): Study specific details from controlled human exposure
studies of individuals with asthma.
Disease Status; n;
Study Sex; (Agea)
Exposure Details
(Concentration; Duration)
Endpoints Examined
Linnetal. (1987)
Healthy; n = 24; 15 M,
9 F (18-37 yr)
Atopic; n = 21; 12 M,
9 F (18-32 yr)
Minimal or mild
asthma; n = 16; 10 M,
6 F (20-33 yr)
Moderate or severe
asthma; n = 24; 10 M,
14 F (18-35 yr)
Moderate or severe
asthma; n = 24
0, 0.2, 0.4, or 0.6 ppm SO2
1 h exposures
3 x 10-min exercise (bicycle) periods
-40 L/min
Two rounds of exposures were conducted
Lung function measure
pre-exposure, -15 min
and -55 min into
exposure
sRaw, FVC, FEV1,
peak expiratory flow
rate, maximal
midexpiratory flow rate
Continuously—EKG
Midway—HR
Before, during, 1-day
after, and 1-wk
after-symptom score,
self-rated activity
Immediately after
exposure—bronchial
reactivity percentage
change in FEV1
induced by 3 min
normocapnic
hyperpnea with cold,
dry air
Linn et al. (1988) Asthma: n = 20; 13 M.
Three pretreatment groups
Lung function—pre,
7 F (19-36 yr)
(1) metaproterenol sulfate (2) placebo
post 60 min, 90 min
(3) no treatment
120 min,
0, 0.3, or 0.6 ppm SO2
Symptoms—pre, post,
20 min post, 60 min
post, 120 min post,
24 h post, 1 wk post
10 min with exercise (bike 50 L/min)
Linnetal. (1990)
Asthma; n = 21; 6 M,
15 F (19-48 yr)
0, 0.3, or 0.6 ppm SO210 min with exercise
50 L/min
(1) low medication use; (2) normal; (3) high
(usual medication supplemented by inhaled
metaproterenol before exposure)
Lung function and
symptoms measured
before and after
exposure
Maqnussen et al.
(1990)
Asthma; n = 46; 24 M,
22 F(28± 14 yr)
Healthy; n = 12
(24 ± 5 yr)
0 or 0.5 ppm SO210 min tidal breathing
followed by 10 min of isocapnic
hyperventilation (30 L/min) via mouthpiece
Histamine challenge—(8 mg/mL)
sRaw
Myers et al. (1986a)
Asthma; n = 10; 7 M,
3 F (19-40 yr)
0, 0.25, 0.5, 1, 2, 4, or 8 ppm SO2 3 min
sequential exposures (mouthpiece,
40 L/min) with pretreatment 30 min prior
with cromolyn (placebo, 20, or 200 mg)
sRaw
Myers et al. (1986b)
(1) Asthma; n = 9;
2 F (19-40 yr)
(2) Asthma; n = 7;
(19-40 yr)
7 M, 0, 0.25, 0.5, 1, 2, 4, or 8 ppm SO2 3 min
sequential exposures (mouthpiece,
7 m eucapnic hyperpnea 40 L/min) with
pretreatment 30 min prior (1) atropine
(2 mg) and cromolyn (200 mg); (2) placebo
and cromolyn (200 mg); (3) atropine (2 mg)
and placebo; (4) placebo
sRaw
5-12
-------
Table 5-1 (Continued): Study specific details from controlled human exposure
studies of individuals with asthma.
Disease Status; n;
Study Sex; (Agea)
Exposure Details
(Concentration; Duration)
Endpoints Examined
Roger et al. (1985)
Asthma; n = 28; 28 M
(19-33 yr)
0, 0.25, 0.5, or 1.0 ppm SO2 with three
10 min periods of exercise (42.4 L/min)
each followed by 15 min of exposure at rest
for a total exposure duration of 75 min
Raw; sRaw; FVC,
FEV1, FEF25-75,
FEFmax, FEF50, FEF75,
sRaw, FVC, FEV1,
single-breath nitrogen
test
Rubinstein et al. Asthma; n = 9; 5 M, 4 F 0 or 0.3 ppm NO2 during exercise in a
(1990) (23-34 yr) chamber followed by challenge with 0.25 to
4.0 ppm SO2, in doubling dose increments,
for 4 min each via mouthpiece during light
exercise (20 L/min) until sRaw increased by
8 SRaw units above baseline
Sheppard et al.
(1983)
Asthma; n = 8; 4 M, 4 F
(22-36 yr)
0.5 ppm SO2 for 3 min eucapnic hyperpnea
via mouthpiece
sRaw, symptoms
Trenqa et al. (1999)
Asthma; n = 47; 14 M,
33 F (18-39 yr)
0.5 ppm SO2 for 10 min via mouthpiece
during moderate exercise
FEV1, FVC, FEV1/FVC,
PEF, FEF25-75,
symptoms ratings
Trenqa et al. (2001)
Asthma; n = 17; 5 M,
12 F (19-38 yr)
0.1 or 0.25 ppm SO2 for 10 min via
mouthpiece with moderate exercise
(treadmill) following exposure to air or 0.12
O3 ppm for 45 min via mouthpiece with
intermittent moderate exercise.
FVC, FEV1, FEF25-75,
PEF, symptoms
Tunnicliffe et al.
(2003)
Asthma; n = 12
(adults, 35.7 yr)
Healthy; n = 12
(adults, 34.5 yr)
0 or 0.2 ppm SO2 via head dome at rest
Symptoms, FEV1, FVC,
MMEF, exhaled NO,
ascorbic and uric acid
in nasal lavage fluid
COPD = chronic obstructive pulmonary disease; EIB = exercise-induced bronchospasm; EKG = electrocardiogram; F = female;
FE\A| = forced expiratory volume in 1 sec; FEF25-75% = forced expiratory flow at 25-75% of forced vital capacity; FEF50% = forced
expiratory flow at 50% of forced vital capacity; FEF75o/o = forced expiratory flow at 75% of forced vital capacity; FEFmax = maximum
forced expiratory flow; FRC = functional residual capacity; FVC = forced vital capacity; HR = heart rate; M = male;
MMEF = maximum midexpiratory flow; MMFR = maximal midexpiratory flow rate; n = sample size; NaCI = sodium chloride;
NO = nitric oxide; N02 = nitrogen dioxide; 03 = ozone; PEF = peak expiratory flow; PEFR = peak expiratory flow rates; ppm = parts
per million; Raw = airway resistance; RH = relative humidity; RT = total respiratory resistance; SD = standard deviation;
sGAW = specific airway conductance; sRaw = specific airway resistance; S02 = sulfur dioxide; VE = minute volume;
Vmax = maximal flow of expired vital capacity; Vmax75 = flow rate with 75% of FVC remaining to be expired; Vmax5o = flow rate with
50% of FVC remaining to be expired; Vmax25 = flow rate with 25% of FVC remaining to be expired.
aRange or Mean ± SD.
Several investigators (Linn et al.. 1990; Linn et al.. 1988; Linn et al.. 1987; Bethel et al..
1985; l.inn et al.. 1984a; Linn et al.. 1983b) demonstrated >100% increase in sRaw or
>15% decrease in FEVi after 5-10-minute exposures to low concentrations
(0.2-0.3 ppm) of SO2 in exercising adults with asthma, with effects being more
pronounced following 5-10-minute exposures to 0.4-0.6 ppm SO2 (Linnetal.. 1990;
Magnussen et al.. 1990; l.inn et al.. 1988; l inn et al.. 1987; Roger et al.. 1985; l.inn et al .
1983b).
5-13
-------
SCh-induced bronchoconstriction in individuals with asthma occurs rapidly when
exposed while at increased ventilation and is transient with recovery following cessation
of such conditions. Bronchoconstriction occurs in as little as 2 minutes from the start of
exposure in adults with asthma who have increased ventilation rates due to exercise or
eucapnic hyperpnea (Horstman ct al.. 1988; Balmes et al.. 1987; Sheppard et al.. 1983).
During exposure to SO2 over a 30-minute period with continuous exercise, the response
to SO2 develops rapidly and is maintained throughout the 30-minute exposure (Kchrl et
al.. 1987). Linn et al. (1984a) reported decrements in lung function in adults with asthma
immediately after each of two 5-min exercise periods (one after entering the chamber and
the second 5 hours later) in two 6-hour exposures to 0.6 ppm SO2 on successive days.
The decrements in lung function observed in the early and late exercise periods were not
statistically significantly different from each other, but tended to be diminished in the late
exercise period relative to the first. The responses observed after the second day of SO2
exposure were also slightly (minimal biologically, but statistically less based on sGaw
data) less than the response observed after the first day of SO2 exposure. These effects are
generally observed to return to baseline levels within 1 hour after cessation of exercise,
even with continued exposure (Linnetal.. 1984a).
Other factors that affect responses to SO2 include temperature and humidity. The
majority of controlled human exposure studies were conducted at 20-25°C and at relative
humidities ranging from -25-90%. Some evidence indicates that the respiratory effects
of SO2 are exacerbated by colder and dryer conditions (Linn et al.. 1985b; Bethel et al..
1984; Linnetal.. 1984b).
Responders versus nonresponders to SO2. At the time of the 2008 SOx ISA (U.S. EPA.
2008d). it was well documented that some individuals have a greater response to SO2
than others with similar disease status (Table 5-2) (Linn et al.. 1990; Magnussen et al..
1990; Linn et al.. 1988; Linn et al.. 1987; Horstman et al.. 1986; Bethel et al.. 1985;
Roger et al.. 1985; Linn et al.. 1984b; Linn et al.. 1983b).
Horstman et al. (1986) reported that individuals required different concentrations of SO2
to produce a doubling of sRaw (>100%) compared to clean air exposure [provocative
concentration of SO2, PC(SC>2)] (Figure 5-1). This study described the distribution of
individual bronchial sensitivity to SO2, measured by sRaw, in 27 subjects with asthma
that were sensitive to methacholine; insensitive volunteers w ere excluded from further
participation in the study. Individuals were exposed to concentrations of SO2 between 0
and 2 ppm for 10 minutes under exercising conditions [minute ventilation
(Ve) = 42 L/minute]. While six of the subjects (22%) reached a PC(SC>2) below 0.5 ppm
SO2, two subjects (7.4%) experienced a moderate decrease <0.3 ppm (Figure 5-1). On the
other end of the spectrum, four subjects (14.8%) did not demonstrate >100% increase in
5-14
-------
sRaw even when exposed to 2.0 ppm SO2 and eight (29.6%) subjects required an SO2
concentration between 1.0 and 2.0 ppm to elicit a response. The authors noted that the
effects of SO2 on sRaw are similar to a variety of nonspecific bronchoconstrictive stimuli.
However, they observed only a weak correlation between airway responsiveness to SO2
and methacholine (r = 0.31,/? = 0.12). This study demonstrates substantial interindividual
variability in sensitivity to the bronchoconstrictive effects of SO2 in exercising adults
with asthma.
5-15
-------
Table 5-2 Percentage of adults with asthma in controlled human exposure studies experiencing sulfur
dioxide-induced decrements in lung function and respiratory symptoms.
SO2 Exposure
Cone Duration
(ppm) (min)
Cumulative Percentage of Subjects
(Number of Subjects)3
Ventil-ation
(L/min)
T>sRaw
sU FEV1
>100%
>200%
>15%
>20%
>300%
>30%
Study
Respiratory
Symptoms:
Supporting
Studies
0.2
10
10
23
40
40
-48
sRaw
9% (2)b
Linn et al. (1983b)
-40
sRaw
7.5% (3)c
2.5% (1)c
Linn et al. (19871°
-40
FEVi
9% (3.5)c
2.5% (1)c
1% (0.5)c Linn et al. (19871°
Limited
¦ evidence of
S02-induced
¦ increases in
respiratory
¦ symptoms in
some
people with
asthma:
. (Linn et al.
(1990); Linn
. etal. (1988);
Linn et al.
.(1987);
Schachter et
. al. (1984):
0.25
10
19
28
-50-60
sRaw
32% (6)
16% (3)
-80-90
sRaw
22% (2)
Bethel etal. (1985)
Bethel etal. (1985)
-40
sRaw
4% (1)
Roger et al. (1985)
0.3
10
10
20
21
-50
sRaw
10% (2)
5% (1)
5% (1)
Linn et al. (1988)d
-50
sRaw
33% (7)
10% (2)
Linn et al. (1990)d
10
20
-50
FEVi
15% (3)
Linn et al. (1988)
Linn et al.
(1983b))
10
21
-50
FEVi
24% (5)
14% (3)
10% (2) Linn et al. (1990)
5-16
-------
Table 5-2 (Continued): Percentage of adults with asthma in controlled human exposure studies experiencing
sulfur dioxide induced decrements in lung function and respiratory symptoms.
SO2 Exposure
Cone Duration
(ppm) (min)
Cumulative Percentage of Subjects
(Number of Subjects)3
Ventil-ation
(L/min)
T>sRaw
sU FEV1
>100%
>200%
>15%
>20%
>300%
>30%
Study
Respiratory
Symptoms:
Supporting
Studies
0.4
10
10
23
40
40
-48
sRaw
13% (3)
4% (1)
Linn et al. (1983b)
-40
sRaw
24% (9.5)c
9% (3.5)c
4% (1,5)c Linn et al. (1987)c
-40
FEVi
27.5% (11)
17.5% (7)c
10% (4)c Linn et al. (1987)c
Stronger
¦ evidence
with some
¦ statistically
significant
¦ increases in
respiratory
¦ symptoms:
Balmes et
¦ al. (1987)f.
Gong et al.
(1995) (Linn
et al. (1987):
Linn et al.
(1983b))
Roger et al.
(1985)
0.5
10
10
10
28
45
-50-60
sRaw
60% (6)
40% (4)
20% (2) Bethel et al. (1983)
-40
sRaw
18% (5)
4% (1)
4% (1)
Roger et al. (1985)
-30
sRaw
36% (16)
16% (7)
13% (6) Magnussen et al. (1990)f
LO
CD
O
23
J
CO
sRaw
39% (9)
26% (6)
17% (4)
Linn et al.
(1983b)
10
40
-40
sRaw
34% (13.5)c
24% (9.5)c
19% (7.5)c
Linn et al.
(1987)°
10
20
J
cn
o
sRaw
60% (12)
35% (7)
10% (2)
Linn et al.
(1988)
10
21
cn
o
sRaw
62% (13)
29% (6)
14% (3)
Linn et al.
(1990)
10
40
-40
FEVi
47.5% (19)c
39% (15.5)c
17.5% (7)c
Linn et al.
(1987)°
10
20
J
cn
o
FEVi
55% (11)
55% (11)
5% (1)
Linn et al.
(1988)
10
21
cn
o
FEVi
43% (9)
38% (8)
14% (3)
Linn et al.
(1990)
5-17
-------
Table 5-2 (Continued): Percentage of adults with asthma in controlled human exposure studies experiencing
sulfur dioxide induced decrements in lung function and respiratory symptoms.
Cumulative Percentage of Subjects
(Number of Subjects)3
SO2
Cone
(ppm)
Exposure
Duration
(min)
Ventil-ation
(L/min)
T>sRaw
>100%
>200%
>300%
sU FEV1
>15%
>20%
>30%
Study
Respiratory
Symptoms:
Supporting
Studies
Clear and
consistent
increases in
S02-induced
respiratory
symptoms:
(Linn et al.
(1990): Linn
et al. (1988):
Linn et al.
(1987): Linn
et al.
(1983b)).
Gong et al.
(1995).
Horstman et
al. (1988)
1.0
10
28
-40
sRaw
50% (14)
25% (7)
14% (4)
Roger et al. (1985)e
10
10
-40
sRaw
60% (6)
20% (2)
0
Kehrl et al. (1987)
Cone = concentration; FE\A| = forced expiratory volume in 1 sec; n = sample size; sRaw = specific airway resistance; S02 = sulfur dioxide.
aData presented from all references from which individual data were available. Percentage of individuals who experienced greater than or equal to a 100, 200, or 300% increase in
specific airway resistance, or a 15, 20, or 30% decrease in FEV-i. Lung function decrements are adjusted for the effects of exercise in clean air (calculated as the difference between
the percent change relative to baseline with exercise/S02 and the percent change relative to baseline with exercise/clean air).
bNumbers in parenthesis represent the number of subjects experiencing the indicated effect.
°Responses of people with mild and moderate asthma reported in Linn et al. (1987) have been combined. Data are the average of the first- and second-round exposure responses
following the first 10 min period of exercise. In some cases, the average had a first decimal place value of 5, which is reported in the table to avoid a high bias in values due to
rounding. In all other cases, the calculated percentages were rounded to the nearest integer.
dAnalysis includes data from only people with mild (Linn et al.. 1988) and moderate (Linn et al.. 1990) asthma who were not receiving supplemental medication.
eOne subject was not exposed to 1 ppm due to excessive wheezing and chest tightness experienced at 0.5 ppm. For this subject, the values used for 0.5 ppm were also used for
1.0 ppm under the assumption that the response at 1.0 ppm would be equal to or greater than the response at 0.5 ppm.
'Indicates studies in which exposures were conducted using a mouthpiece rather than a chamber.
5-18
-------
100 ¦
>.
o
c
CD
3
CT
tu
0)
_>
re
E
3
o
75-
50.
25
X
X
X
X*
f
~7fHL
0.25
X
X
0.5
-1 1 r~
0.75 1.0 2.0
PC{S02) (ppm)
5.0
10.0
PC = provocative concentration; S02 = sulfur dioxide.
Note: Each data point represents the PC(S02) for an individual subject.
Source: Horstman et al. (1986V Reprinted with permission of Sage Publications.
Figure 5-1 Distribution of individual airway sensitivity to sulfur dioxide. The
cumulative percentage of subjects is plotted as a function of
provocative concentration, which is the concentration of sulfur
dioxide that provoked a 100% increase in specific airway
resistance compared to clean air.
Completed after the 2008 SOx ISA (U.S. EPA. 2008d). an analysis by Johns et al. (2010)
of publicly available primary data from published studies clearly demonstrates disparate
responses among 177 adults with asthma. Data from five studies of individuals with
asthma exposed to multiple concentrations of SO2 for 5-10 minutes with elevated
ventilation rates (Linnetal.. 1990; Linn et al.. 1988; Linn et al.. 1987; Roger et al.. 1985;
Linn et al.. 1983b) were analyzed after classifying individuals by responder status.
Classification of responders versus nonresponders was based on the magnitude of sRaw
and FEVi changes in response to the highest SO2 concentration to which subjects were
exposed (0.6 or 1.0 ppm). Responders were defined as subjects experiencing >100%
increase in sRaw or >15% decrease in FEVi after exposure. Response status was assigned
5-19
-------
separately for sRaw and FEVi. Among responders, significant decreases in FEVi were
observed for concentrations as low as 0.3 ppm SO2 (p = 0.005) (Table 5-3). In addition,
marginally significant increases in sRaw were demonstrated at 0.3 ppm SO2 (p = 0.009),
with statistically significant increases observed at 0.4 and 0.5 ppm (p < 0.001)
(Table 5-4). [Due to multiple comparisons, Johns et al. (2010) designated a critical
/?-value of 0.005 as significant, using the Bonferroni multiple comparison correction.]
Overall, these data demonstrate a bimodal distribution of airway responsiveness to SO2 in
individuals with asthma, with one subpopulation that is insensitive to the
bronchoconstrictive effects of SO2 even at concentrations as high as 1.0 ppm, and another
subpopulation that has an increased risk for bronchoconstriction at low concentrations of
SO2. The Winterton et al. (2001) study suggests that a tumor necrosis factor alpha
(TNF-a) promoter polymorphism in some individuals with asthma may be associated
with increased airway responsiveness to SO2.
Table 5-3 Percent change in post- versus pre-exposure measures of forced
expiratory volume in 1 second relative to clean air control after
5-10-minute exposures to sulfur dioxide during exercise.
FEVi
qq2 95% Confidence Limits
Concentration Number of
ppm
Exposures
% Decrease
Lower
Upper
p-Value
Responders
0.2
37
-5.0
-8.9
-1.1
0.012
0.3
20
-7.6
-13.0
-2.3
0.005ab
0.4
37
-17.4
-21.3
-13.6
<0.001ab
Nonresponders
0.2
43
0.4
-4.3
5.2
0.854
0.3
21
-3.6
-9.6
2.5
0.252
0.4
43
-4.3
-9.2
0.6
0.086
FE\A| = forced expiratory volume in 1 sec; ppm = parts per million; S02 = sulfur dioxide,
indicates significance at 0.05 level using the Bonferroni multiple comparison correction,
indicates significance at 0.05 level using Dunnett's test.
A generalized linear latent and mixed models (GLLAMM) procedure was used that included study as a fixed effect, concentration
dummy variables as a covariate, and subject and the times a subject was exposed to a sequence of exposures as random
variables. Data were included from Linn et al. (1987). Linn et al. (1988). and Linn et al. (1990).
Source: Table 2 of Johns et al. (2010).
5-20
-------
Table 5-4 Percent change in post- versus pre-exposure measures of specific
airway resistance relative to clean air control after 5-10-minute
exposures to sulfur dioxide during exercise.
sRaw
SO2
Concentration
ppm
Number of
Exposures
95% Confidence Limits
% Increase
Lower
Upper
p-Value
Responders
0.2
36
10.2
-3.6
24.0
0.147
0.25
14
19.5
-4.0
43.1
0.104
0.3
25
25.4
6.5
44.3
0.009
0.4
36
75.7
53.4
98.0
<0.001ab
0.5
14
68.0
33.2
102.8
<0.001ab
Nonresponders
0.2
67
7.9
-4.9
20.7
0.227
0.25
14
12.6
-10.5
35.7
0.286
0.3
16
16.4
-5.2
38.1
0.137
0.4
67
16.2
1.8
30.6
0.028
0.5
14
14.7
-12.3
41.7
0.285
sRaw = specific airway resistance; ppm = parts per million
S02 = sulfur dioxide.
indicates significance at 0.05 p level, using the Bonferroni multiple comparison correction,
indicates significance at 0.05 level using Dunnett's test.
A generalized linear latent and mixed models (GLLAMM) procedure was used that included study as a fixed effect, concentration
dummy variables as a covariate, and subject and the times a subject was exposed to a sequence of exposures as random
variables. Data were included from Linn et al. (1983b). Linn et al. (1987). Linn et al. (1988). Linn et al. (1990). and Roger et al.
(1985).
Source: Table 1 of Johns et al. (2010).
A recent analysis of four previously published studies (Horstman et al.. 1988; Horstman
et al.. 1986; Schachter et al.. 1984; Sheppard et al.. 1980) in which individuals with
asthma were exposed to multiple SO2 concentrations or had their response recorded over
multiple durations of SO2 exposure was provided by Goodman et al. (2015). Eight of
56 individuals were identified as sensitive to the effects of SO2 by Goodman et al. (2015).
However, the analysis conducted by Goodman et al. (2015) did not consider the
log-normal distribution of airway responsiveness data and instead used an arithmetic
mean and standard deviation in their analysis.
5-21
-------
Effects of asthma severity on SCh-induced response. The influence of asthma severity
on the degree of responsiveness to SO2 exposure has been examined (Trenga et al.. 1999;
Linn et al.. 1987). One study involved exposure to SO2 under conditions of increased
ventilation (i.e., exercise) (Linn et al.. 1987). Adults with asthma were divided into two
groups, minimal/mild and moderate/severe, mainly based on the individual's use of
medication to control asthma. Individuals that did not regularly use asthma medication
were classified as minimal/mild; however, even the moderate/severe group consisted of
adults who had well-controlled asthma, were generally able to withhold medication, were
not dependent on corticosteroids, and were able to engage in moderate to heavy levels of
exercise. Thus, this moderate/severe group would likely be classified as moderate by
today's classification standards (Johns et al.. 2010; Reddel. 2009). Linn et al. (1987)
found similar relative decrements in lung function in response to SO2 exposure between
the groups. Nevertheless, the moderate/severe group demonstrated larger absolute
changes in lung function compared with the mild group (Linnetal.. 1987). This greater
decrement in lung function was attributable to a larger response to the exercise
component of the exposure protocol in the moderate/severe group compared with the
mild group. Trenga et al. (1999) found a correlation between asthma severity and
response to SO2. Adults with asthma were divided into four groups based on medication
usage as an indicator of asthma severity. The role of exercise was not determined in this
study, so it was unclear whether individuals with more severe asthma had a greater
response to exercise compared to individuals with less severe asthma. However, both
studies suggest that adults with moderate/severe asthma may have more limited reserve to
deal with an insult compared with individuals with mild asthma.
Asthma with medication. Asthma medications have been shown to mitigate
SCh-induced bronchoconstriction (U.S. EPA. 2008d). Medications evaluated include
short-acting and long-acting beta-adrenergic bronchodilators (Gongetal.. 1996; Linn et
al.. 1990; Linn et al.. 1988; Koenig et al.. 1987). cromolyn sodium (Koenig et al.. 1988;
Myers etal.. 1986b). theophylline (Koenig et al.. 1992). and leukotriene receptor
antagonists (Gong et al.. 2001; Lazarus et al.. 1997). While these therapies have been
shown to mitigate the respiratory effects of SO2, they did not completely eliminate these
effects in all studies.
Children and adolescents. Several studies have examined the responsiveness to SO2 of
adolescents (ages 12-18 years) with asthma or allergic with exercise-induced
bronchospasm (EIB) (Koenig et al.. 1990; Koenig et al.. 1988; Koenig et al.. 1987). Of
these studies, only Koenig et al. (1987) included a control air exposure, so that the
bronchoconstrictive effects of SO2 itself (rather than, for example, due to EIB), can be
assessed. On average, based on the data provided in Table 1 of this paper, adolescents
experienced a pre- to post-exposure reduction in FEVi of 15.4% following exposure to
5-22
-------
0.75 ppm SO2 and a reduction in FEVi of 3.46% following air exposure. Although the
adolescents in this study were allergic with EIB, they did not have extrinsic asthma.
Nevertheless, they are discussed here because allergies affect airway responsiveness
(Burrows et al.. 1995) and because their response to SO2 is similar to that observed in
other studies of individuals with asthma. The pre- to post-reduction in FEVi of 15.4%
following 0.75 ppm SO2 observed by Koenig etal. (1987) is similar to the pre- to post-
reduction in FEVi of 13.9% found in adolescents with asthma following exposure to
1.0 ppm SO2 observed by Koenig et al. (1988). For potential comparison to the results of
adolescents, three studies of adults with asthma were conducted at 0.75 ppm (Gong et al..
2001; Gong et al.. 1996; Linn et al.. 1983a). Of these, only Gong et al. (2001) provided
pre-to-post data for both exposures to air and SO2. Similar to the Koenig et al. (1987)
results, Gong et al. (2001) observed a pre- to post-reduction of 15.8% in FEVi following
SO2 exposure in adults based on Table 2 of their paper. Adjusted for the responses
occurring with air exposure, Koenig et al. (1987) observed an 11.8% reduction in FEVi
in adolescents, similar to the 12.7% reduction observed in adults by Gong et al. (2001).
These two studies differ in that the adolescents were exposed via a mouthpiece, whereas
the adults were exposed in a chamber without a mouthpiece. Breathing on a mouthpiece
is expected to produce a somewhat larger FEVi decrement than unencumbered breathing
(Linn etal.. 1983a). Although generally similar effects of SO2 on adolescents and adults
have been observed, exact comparisons of SO2 effects between adolescents and adults are
not possible given the available data.
There is also evidence that adolescents (ages 12-18 years) with asthma or atopy are
responsive to coexposures of SO2 and sodium chloride (NaCl) droplet aerosol (Koenig et
al.. 1983. 1981; Koenig et al.. 1980). Exposure concentrations in these studies ranged
from 0.1 to 1.0 ppm SO2. Koenig etal. (1983) observed average FEVi decrements of 15
and 23% in exercising adolescents (12-16 years old) with asthma after a 10-minute
exposure to 0.5 ppm SO2 plus 1 mg/m3 NaCl droplet aerosols or 1.0 ppm SO2 plus
1 mg/m3 NaCl droplet aerosols, respectively. No significant changes were observed
following exposure to the NaCl droplet aerosol alone. However, the observed effect may
be the result of the presence of hygroscopic particles that carry SO2 deeper into the lung.
There are no controlled human exposure studies for children less than 12 years of age
who were exposed to SO2. However, the responsiveness of children to SO2 relative to
adolescents and adults may be inferred by the responses to other nonspecific
bronchoconstrictive stimuli. Horstman et al. (1986) noted that the effects of SO2 on sRaw
are similar to that of a variety of nonspecific bronchoconstrictive stimuli. Indeed, SO2 is a
nonspecific bronchial challenge agent that has been used to assess changes in airway
responsiveness of individuals with asthma following NO2 and O3 exposures (Trenga et
al.. 2001; Jorres and Magnussen. 1990; Rubinstein et al.. 1990). Airway responsiveness
5-23
-------
to methacholine, a history of respiratory symptoms, and atopy were significant predictors
of airway responsiveness to SO2 in healthy adults (Nowak et al.. 1997). Thus, potential
differences in airway responsiveness of children to SO2 relative to adolescents and adults
may be gleaned from the literature on airway responsiveness to other nonspecific stimuli
such as methacholine.
A number of cross-sectional studies have assessed airway responsiveness of children with
and without asthma to methacholine [e.g., (Mochizuki et al.. 1995; Morikawa et al.. 1994;
A vital et al.. 1991; Hopp et al.. 1986; Hopp et al.. 1985)1. Studies show a clear decrease
in airway responsiveness of healthy children with increasing age beyond 5-7 years of age
through adolescence (Mochizuki et al.. 1995; Hopp et al.. 1986; Hopp et al.. 1985). In
studies of children with asthma, some have reported airway responsiveness increased
with asthma severity but was not affected by age (Avital et al.. 1991; Hopp et al.. 1986).
whereas others have found airway responsiveness to increase with asthma severity and
decrease with age beyond 6-7 years of age (Mochizuki et al.. 1995; Morikawa et al..
1994). The study by Mochizuki et al. (1995) suggested that airway responsiveness in both
healthy children and those affected by asthma increases from ages 2-3 years up to
6-7 years, after which airway responsiveness begins decreasing.
More confidence in the effect of age on airway responsiveness may be placed on data
from longitudinal studies than from the cross-sectional studies discussed above. In a
longitudinal study of methacholine responsiveness conducted at 9, 11, 13, and 15 years of
age, Le Souef et al. (1995) found that responsiveness (1) decreases with age; (2) is
greater in boys (n = 389) than girls (n = 429); and (3) is greater in those reporting
wheeze, although responsiveness decreased with age in these individuals as well. Asthma
prevalence and symptoms such as wheeze are greater in boys than girls during childhood
and become similar or reversed around the time of puberty (Almqvist et al.. 2008). In a
subset of the cohort as used by Le Souef et al. (1995). Burrows et al. (1995) investigated
the effects of age (n = 573, 49% female), atopy (n = 558), and serum immunoglobulin E
(IgE) (n = 473) on airway responsiveness. At 9 years of age, a larger fraction of boys
experienced AHR compared with girls. By the age of 15 years, there was little to no
difference in the fraction with AHR between the sexes. Relative to atopic children, those
without atopy or with only minimal atopy had a lower fraction with AHR and showed a
more evident decrease in the fraction having AHR with increasing age. In the most atopic
children (41 of 558), about 20-30% experienced severe AHR, which did not decrease
with age. Across all ranges of serum IgE, there was a decrease in the fraction having
AHR from age 9 to age 15 years. By 15 years of age, only a small fraction of the children
with low serum IgE levels had AHR. At both 9 and 15 years of age, the fraction having
AHR increased with increasing serum IgE levels (p < 0.0001). In biennial assessments of
childhood responsiveness, Burrows et al. (1995) observed considerable intra-individual
5-24
-------
variability in bronchial reactivity, but they observed a statistically significant trend for the
more allergic children to experience persistent AHR among their biennial assessments.
A number of factors may influence bronchial responsiveness to SO2 including innate
responsiveness of the airways, route of breathing, disease status, and age. To the extent
that variability in bronchial responsiveness to SO2 may be inferred from studies
evaluating responsiveness to methacholine, these studies suggest that greater airway
responsiveness to SO2 may occur in school-aged children (-5-11 years of age),
particularly boys, than in adolescents. Additionally, the methacholine data also suggest
that greater airway responsiveness to SO2 in school-aged children and adolescents who
are allergic or experience wheeze is expected to occur than in those without these
conditions. Children, particularly boys, breathe more through the mouth than adults, and
ventilation rates relative to body mass are greater in children than adults (see
Section 4.1.2V Allergic rhinitis can lead to increased nasal resistance, which also results
in less nasal and more oral breathing. Obese children also tend to have increased nasal
resistance, increased oral breathing, and increased ventilation rates relative to
normal-weight children (see Section 4.1.2). Oral breathing allows greater SO2 penetration
into the lower airways, where it may cause bronchoconstriction, than does nasal breathing
(see Section 4.2.2). Overall, normal-weight and obese school-aged children (-5-11 years
of age) having asthma-like symptoms might be expected to experience greater
responsiveness (i.e., larger decrements in pulmonary function) following exposure to SO2
than normal-weight adolescents and adults. Boys may be particularly affected due the
combined effects of increased bronchial responsiveness and a greater degree of mouth
breathing.
Mixtures effects. The health effects of SO2 can be potentially modified by the interaction
with other pollutants during or prior to exposure. A few studies involving mixtures with
NaCl droplet aerosol are discussed above. A few controlled human exposure studies have
examined the interactive effects of O3 and S02both sequentially and in combination.
Exercising adolescents with asthma exposed to 0.1 ppm SO2 for 15 minutes after a
45-minute exposure to 0.12 ppm O3 had a significant decrease (8%) in FEVi (8%)
(p < 0.05,), a significant increase in total respiratory resistance (Rt) (19%) (p < 0.05). and
a significant decrease in maximal flow at 50% of expired vital capacity (Vmaxso) (15%)
(p < 0.05); while air followed by SO2, and O3 followed by O3 exposures did not cause
significant changes in lung function (Koenig et al.. 1990). In a more recent study in
exercising adults with asthma, Trenga et al. (2001) observed greater decrements in lung
function after 45 minutes of exposure to 0.12 ppm O3 followed by 15 minutes of
0.25 ppm SO2 compared to air followed by SO2.
5-25
-------
Jorres and Magnussen (1990) and Rubinstein et al. (1990) investigated the effects of prior
NO2 exposure on SC>2-induced bronchoconstriction in adults with asthma. Jorres and
Magnussen (1990) observed that tidal breathing of NO2 at rest increased airway
responsiveness to subsequent hyperventilation of SO2. Rubinstein et al. (1990) noted NO2
exposure during exercise induced greater airway responsiveness to inhaled SO2 in only
one subject of nine. The effect of exercising versus resting exposures to NO2 on airway
responsiveness is discussed elsewhere (Brown. 2015).
While SO2 acts as a nonspecific bronchial challenge agent that causes reductions in lung
function in individuals with asthma after brief exposure, it can also affect airway
responsiveness to subsequent exposures involving other stimuli such as allergens or
methacholine. Two studies of adults with asthma exposed at rest to SO2 in combination
with NO2 demonstrated increases in airway responsiveness to subsequent allergen
challenge (Rusznak et al.. 1996; Devalia et al.. 1994). In the first of these studies,
exposure to 0.2 ppm SO2 or 0.4 ppm NO2 did not affect airway responsiveness to house
dust mite allergen immediately after a 6-hour exposure. In considering the effect of SO2
alone, it is unlikely that enough SO2 reached the bronchial airways to cause an effect
because volunteers were exposed at rest. Following exposure to the two pollutants in
combination, volunteers demonstrated an increased response to inhaled allergen (Devalia
et al.. 1994). Rusznak et al. (1996) confirmed these results in a similar study and found
that increased airway responsiveness to dust mites persisted up to 48 hours
post-exposure. These results provide evidence that exposure to SO2 in combination with
NO2 increases airway responsiveness to a subsequent allergen challenge. This effect is
longer in duration than other effects typically associated with exposure to SO2.
Epidemiologic Studies
Unlike controlled human exposure studies, epidemiologic studies inconsistently indicate
SC>2-related lung function decrements in populations with asthma. This inconsistency
applies to previous (U.S. EPA. 2008d) and recent (Tables 5-5 and 5-6) studies, as well as
those involving adults and children with asthma. Epidemiologic studies examined longer
SO2 averaging times and lags and had uncertainty in exposures because the exposures
were estimated from fixed-site monitors. For the few findings of S02-associated lung
function decrements, confounding by moderately to highly correlated PM and NO2
(r = 0.54-0.9) was not examined. A few recent studies address some of these
uncertainties, but they persist in the evidence overall.
5-26
-------
Table 5-5 Epidemiologic studies of lung function in adults with asthma published since the 2008 ISA for Sulfur
Oxides.
Study Population
and Methodological Details
SO2 Exposure
Estimates (ppb)
SO2 Averaging
Time and Lag Day
Effect Estimate (95% CI)
Single-Pollutant Modela
SO2 Copollutant Model
Results and Correlations
tQian et al. (2009b)
Boston, MA; New York City, NY; Philadelphia,
PA; Madison, Wl; Denver, CO; San Francisco,
CA; 1997-1999
N = 154, ages 12-65 yr. 100% persistent
asthma. 1/3 ICS use, 1/3 beta-agonist use,
1/3 placebo use.
Daily measures for 16 wk. Home PEF. Recruited
from clinics as part of an asthma medication trial.
Multiple comparisons—many pollutants, lags,
medication use analyzed.
Monitors avg within
32 km of subject ZIP
code centroid.
Mean (SD): 4.8 (3.9)
75th percentile: 6.2
Max: 32
24-h avg
0
0-2 avg
Change in PEF (L/min)
All subjects: -0.12 (-3.0, 2.7)
ICS: -8.4 (-13, -3.4)
Beta-agonist: 4.4 (-0.49, 9.3)
Placebo: 3.3 (-1.4, 8.0)
All subjects: -1.9 (-5.6, 1.7)
ICS: -13 (-18, -6.4)
Beta-agonist: 6.4 (0.14, 13)
Placebo: 0.85 (-5.2, 6.9)
Persist with: PM10, NO2, orC>3
(ICS users)
PM2.5 not examined.
r= 0.58 NO2, NR for PM10.
tMaestrelli et al. (2011)
Padua, Italy, 2004-2005
N = 32, mean (SD) age 40 (7.5) yr. 81%
persistent asthma. 69% ICS use. 90% atopy.
6 measures over 2 yr. Supervised spirometry.
Recruited from database of beta-agonist users
(>6 times per yr for 3 yr).
Two monitors in city 24-h avg
Medians across 0
seasons: 0.87-2.7
75th percentiles across
seasons: 1.3-4.1
Change in % predicted FEV1
All subjects: -2.1 (-6.6, 2.3)
Nonsmokers: -11 (-40, 18)
No copollutant model
Copollutant correlations NR.
tCanova et al. (2010)
Padua, Italy, 2004-2005
N = 19, ages 15-44 yr. 79% moderate/severe
asthma. 58% ICS use.
Daily measures for five 30-day periods over 2 yr.
Home PEF/FEV1. Part of same cohort as
Maestrelli et al. (2011) above.
Two monitors in city
Mean (SD): 1.4 (1.1)
Max: 4.9
24-h avg
0, 1, 2, 3, 0
0-3 avg
'-1 avg,
Quantitative effect estimates
NR. Figure shows negative
but imprecise associations for
PEF and FEV1 with wide 95%
CIs.
No copollutant model
PM2.5 not examined.
Spearman r= 0.50 CO, 0.51
PM10, 0.54 NO2.
5-27
-------
Table 5-5 (Continued): Epidemiologic studies of lung function in adults with asthma published since the 2008
Integrated Science Assessment for Sulfur Oxides.
tWiwatanadate and Liwsrisakun (2011)
Chiang Mai, Thailand, 2005-2006
N = 121, ages 13-78 yr. 48% moderate/severe
persistent asthma.
Mean (SD): 1.7 (0.62)
90th percentile: 2.4
Monitor within 10 km of 24-h avg
home
4
NR
Only multipollutant models
analyzed
r= 0.23 NO2, -0.07 PM2.5.
Daily measures for 10 mo. Home PEF. Recruited Max: 3.9
from allergy clinics.
Avg = average; CI = confidence interval; CO = carbon monoxide; FE\A| = forced expiratory volume in 1 sec; ICS = inhaled corticosteroid; ISA = Integrated Science Assessment;
max = maximum; n = sample size; N = population number; N02 = nitrogen dioxide; NR = not reported; 03 = ozone; PEF = peak expiratory flow; PM2.5 = particulate matter with a
nominal aerodynamic diameter less than or equal to 2.5 |jm; PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm; r = correlation coefficient;
SD = standard deviation; S02 = sulfur dioxide.
aEffect estimates are standardized to a 10-ppb increase in 24-h avg S02.
fStudies published since the 2008 ISA for Sulfur Oxides.
5-28
-------
Adults. Previous studies were limited to Europe and Asia. A recent study shows an
SCh-associated decrease in lung function in adults with asthma in the U.S. (Qian ct al..
2009b); however, recent studies in Europe and Asia do not show this decrease (Macstrclli
et al.. 2011; Wiwatanadate and Liwsrisakun. 2011; Canova et al.. 2010) (Table 5-5).
Mean and upper percentile SO2 concentrations tended to be lower in recent studies than
in previous studies [e.g., means for 24-h avg 0.87-4.8 ppb vs. 1.6-90 ppb in (Boczcn et
al.. 2005; Neukirch et al.. 1998; Peters et al.. 1996a)l. However, lower concentrations do
not appear to account for the weak recent evidence in adults with asthma because
previous studies with mean SO2 concentrations of 5.2 to 90 ppb did not observe
S02-associated lung function decrements (Park et al.. 2005; Peters et al.. 1996a). Recent
studies did not differ in temporal variability (e.g., ratio of the mean concentration to
standard deviation) in SO2 concentrations, which is the basis of analysis in these repeated
measures studies.
The U.S. multicity study provides supporting evidence but has the same uncertainty in
the exposure estimate as do other studies in adults with asthma. All studies estimated SO2
exposure from fixed-site monitors, either a single monitor or average of many monitors.
Ambient SO2 concentrations tend to show high spatiotemporal variability within a city,
and correlations with personal exposure are low to moderate (Section 3.4.1.3). The
studies did not discuss whether measurements at the monitors adequately represented the
spatiotemporal variability in ambient SO2 concentrations in the study area. Uncertainty is
high in the U.S. study, which averages SO2 concentrations across monitors within 32 km
of subjects' ZIP code centroid (Qian et al.. 2009b). Ambient SO2 concentrations show
large, transient peaks (Section 2.5.3). which may be important based on results from
controlled human exposure studies showing that 5- to 10-minute exposures to
200-600 ppb SO2 induce rapid and short-lived lung function decrements. Epidemiologic
studies examined same-day (lag 0) SO2 concentrations. Daily average SO2 concentrations
may not represent peak exposures or capture the transient effects of peak exposures
implicated in controlled human exposure studies.
Some recent studies that did not observe S02-related lung function decrements had small
sample sizes (N = 19 or 32) (Maestrelli et al.. 2011; Canova et al.. 2010). It is unclear,
however, whether sample size explains the inconsistency among adults with asthma or
AHR overall. Similarly sized studies (Boezen et al.. 2005; Neukirch et al.. 1998)
observed associations, and larger studies do not show evidence for association
(Wiwatanadate and Liwsrisakun. 2011; Park et al.. 2005; Peters et al.. 1996a). In panel
studies, the number of repeated measurements is also important, and Canova et al. (2010)
measured lung function for five 30-day periods. Many studies that had a large number of
repeated measurements examined lung function measured by subjects at home not
supervised by a trained technician. Results were inconsistent for both methodologies.
5-29
-------
A few epidemiologic studies add information on response modification by asthma
phenotype but produce no clear finding. Previous results found an association between
SO2 exposure and decreased lung function in adults with AHR and elevated IgE (Boezen
ct al.. 2005). and an association between SO2 exposure and increased AHR in adults with
physician-diagnosed asthma and AHR (Taggart et al.. 1996). A recent study did not find
an association between SO2 exposure and lung function decrements in adults with
physician-diagnosed asthma (Macstrclli et al.. 2011). Most of the subjects in Taggart et
al. (1996) and Maestrelli et al. (2011) were atopic. A 10-ppb increase in 24-h avg SO2
was associated with a -2.1 point change [95% confidence interval (CI): -6.6, 2.3] in
percent predicted FEVi. Like the controlled human exposure studies, the epidemiologic
studies do not clearly show that SCh-associated lung function decrements depend on
asthma severity. An association was observed in adults with mild to moderate asthma
(Neukirch et al.. 1998). and the results varied among populations with more severe
asthma (Maestrelli et al.. 2011; Canova et al.. 2010; Oian et al.. 2009b). In contrast with
the controlled human exposure studies, the U.S. asthma medication trial observed an
SCh-related decrease in lung function in adults randomized to daily inhaled corticosteroid
(ICS) use [-8.4 L/minute (95% CI: -13, -3.4) change in peak expiratory flow (PEF) per
10-ppb increase in 24-h avg SO2] (Qian et al.. 2009b). Decrements were not observed in
the beta-agonist or placebo groups (Table 5-5). These two groups had more frequent
asthma exacerbation during the study than the corticosteroid group but similar PEF and
mean age (Lazarus et al.. 2001). All three groups had persistent asthma. Thus, a clear
explanation for the pattern of SO2 associations is not apparent. There is no clear rationale
for attributing null findings to the lack of analysis stratified by corticosteroid use,
particularly for results that were adjusted for such use (Maestrelli et al.. 2011; Canova et
al.. 2010).
Across studies, the potential influence of copollutants is largely unaddressed. No study in
adults with asthma examined PM2 5 total mass, and previous studies observed lung
function decrements in association with larger sized PM metrics that were highly
correlated with SO2 concentrations (r = 0.8-0.9) and sulfate (Neukirch et al.. 1998; Peters
ct al.. 1996a). That some cities had a coal-fired power plant or used coal for heating may
explain some of the high correlations with PM and moderate correlations with NO2
(r = 0.54) (Neukirch et al.. 1998; Taggart et al.. 1996). Copollutant interactions were not
assessed. Only the recent U.S. study analyzed confounding. SO2 was moderately
correlated with NO2 [r = 0.58, correlations with other pollutants NR)] (Oian et al..
2009b). SO2 was negatively associated with PEF in the corticosteroid group, and effect
estimates persisted with adjustment for PM10, NO2, or O3 (Table 5-5). However, inference
from the results is weak due to numerous comparisons across pollutants, lags, and
medication groups and questionable reliability in the exposures estimated from monitors
up to 32 km away.
5-30
-------
Children. As with adults, evidence from neither the 2008 ISA for Sulfur Oxides (U.S.
EPA. 2008d) nor recent studies (Table 5-6) consistently links increases in ambient SO2
concentration with lung function decrements in children with asthma, including recent
U.S. multicity studies (Ierodiakonou et al.. 2015; O'Connor et al.. 2008). The
inconsistency does not appear to be explained by lung function measured under
supervised conditions or by subjects at home, asthma severity, or prevalence of asthma
medication use. In contrast to adults with asthma, S02-associated lung function
decrements were not observed in children with asthma who took inhaled corticosteroids
(Ierodiakonou et al.. 2015; Liu et al.. 2009b). Among children with asthma in Windsor,
ON, the association was limited to nonusers (Liu et al.. 2009b). For some recent studies,
including a U.S. multicity study, inference about an SO2 effect is weak because the
association was isolated to one lung function parameter or exposure lag among numerous
lung function parameters, lags, pollutants, and/or asthma medication groups examined
(Ierodiakonou et al.. 2015; Wiwatanadate and Trakultivakorn. 2010).
5-31
-------
Table 5-6 Epidemiologic studies of lung function in children with asthma published since the 2008 ISA for
Sulfur Oxides.
SO2 Averaging
Study Population and Methodological
SO2 Exposure
Time and Lag
Effect Estimate (95% CI)
SO2 Copollutant Model Results
Details
Estimates (ppb)
Day
Single-Pollutant Modela
and Correlations
tGreenwald et al. (2013)
Monitor at school
24-h avg
Percent change in FEV1
No copollutant model
El Paso, TX, Mar-Jun 2010
A: residential area
0-3 avg
A: 15 (-60, 210)
PM2.5 not associated
N = 38, mean age 10 yr. 47% daily asthma
B: 91 m from major
B: -31 (-52, -2.0)
Pearson r= -0.14 BC, -0.22 NO2,
medication use.
road
-0.07 BTEX, 0.14 cleaning product
Weekly measures for 13 wk. Supervised
Mean (SD): 1.2 (0.44)
VOCs.
spirometry. Recruited from schools.
and 0.84 (0.54)
Upper percentiles NR.
tDales et al. (2009)
Two monitors avg
12-h avg
Percent change in FEV1
Persists with: PM2.5, NO2, or O3.
Windsor, ON, Oct-Dec 2005
99% homes within
8 a.m.-8 p.m.
Bedtime: 0 (-0.92, 0.93)
Pearson r= 0.43 PM2.5, 0.31 NO2.
N = 182, ages 9-14 yr. 37% ICS use,
10 km of sites.
Diurnal: -1.41 (-2.73, -0.08)
35% beta-agonist use.
Median: 4.5
8 p.m.-8 a.m.
Bedtime: -0.17 (-0.98, 0.65)
Daily measures for 4 wk. Home FEV-i. Recruited
95th percentile: 16
from schools. Mean 1.6 and 2.2 h/day spent
8-h avg
Morning: 0.63 (-0.28, 1.55)
outdoors for two study groups.
12 a.m.-8 a.m.
24-h avg
Bedtime: -0.14 (-1.03, 0.76)
tLiuetal. (2009b). Liu (2013)
Two monitors avg
24-h avg
Percent change
FEF25-75%, lag 0-2 avg
Windsor, ON, Oct-Dec 2005
99% homes within
0
FEV1: -0.46 (-2.0, 1.1)
Persists with: O3
N = 182, ages 9-14 yr. 37% ICS use,
10 km of sites.
FEF25-75%: -1.5 (-4.7, 2.0)
Does not persist with: PM2.5 or NO2
35% beta-agonist use.
Median: 4.5
Spearman r- 0.56 PM2.5, 0.18 NO2,
Weekly measures for 4 wk. Supervised
95th percentile: 16
0-2 avg
Change in percent predicted
-0.02 O3.
spirometry. Same cohort as Dales et al. (2009)
FEV1: -2.0 (-4.6, 0.74)
above.
FEF25-75%: -5.7 (-11, -2.2)
5-32
-------
Table 5-6 (Continued): Epidemiologic studies of lung function in children with asthma published since the 2008
ISA for Sulfur Oxides.
Study Population and Methodological
Details
SO2 Exposure
Estimates (ppb)
SO2 Averaging
Time and Lag
Day
Effect Estimate (95% CI)
Single-Pollutant Modela
SO2 Copollutant Model Results
and Correlations
tO'Connor et al. (2008)
Inner-City Asthma Study cohort: Boston, MA;
Bronx, NY; New York City, NY; Chicago, IL;
Dallas, TX; Tucson, AZ; Seattle, WA;
1998-2001
N = 861, ages 5-12 yr. 100% persistent
asthma. 100% atopy.
Daily measures for four 2-wk periods. Home
FEV1/PEF. Recruited from intervention study.
Monitors avg close to
home and not near
industry.
Median 2.3 km to site.
Quantitative SO2 data
NR.
24-h avg
1 -5 avg
Change in percent predicted
FEVi: -1.29 (-2.04, -0.54)
PEF: -1.73 (-2.49, -0.96)
No association for lag 1.
No copollutant model
r= 0.37 PM2.5, 0.59 NO2.
tAmadeo et al. (2015)
Pointe-a-Pitre, Guadeloupe, 2008-2009
N = 71, ages 8-13 yr.
Cross-sectional. Supervised spirometry.
Recruited from schools.
Monitors in city
Number and distance
NR
Mean (SD): 1.8 (1.4)
Max: 4.9
24-h avg
0-13 avg
Change in prerun PEF (L/min)
93 (-28, 214)
Percent change post 6-min run
-1.6 (-36, 33)
No copollutant model
PM2.5 not examined.
Copollutant correlations NR.
tlerodiakonou etal. (2015)
Childhood Asthma Management Program
cohort: Boston, MA; Baltimore, MD; St. Louis,
MO; Denver, CO; Albuquerque, NM; San Diego,
CA; Toronto, ON, 1993-1999
N = 1,003, ages, 5-12 yr. 100% mild/moderate
asthma. 30% ICS use. 30% mast cell inhibitor
use.
14 measures over 4 yr. Supervised spirometry.
Recruited from clinics. Multiple
comparisons—many pollutants, lags, exposure
durations, medication use analyzed.
Nearest monitor
within 50 km of ZIP
code centroid.
Medians across cities:
2-6
90th percentiles
across cities: 5-24
24-h avg Change in percent predicted
0 Prebronchodilator FEVi
All subjects 0.25 (-0.13, 0.63)
ICS: 0.38 (-0.30, 1.1)
Post-bronchodilator FEVi
ICS: 0 (-0.73, 0.75)
Change in methacholine that
induces a 20% drop in FEVi
Mast cell inhibitor:
-13% (-25, 1.3)
No copollutant model
PM2.5 not examined.
Spearman r across
cities = 0.19-0.34 CO, -0.41 to
-0.05 Os, 0.15-0.54 NO2.
5-33
-------
Table 5-6 (Continued): Epidemiologic studies of lung function in children with asthma published since the 2008
ISA for Sulfur Oxides.
Study Population and Methodological
Details
SO2 Exposure
Estimates (ppb)
SO2 Averaging
Time and Lag
Effect Estimate (95% CI)
SO2 Copollutant Model Results
Day
Single-Pollutant Modela
and Correlations
24-h avg
Change in PEF (L/min)
Daily avg PEF lag 4
Persists with: O3
0
Evening PEF
PM2.5 not associated
4
-8.1 (-25, 9.2)
r= -0.04 O3, -0.07 PM2.5, 0.38 CO,
-21 (-38, -4.1)
0.23 NO2.
tWiwatanadate and Trakultivakorn (2010)
Chiang Mai, Thailand, 2005-2006
N = 31, ages 4-11 yr. 100% with symptoms in
previous yr. 52% mild intermittent asthma
Daily measures for 1 yr. Home PEF. Recruited
from allergy clinic. Multiple comparisons—many
pollutants, lags, lung function parameters
analyzed.
Monitor within 25 km
of home
Mean (SD): 1.7 (0.62)
90th percentile: 2.4
Max: 3.9 ppb
Daily avg PEF
-0.3 (-15, 15)
-18 (-32, -2.8)
Avg = average; BC = black carbon; BTEX = benzene, toluene, ethylbenzene, xylene; CI = confidence interval; CO = carbon monoxide; FEF25-75% = forced expiratory flow at 25-75%
of forced vital capacity; FE\A| = forced expiratory volume in 1 sec; ICS = inhaled corticosteroid; ISA = Integrated Science Assessment; max = maximum; n = sample size;
N = population number; N02 = nitrogen dioxide; NR = not reported; 03 = ozone; PEF = peak expiratory flow; PM25 = particulate matter with a nominal aerodynamic diameter less than
or equal to 2.5 |jm; r= correlation coefficient; SD = standard deviation; S02 = sulfur dioxide; VOC = volatile organic compound.
aEffect estimates are standardized to a 10-ppb increase in 8-h to 24-h avg S02.
fStudies published since the 2008 ISA for Sulfur Oxides.
5-34
-------
Considerations in Interpreting Epidemiologic Evidence for Lung Function. A few
recent studies aimed to address uncertainty in the exposure estimates, lag structure of
associations, or copollutant confounding (Grccnwald et al.. 2013; Dales et al.. 2009; Liu
et al.. 2009b) and provide limited indication of SCh-associated lung function decrements.
For children in El Paso, TX, Greenwald et al. (2013) measured SO2 concentrations inside
and outside of two schools. For children attending the school near a major road, a 10-ppb
increase in lag 0-3 avg SO2 was associated with a -31% change (95% CI: -52, -2.0) in
FEVi. This is the largest effect estimate among children or adults with asthma, but a
10-ppb increase in 4-day avg SO2 is unlikely in the area [school mean 0.84 (SD: 0.54)
ppb]. Results are inconsistent for 24-h avg SO2 assigned from monitors 2.3 to 50 km
from children's homes or schools (Amadeo et al.. 2015; Ierodiakonou et al.. 2015; Dales
et al.. 2009; Liu et al.. 2009b; O'Connor et al.. 2008). Lung function decreased with
increases in SO2 concentrations at a monitor located a median distance of 2.3 km from
children's homes (O'Connor et al.. 2008) but not a monitor within 50 km of children's
ZIP code centroid (Ierodiakonou et al.. 2015) (Table 5-6). The studies did not discuss the
adequacy of using monitors at these distances to represent temporal variation in SO2
exposure. No association was observed with the change in PEF after a 6-minute exercise
period (Amadeo et al.. 2015). but this protocol does not mimic controlled human
exposure studies because PEF was examined in relation to 13-day avg SO2.
The lag structure of associations between ambient SO2 concentration and observed
reductions in lung function varies among epidemiologic studies. Previous studies
reported associations with same day (i.e., lag 0) SO2 concentrations (Delfino et al..
2003b; Peters et al.. 1996a). Recent studies point to a potential prolonged response to SO2
exposure through evidence of associations with 3-to 5-day avg SO2 concentrations
(Greenwald et al.. 2013; Liu et al.. 2009b; O'Connor et al.. 2008) that are larger in
magnitude than those for lag 0 or 1 day (Table 5-6).
Correlations of copollutants [PM2 5, PM10, sulfate, black carbon (BC), organic carbon
(OC), total suspended solids (TSP), NO2 or volatile organic compound (VOCs)] with SO2
were moderate (r = 0.56-0.59) in some recent studies (Liu et al.. 2009b; O'Connor et al..
2008) and high in previous studies (r = 0.8-0.9) (Delfino et al.. 2003b; Peters et al..
1996a). SO2 averaging times varied across studies, making it difficult to assess whether
higher correlations were due to higher air pollution levels in the past. Correlations were
weak for school measurements but were not reported specifically for a school near a
major road (Greenwald et al.. 2013) where confounding by BC and VOCs could be more
likely. Copollutant confounding and interactions are poorly studied, and unstudied for
children living near a coal-fired power plant (Peters et al.. 1996a). SO2 and O3
measurements at fixed-site monitors were not correlated (r = -0.02), and SO2
associations persisted with adjustment for O3 (Dales et al.. 2009; Liu et al.. 2009b). A
5-35
-------
recent study adds information on SO2 results adjusted for correlated copollutants. Among
children with asthma in Windsor, ON, the SO2 association persisted with adjustment for
PM2 5 or NO2 for 12-h avg SO2 (Dales et al.. 2009) but not 24-h avg SO2 (Liu. 2013; Liu
et al.. 2009b) (Table 5-6). Associations for PM2 5 were robust to SO2 adjustment, but
inference about confounding is weak due to the moderate SO2-PM2 5 correlation
(r = 0.56) and the potential differences in exposure error for SO2 and PM2 5, which were
made up to 10 km from subjects' homes. Weak inference also applies to results in a Los
Angeles, CA cohort not supporting an association for SO2 after adjustment for benzene
[-34 L/minute change in PEF (95% CI: -120, 52) per 40-ppb increase in 1-h max SO2]
(Delfino et al.. 2003b). SO2 was highly correlated with benzene (r = 0.70), and pollutants
were measured up to 4.8 km from home or school.
Summary of Lung Function Changes in Populations with Asthma
Controlled human exposure studies provide strong evidence for S02-induced lung
function decrements in adults with asthma under increased ventilation conditions.
Short-term exposures for 5-10 minutes to 0.2-0.3 ppm SO2 resulted in 5-30% of
exercising individuals with asthma experiencing moderate or greater decrements (defined
in terms of a >15% decrease in FEVi or >100% increase in sRaw; Table 5-2). Exposures
for 5-10-minutes to SO2 at concentrations >0.4 ppm results in moderate or greater
decrements in lung function in 20-60% of exercising individuals with asthma. A group
of responders (defined as having >15% decrease in FEVi after exposure to 0.6 or 1.0 ppm
SO2) showed statistically significant decrements in FEV1 following exposure for
5-10 minutes to 0.3 ppm SO2 (Table 5-3). Less evidence is available from controlled
human exposure studies to assess SC>2-induced lung function decrements in children with
asthma. However, school-aged children (-5-11 years of age), particularly boys and
perhaps obese children, might be expected to experience greater responsiveness
(i.e., larger decrements in lung function) following exposure to SO2 than normal-weight
adolescents and adults.
For both adults and children with asthma, epidemiologic evidence is inconsistent for lung
function decrements associated with ambient SO2 concentrations (Tables 5-5 and 5-6).
However, one study found an association between ambient SO2 concentration and lung
function decrements in a population with AHR and elevated IgE, and another found an
association between ambient SO2 concentration and AHR in a population with asthma
and a high prevalence of atopy. Evidence from animal toxicological studies, provides
coherence for the timing of effects observed in recent epidemiologic studies. Specifically,
when examining associations between ambient SO2 concentration and observed
reductions in lung function, the pattern of associations was found to vary among
epidemiologic studies, with some studies reporting associations with same day SO2
5-36
-------
concentrations and recent studies pointing to a potential prolonged response (Table 5-6).
The potential prolonged effect of SO2 is supported by rodent studies demonstrating
enhanced allergic inflammation after repeated SO2 exposures. Allergic inflammation may
mediate lung function decrements and provide biological plausibility for the
epidemiologic associations observed due to multiday SO2 concentrations, particularly in
populations with elevated IgE or atopy. Findings of increased airway responsiveness
could not be attributed to exposure to SO2 alone as epidemiologic studies did not examine
copollutant confounding, and controlled human exposure studies only examined SO2 and
NO2 coexposures. A limitation across epidemiologic studies is the uncertainty in the SO2
exposure estimates. A recent study observed an association with SO2 measured at
children's schools, but others used monitors located 2.3-50 km from subjects' homes or
schools. It is unclear whether the SO2 concentrations at the available fixed-site monitors
adequately represent the variation in personal exposure, especially if peak exposures are
as important as indicated by controlled human exposure studies. The influence of
copollutants on epidemiologic results remains largely uncharacterized, including
associations in populations with AHR and elevated IgE or asthma and a high prevalence
of atopy, and populations living near SO2 sources. S02-related lung function decrements
in adults and children with asthma are inconsistently observed after adjustment for PM2 5,
PM10, or NO2, but the implications of these results are unclear because of the uncertainty
in the SO2 exposure estimates and potential differences in exposure error for PM2 5
(Table A-l).
Respiratory Symptoms in Populations with Asthma
The 2008 SOx ISA (U.S. EPA. 2008d) reported strong evidence for the effects of SO2
exposure on respiratory symptoms in controlled human exposure studies in individuals
with asthma under increased ventilation conditions. The same studies also observed
SCh-induced decrements in lung function, although respiratory symptoms occurred
consistently at higher SO2 concentrations (Table 5-2). No new controlled human
exposure studies have been reported since the previous ISA. The available epidemiologic
studies do not provide insight into the concurrence between lung function and symptom
changes. In contrast to evidence from controlled human exposure studies, previous and
recent epidemiologic evidence for S02-associated increases in respiratory symptoms is
weak in adults with asthma. However, epidemiologic evidence supports associations in
children with asthma, and recent studies add evidence for estimates of SO2 exposure at
school and/or home. Overall, the influence of copollutants remains largely unexamined.
5-37
-------
Controlled Human Exposure Studies
As reviewed in the 2008 ISA for Sulfur Oxides and the 1986 Supplement to the Second
Addendum (U.S. EPA. 2008d. 1994). controlled human exposure studies demonstrate
increases in incidence or severity of respiratory symptoms (i.e., cough, chest tightness,
throat irritation) in individuals with asthma exposed to SO2 concentrations between 0.2
and 0.6 ppm for 5-10 minutes during exercise (Tables 5-2 and 5-7). Statistically
significant increases are observed at SO2 concentrations >0.4 ppm [e.g., Linn et al.
(1983b)].
Table 5-7 Study-specific details from controlled human exposure studies of
respiratory symptoms.
Study
Disease Status; n;
Sex; (Agea)
Exposure Details
(Concentration; Duration)
Time of Symptom
Assessment
Balmes et al. (1987)
Asthma; n = 8; 6 M, 2 F
(23-39 yr)
0, 0.5, or 1 ppm SO2 for 1, 3, or 5 min
during eucapnic hyperpnea (60 L/min)
After exposure
Gonq et al. (1995)
Asthma;
n = 14; 12 M, 2 F;
(27 ±11 yr)
0, 0.5, or 1.0 ppm SO2 with light, medium,
and heavy exercise (avg ventilation 30, 36,
and 43 L/min) for 10 min
Before, during, and
immediately after
exposure
Gonq et al. (1996)
Asthma;
n = 10; 2 M, 8 F;
(30.3 ± 9.2 yr)
0 or 0.75 ppm SO2 with exercise (29 L/min)
for up to 24 h with or without pretreatment
with salmeterol (long-acting p2-agonist)
Before and
immediately after
exposure
Gonq et al. (2001)
Asthma;
n = 11; 2 M, 9 F;
(30.8 ±11.3 yr)
0 or 0.75 ppm SO2 for 10 min with exercise
(35 L/min) with or without pretreatment to
montelukast sodium (10 mg/day for 3 days)
Before, immediately
after, and 1 and 2 h
after exposure
Horstman et al. (1988)
Asthma;
n = 12 M;
(28.6 ± 5.5 yr)
0 or 1.0 ppm SO2 for 0, 0.5, 1.0, 2.0, and
5.0 min with exercise (treadmill, 40 L/min)
Before and
immediately after
exposure
Kehrl et al. (1987)
Asthma;
n = 10 M;
(26.8 ± 4.4 yr)
0 or 1 ppm SO2 for 1 h with exercise
(3x10 min, 41 L/min, treadmill)
Before and during
exposure/exercise
Koeniq et al. (1980)
Asthma;
n = 9; 7 M, 2 F;
(15.7 ± 1.1 yr)
0 or 1 ppm SO2 with 1 mg/m3 of NaCI
droplet aerosol, 1 mg/m3 NaCI droplet
aerosol for 60-min exposure with
mouthpiece at rest
Before, during, and
immediately after
exposure
Koeniq et al. (1981)
Asthma;
n = 8; 6 M, 2 F;
(14-18 yr)
0 or 1 ppm SO2 with 1 mg/m3 of NaCI
droplet aerosol, 1 mg/m3 NaCI droplet
aerosol for 30-min exposure via mouthpiece
at rest followed by 10 min exercise on a
treadmill (sixfold increase in Ve)
Before, during, and
immediately after
exposure
5-38
-------
Table 5-7 (Continued): Study specific details from controlled human exposure
studies of respiratory symptoms.
Disease Status; n;
Exposure Details
Time of Symptom
Study
Sex; (Agea)
(Concentration; Duration)
Assessment
Koenia et al. (1983)
Phase 1:
Phase 1:
Before and
Asthma with EIB;
1 g/m3 of NaCI droplet aerosol, 1 ppm SO2,
immediately after
n = 9; 6 M, 3 F;
1 mg/m3 NaCI, 0.5 ppm SO2 + 1 mg/m3
exposure
(12-16 yr)
NaCI for 30-min exposure via mouthpiece at
Phase 2:
rest followed by 10 min exercise on
Asthma with EIB;
treadmill (five-to sixfold increase in Ve)
n = 7 (sex NR);
Phase 2:
(12-16 yr)
0.5 ppm SO2 + 1 mg/m3 NaCI via a face
mask with no nose clip with exercise
conditions the same as Koenia et al. (1981)
Koenia et al. (1987)
Allergy with EIB;
0 or 0.75 ppm SO2 (mouthpiece) with
Before and
n = 10; 3 M, 7 F;
exercise (33.7 L/min) for 10 min and 20 min
immediately after
(13-17 yr)
prior pretreatment (0 or 180 |jg albuterol)
pretreatment and
exposure
Koenia et al. (1990)
Asthma with EIB;
0.1 ppm SO2 for 15 min preceded by air or
Before and
n = 13; 8 M, 5 F
0.12 ppm O3 for 45 min during intermittent
immediately after
(14.3 ± 1.8 yr)
exercise (2><15 min, 30 L/min, treadmill),
exposure
no control, air exposure
Koenia et al. (1992)
Asthma;
1 ppm SO2 for 10 min with exercise
Before and
n = 8; 2 M, 6 F;
(13.4-31.3 L/min) with or without
immediately after
(27.5 ± 9.6 yr)
pretreatment to theophylline
exposure
Linn et al. (1983b)
Asthma;
0, 0.2, 0.4, or 0.6 ppm SO2 with low
Before and
n = 23; 13 M, 10 F;
humidity or high humidity for 10 min with
immediately after
(23.3 ± 4.4 yr)
exercise (bicycle, 5 min 50 L/min)
exposure
0 or 0.6 ppm SO2 with warm air or cold air
with exercise (bicycle, 50 L/min, ~5 min)
Linn et al. (1983a)
Asthma;
0 or 0.75 ppm SO2 with unencumbered
Before and
n = 23; 15 M, 8 F
breathing and mouth only breathing with
immediately after
(23 ± 4 yr)
exercise (40 L/m, 10 min, bicycle)
exposure
Linn et al. (1984a)
Asthma;
0, 0.3, or 0.6 ppm SO2 at 21°C, 7°C, and
Before, during,
n = 14; 12 M, 2 F
-6°C, RH 80% with exercise (bicycle,
immediately after,
(24.1 ± 4.7 yr)
50 L/min, ~5 min)
and a week after
exposure
Linn et al. (1984c)
Asthma;
0, 0.3, or 0.6 ppm SO2 at 21°C, 7°C, and
Before, immediately
n = 24; 13 M, 11 F;
-6°C and 80% RH with exercise (5 min,
after, and 24 h after
(24.0 ± 4.3 yr)
50 L/min)
exposure
Linn et al. (1984b)
Asthma;
Phase 1:
Phase 1:
Phase 1 (Pilot)
0, 0.2, 0.4, or 0.6 ppm SO2 at 5°C, 50, and
before and
n = 8; 4 M, 4 F;
85% RH with exercise (5 min, 50 L/min)
immediately after
(24.5 ± 3.9 yr)
Phase 2:
exposure
Phase 2
0 and 0.6 ppm SO2 at 5°C and 22°C, 85%
Phase 2:
n = 24; 19 M, 5 F;
RH with exercise (5 min, 50 L/min)
before, immediately
(24.0 ± 4.3 yr)
after, 1 day after,
and 1 wk after
exposure
5-39
-------
Table 5-7 (Continued): Study specific details from controlled human exposure
studies of respiratory symptoms.
Disease Status; n;
Exposure Details
Time of Symptom
Study
Sex; (Agea)
(Concentration; Duration)
Assessment
Linnetal. (1985b)
Asthma;
0 or 0.6 ppm SO2 at 21 and 38°C, 20 and
Before, immediately
n = 22; 13 M, 9 F;
80% RH with exercise (~5 min, 50 L/min)
after, and 24 h after
(23.5 ± 4.0 yr)
exposure
Linn et al. (1985a)
Asthma with COPD;
0, 0.4, or 0.8 ppm SO2 for 1 h with exercise
Before, during,
n = 24; 15 M, 9 F;
(2 x 15 min, bicycle, 18 L/min)
immediately after,
(60 yr;
24 h after, and
Range: 49-68 yr)
7 days after
exposure
Linnetal. (1987)
Healthy;
0, 0.2, 0.4, or 0.6 ppm SO2 for 1 h with
Before and during
n = 24; 15 M, 9 F;
exercise (3 * 10-min, bicycle, -40 L/min)
exposure (after first
(18-37 yr)
exercise and after
Atopic (sensitive to
last exercise)
common airborne
allergens but no
asthma);
n = 21; 12 M, 9 F;
(18-35 yr)
Minimal or mild
asthma; n = 16; 10 M,
6 F;
(20-33 yr)
Moderate or severe
asthma;
n = 24; 10 M, 14 F;
(18-35 yr)
Linnetal. (1988)
Asthma;
Three pretreatment groups
Before, immediately
n = 20; 13 M, 7 F;
(1) metaproterenol sulfate, (2) placebo,
after, 10 min,
(28 ± 5 yr)
(3) no treatment
30 min, 60 min,
0, 0.3, and 0.6 ppm SO2 for 10 min with
120 min, 24 h, and
exercise (bike, 50 L/min)
1 wk after exposure
Linnetal. (1990)
Asthma;
0, 0.3, or 0.6 ppm SO210 min with exercise
Before exposure,
n = 21; 6 M, 15 F;
(50 L/min)
after pretreatment,
(34.8 ± 8.9 yr)
(1) low medication use, (2) normal, (3) high
immediately after,
usual medication supplemented by inhaled
30 min after, and
metaproterenol before exposure
60 min after
exposure
Maanussen et al.
Asthma;
0 or 0.5 ppm SO2 for 20 min. 10-min rest
Before exposure
(1990)
n = 46; 21 M, 25 F;
followed by 10 min isocapnic
and immediately
(28 ± 14 yr)
hyperventilation (30 L/min)
after
hyperventilation
Mvers et al. (1986a)
Asthma;
Three pretreatment groups
Before and after
n = 10; 7 M, 3 F;
(1) 200 mg cromolyn, (2) 20 mg cromolyn,
each 3-min
(27.6 ± 5.5 yr)
(3) placebo
exposure to an
Doubling concentrations of SO2 during
increasing SO2
sequential 3 min exposures, from 0.25 to
concentration
8 ppm
5-40
-------
Table 5-7 (Continued): Study specific details from controlled human exposure
studies of respiratory symptoms.
Study
Disease Status; n;
Sex; (Agea)
Exposure Details
(Concentration; Duration)
Time of Symptom
Assessment
SheDDard et al. (1983)
Asthma;
n = 8; 4 M, 4 F;
(26.6 ± 4.3 yr)
0.5 ppm SO2 for 3 min eucapnic hyperpnea
Before and
immediately after
exposure
Trenaa et al. (1999)
Asthma;
n = 47; 14 M, 33 F;
(21.1 yr;
Range: 18-39 yr)
0.5 ppm SO2 for 10 min with moderate
exercise
Before and
immediately after
exposure
Trenqa et al. (2001)
Asthma;
n = 17; 5 M, 12 F;
(27.4 ± 6.3 yr)
0.5 ppm SO2 for 10 min with moderate
exercise (treadmill)
Before and
immediately after
exposure
COPD = chronic obstructive pulmonary disease; EIB = exercise-induced bronchospasm; F = female; M = male; n = sample size;
NaCI = sodium chloride; NR = not reported; 03 = ozone; ppm = parts per million; RH = relative humidity; SD = standard deviation;
S02 = sulfur dioxide; VE = minute volume.
aRange or Mean ± SD.
Linn et al. (1983b> reported the severity of respiratory symptoms following 5-minute
exposures to 0, 0.2, 0.4, and 0.6 ppm SO2 in heavily exercising individuals with mild to
moderate asthma. Total symptom score changes were significant (0.01 0.5 ppm SO2 demonstrated SC>2-induced
increases in respiratory symptoms. Total and lower respiratory symptom scores were
significantly increased with increasing SO2 concentrations (0, 0.5, and 1.0 ppm SO2)
following 10-minute exposures with varying levels of exercise (Gong et al.. 1995).
Trenga et al. (1999) confirmed these results, observing a significant correlation between
FEVi decrements and increases in respiratory symptoms following 10-minute exposures
to 0.5 ppm SO2 via mouthpiece. Respiratory symptoms have also been observed
following exposure durations as low as 3 minutes to 0.5 ppm SO2 via mouthpiece during
eucapnic hyperpnea (Ve = 60 L/minute), in which seven out of eight individuals with
asthma developed respiratory symptoms (Balmes etal.. 1987).
5-41
-------
As with lung function, increased respiratory symptoms in response to short-term
exposure to SO2 in individuals with asthma is dependent on exercise. Linn et al. (1983b)
reported significant changes in total symptom scores after 0.2 ppm SO2 exposure in
heavily exercising individuals with asthma. In contrast, in adults with asthma at rest, no
association was found between respiratory symptoms (i.e., throat irritation, cough,
wheeze) and 1-hour exposures to 0.2 ppm SO2 (Tunnicliffe et al.. 2003).
Epidemiologic Studies
Compared with controlled human exposure studies, epidemiologic evidence for
S02-associated increases in symptoms is variable, being supportive in children with
asthma but weak in adults with asthma. A recent study of children and adults combined
does not support an association with asthma medication use. The analysis, which only
reported the lack of statistically significant associations, was limited by analysis of
beta-agonist levels in wastewater rather than medication use ascertained for individual
subjects (Fattorc et al.. 2016) (Table 5-8). The evidence base specifically in children with
asthma is larger and more informative, providing results for home and/or school SO2
exposure estimates and temporally resolved SO2 metrics. Also, while they do not settle
questions, studies in children with asthma aim to assess copollutant confounding and
interactions. Although the evidence overall is less consistent in recent than previous
studies, the aforementioned strengths are features of many recent studies of children with
asthma.
Adults. Ambient air SO2 concentrations were lower in recent than previous studies
(0.87-2.7 ppb vs. 1.6-90 ppb for means), but this reduction does not appear to explain
the weak evidence because previous results are also inconsistent [Supplemental
Figure 5S-1 and Table 5S-3 (U.S. EPA. 2017c)l. All studies have uncertainty in the SO2
exposure estimates assigned from a single monitor or averaged across multiple monitors.
No study indicated whether measurements at the monitors adequately represented the
spatiotemporal variability in ambient SO2 concentrations in the study area or the temporal
variation in people's exposures.
5-42
-------
Table 5-8 Epidemiologic studies of respiratory symptoms in populations with asthma published since the 2008
ISA for Sulfur Oxides.
Study Population and
Methodological Details
SO2 Exposure
Estimates (ppb)
SO2 Averaging
Time and
Lag Day
Effect Estimate (95% CI)
Single-Pollutant Modela
SO2 Copollutant Model Results and
Correlations
Adults with asthma
tMaestrelli et al. (2011)
Padua, Italy, 2004-2005
N = 32, mean (SD) age 40 (7.5 yr). 81%
persistent asthma. 69% ICS use. 90% atopy.
Six measures over 2 yr. Symptoms assessed in
clinic. Recruited from database of beta-agonist
users (>6 times per year for 3 yr).
Two monitors in city
Medians across
seasons: 0.87-2.7
75th percentiles
across seasons:
1.3-4.1
24-h avg
0
Asthma control score
Increase = better control
All subjects: 0.77 (-1.1, 2.6)
Nonsmokers: 0.10 (-2.2, 2.4) Copollutant correlations NR
n = 22
No copollutant model
No association with personal or
fixed-site PM2.5
tWiwatanadate and Liwsrisakun (2011)
Chiang Mai, Thailand, 2005-2006
N = 121, ages 13-78 yr. 48% moderate/severe
persistent asthma.
Daily diary for 10 mo. Recruited from allergy
clinics. Multiple comparisons—many pollutants,
lags, health endpoints analyzed.
Monitor within 10 km
of home
Mean (SD): 1.7 (0.62)
90th percentile: 2.4
Max: 3.9
24-h avg
SO2 increment NR. Results Not statistically significant with: NO2
reported only for statistically Quantitative results NR.
PM2 5 not examined.
r= 0.23 NO2 and PM10
significant lags
2
Daytime symptoms
OR: 0.90 (0.81, 0.99)
5
Nighttime symptoms
OR: 1.16 (1.04, 1.29)
24-h avg
Cough
0
0.67 (0.34, 1.31)
2
2.19 (1.34, 3.54)
0-2 avg
2.53 (1.05, 6.08)
tAnvenda et al. (2016)
Kanazawa, Japan, Jan-June 2011
N = 83, ages 23-84 yr. 54% atopy.
Daily diary for mean 153 days. Recruited from
hospital outpatients.
One monitor in city
Mean (SD): 1.6 (1.3)
Max: 7.3
Persists with: PAH or NO2 (lag 2)
Spearman r= 0.60 PAH, 0.56 NO2
5-43
-------
Table 5-8 (Continued): Epidemiologic studies of respiratory symptoms in populations with asthma published
since the 2008 ISA for Sulfur Oxides.
Study Population and
Methodological Details
SO2 Exposure
Estimates (ppb)
SO2 Averaging
Time and
Lag Day
Effect Estimate (95% CI)
Single-Pollutant Modela
SO2 Copollutant Model Results and
Correlations
Children with asthma
tSDira-Cohen et al. (2011). SDira-Cohen (2013)
Monitor at school
1-h max (a.m.)
Cough
Cough
Bronx, NY, 2002-2005
Concentrations NR
0
RR: 1.60 (1.20, 2.12)
Does not persist with: school EC
N = 40, ages 10-12 yr. 44% with asthma ED
Most children walk to
Wheeze
PM2.5 not associated
visit or hospital admission in previous 12 mo.
school
RR: 1.81 (1.15, 2.84)
r= 0.45 EC
Daily diaries for 1 mo. Recruited from schools
Shortness of breath
by referrals from school nurses.
RR: 1.45 (0.90, 2.84)
tVelicka etal. (2015)
Five monitors and
24-h avg
Cough
No copollutant model
Ostrava, Czech Republic, Nov 2013-Feb 2014
dispersion model
0
OR: 0.92 (0.74, 1.17)
PM2.5 not examined.
N = 147, ages 6-18 yr. 67% mild persistent
0.5 x 0.5 km
Breathing difficulty-wheeze
Copollutant correlations NR
asthma. 33% moderate persistent asthma. 79%
resolution
OR: 2.29 (1.55, 3.39)
atopy. 97% regular asthma medication use.
Weighted avg by time
Reliever inhaler use
Daily diaries for 4 mo. Recruited from clinics.
at home and school
Median: 4.0
75th percentile: 12
OR: 1.84 (1.32, 2.56)
Restricted activities
OR: 1.25 (1.00, 1.62)
tDales etal. (2009)
Two monitors avg
24-h avg
OR for SO2 S8.8 vs. <2.3 ppb
No copollutant model
Windsor, ON, Oct-Dec 2005
99% homes within
Chest tightness
Quantitative results NR
N = 182, ages 9-14 yr. 37% ICS use. 35%
10 km of sites
1.30 (1.06, 1.58)
beta-agonist use.
Median: 4.5
ORs for difficulty breathing,
Daily diaries for 4 wk. Recruited from schools.
95th percentile: 16
cough, and wheeze reported
Mean 1.6 and 2.2 h/day spent outdoors.
not statistically significant
tO'Connor et al. (2008)
Monitors avg close to
24-h avg
Wheeze-cough
No copollutant model
Inner-City Asthma Study cohort: Boston, MA;
home and not near
1-19 avg
RR: 1.05 (0.89, 1.23)
r= 0.59 NO2, 0.32 CO, 0.37 PM2.5
Bronx, NY; New York City, NY; Chicago, IL;
industry
Nighttime asthma
Dallas, TX; Tucson, AZ; Seattle, WA;
Median 2.3 km to site
RR: 1.11 (0.91, 1.36)
1998-2001
Quantitative SO2 data
Slow play
N = 861, ages 5-12 yr. 100% persistent
NR
RR: 1.06 (0.88, 1.27)
asthma. 100% atopy.
Missed school
Daily diaries for four 2-wk periods. Recruited
RR: 1.10 (0.82, 1.49)
from intervention study.
5-44
-------
Table 5-8 (Continued): Epidemiologic studies of respiratory symptoms in populations with asthma published
since the 2008 ISA for Sulfur Oxides.
Study Population and
Methodological Details
SO2 Exposure
Estimates (ppb)
SO2 Averaging
Time and
Lag Day
Effect Estimate (95% CI)
Single-Pollutant Modela
SO2 Copollutant Model Results and
Correlations
tGent et al. (2009)
New Haven county, CT, 2000-2004
N = 149, ages 4-12 yr. 45% intermittent
asthma.
Daily diaries reported monthly for 1 yr.
Recruited from larger cohort, clinic, and school.
Monitor 0.9-30 km of
home
Mean 10 km to site
Concentrations NR
24-h avg
0
NR
Only multipollutant model analyzed
with six PM2.5 component factors
r= 0.45 motor vehicle factor
Children and adults with asthma
tFattore et al. (2016)
Milan, Italy, Sep-Dec2013
N = 84 days
Daily wastewater samples for 84 days analyzed
for levels of the beta-agonist salbutamol.
Three monitors avg
Mean (SD): 2.2 (1.3)
Max: 5.9
24-h avg
Oto 10
(Single-day)
Beta-agonist levels in
wastewater
No quantitative results. RRs
reported not statistically
significant
No copollutant model
Pearson r= 0.66 PM2.5, 0.65 PM10
Avg = average; CI = confidence interval; CO = carbon monoxide; EC = elemental carbon; ED = emergency department; ICS = inhaled corticosteroids; ISA = Integrated Science
Assessment; max = maximum; n = sample size; N = population number; N02 = nitrogen dioxide; NR = not reported; OR = odds ratio; PAH = polycyclic aromatic hydrocarbon;
PM2.5 = particulate matter with a nominal aerodynamic diameter less than or equal to 2.5 |jm; PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10
|jm; r = correlation coefficient; RR = relative risk; SD = standard deviation; S02 = sulfur dioxide.
aEffect estimates are standardized to a 10-ppb increase in 24-h avg S02 and 40-ppb increase in 1-h max S02.
fStudies published since the 2008 ISA for Sulfur Oxides.
5-45
-------
All epidemiologic studies of adults examined 24-h avg SO2 concentrations (Table 5-8).
longer than the 5-10-minute exposures used in controlled human exposure studies
(Table 5-2). As in previous studies, recent epidemiologic evidence does not indicate
associations for respiratory symptoms with same-day (lag 0) SO2 concentrations
(Anvenda et al.. 2016; Maestrelli et al.. 2011). Atopy was prevalent in Maestrelli et al.
(2011) (90%); previous findings supported an association in adults with AHR and
elevated IgE (Boezen et al.. 2005). A recent study linked an increase in SO2 concentration
to an increase in nighttime asthma symptoms with a 5-day lag (Wiwatanadate and
Liwsrisakun. 2011). but inference is weak because results were inconsistent among the
many lags, pollutants, and health effects examined. Also, SO2 exposures were assessed
from a monitor up to 10 km from subjects' homes. There is some consistency for SO2
concentrations lagged 2 or 5 days, or averaged over 3 or 5 days, [Supplemental
Figure 5S-1 and Table 5S3 (U.S. EPA. 2017c) and Anvenda et al. (2016)1. PM metrics
also were associated with symptoms and moderately to highly correlated with SO2
(r = 0.60-0.9) (Boezen et al.. 2005; Neukirch et al.. 1998; Peters et al.. 1996a). Whether
the magnitude of copollutant correlations influences the consistency of association for
SO2 with respiratory symptoms in adults with asthma cannot be determined in this small
evidence base. As examined only in a recent study, SO2 associations persisted with
adjustment for polycyclic aromatic hydrocarbon (PAH) or NO2 (Anvenda et al.. 2016).
However, uncertainty in the exposures estimated from a single monitor and a different
site for PAH limits the inferences that can be drawn about an independent association for
SO2. Controlled human exposure studies show that symptoms resolve once exposure
ends, but SC>2-induced allergic inflammation could be a pathway by which SO2 exposure
induces symptoms after several days or over multiple days.
Children. Overall, the epidemiologic evidence indicates associations between higher SO2
concentrations and increased respiratory symptoms in children with asthma, particularly
when effects are examined as a composite index of multiple symptoms (Figure 5-2).
Associations also are observed for asthma medication use or activity restriction but not
consistently for wheeze or cough. Results vary in magnitude and precision (Figure 5-2).
In some study areas, the SO2 concentrations were much lower (Spira-Cohen et al.. 2011;
Delfino et al.. 2003a; Delfino et al.. 2003b) or higher (Mortimer et al.. 2002) than the
10-ppb increment used to standardize the effect estimates. Although recent studies give
inconsistent results (Table 5-8). associations are observed with SO2 exposure estimates
that are measured or modeled for the school or home. Recent studies reported lower SO2
concentrations than many previous studies [for 24-h avg, median ~4 ppb vs. means 8.3
and 90 ppb in (Segala et al.. 1998; Romieu et al.. 1996)1. It is unclear whether the
inconsistency is due to lower concentrations; previous studies observed associations in
locations with similar SO2 concentrations [median 24-h avg 2.2-7.4 ppb in Schildcrout et
al. (2006). mean 8-h max 4.6 ppb in Delfino et al. (2003a). Delfino et al. (2003b)l.
5-46
-------
Study
Wheeze
Exposure Assessment
tSpira-Cohen etal. (2011)School
Cough
tSpira-Cohen etal. (2011)School
tVelicka et al. (2015) Modeled home/school
Romieuetal. (1996) Monitor within 5 km
Segala et al. (1998) Average of 11 monitors
Composite ofsvmptoms
tVelicka et al. (2015) Modeled home/school
Delfinoetal. (2003a)a Monitor within 4.8 km
Delfino et al. (2003b)
Romieuet al. (1996)
Boezenetal. (1999)
Mortimer et al. (2002)
Segala et al. (1998)a
Monitor within 4.8 km
Monitor within 5 km
1 monitor
Average of city monitors
Average of 11 monitors
Peters etal. (1996) 1 monitor
Schildcrout et al. (2006) Monitors within 80 km
tO'Connor et al. (2008) Monitors within median 2.3
km
Asthma Medication
tVelicka etal. (2015)
Segala et al. (1998)a
Modeled home/school
Average of 11 monitors
-20
129 (55, 229)
136(16, 381)
100
Percent increase (95% confidence interval)13
aThe two results for Delfino et al. (2003a) refer to symptoms not interfering with activity and symptoms interfering with activity. The
two results for Segala et al. (1998) refer to children with mild asthma and children with moderate asthma.
bEffect estimates are standardized to a 10-ppb increase in 24-h avg sulfur dioxide concentration and a 40-ppb increase in 1-h max
concentrations. Study details are presented in Table 5-8. Results from Gent et al. (2009) are not presented in the figure because
they are based on a multipollutant model. Corresponding quantitative results are reported in Supplemental Table 5S-4 (U.S. EPA,
2017c).
Note: f and red = recent studies published since the 2008 ISA for Sulfur Oxides, black = studies from the 2008 ISA for Sulfur
Oxides.
Figure 5-2 Associations between short-term average ambient sulfur dioxide
concentrations and respiratory symptoms and asthma medication
use in children with asthma.
Spira-Cohen et al. (2011) is notable not only for monitoring SO2 at schools but also for
examining 1-h max concentrations. In the population of children in Bronx, NY, increases
in SO2 were linked to increased odds of cough and wheeze but not shortness of breath
(Table 5-8). Previous U.S. studies also associated symptoms with temporally resolved
SO2 metrics [i.e., 1-h max, 8-h max, 3-h avg (8-11 a.m.)] but had more uncertainty in
5-47
-------
exposures estimated from monitors up to 4.8 km from children's homes/schools (Delfino
et al.. 2003a; Delfino et al.. 2003b') or monitors averaged across the city (Mortimer ct al..
2002). Spira-Cohen et al. (2011) did not report SO2 concentrations to compare to
previous studies but reported that most children walked to school. Velicka et al. (2015)
also aimed to improve exposure assessment for children in Ostrava, Czech Republic. A
dispersion model and five monitors were used to estimate SO2 concentrations at 0.5-km
resolution and calculate a time-weighted 24-h avg for each child based on the school and
home location. SO2 was associated with breathing difficulty-wheeze, reliever inhaler use,
and restricted activities, but not cough (Table 5-8). The study population had a high
prevalence of atopy (79%); thus, results agree with Segalaetal. (1998). However,
compared to the previous studies, Velicka et al. (2015) may have less uncertainty in
exposure estimates (Section 3.5).
Other recent studies largely do not provide evidence for S02-associated increases in
respiratory symptoms in children with asthma (Dales et al.. 2009; Gent et al.. 2009;
O'Connor et al.. 2008). But, they are limited because of (1) the large distance between the
SO2 monitor and children's homes (e.g., up to 10 km, median 2.3 km, mean 10 km); (2) a
lack of quantitative results (Dales et al.. 2009); (3) use of 19-day avg SO2 concentrations,
which are more subject to residual temporal confounding (O'Connor et al.. 2008); or
(4) use of SO2 only as part of a multipollutant model with six PM2 5 component source
factors (Gent et al.. 2009).
For the associations observed between SO2 and respiratory symptoms in children with
asthma, including those with atopy, the influence of copollutants is poorly addressed.
Symptoms were not associated with personal or school PM2 5 but with other PM metrics,
including PM10, elemental carbon (EC), OC, black smoke (BS), and TSP. Associations
were also observed with NO2, VOCs such as benzene and xylene, and O3 (Table 5-8).
Except for O3, these copollutants were moderately to highly correlated with SO2
(r = 0.45-0.9). Correlations were highest in previous studies, but recent studies did not
report SO2 concentrations (Spira-Cohen et al.. 2011) or copollutant correlations (Velicka
et al.. 2015) to assess whether the magnitude of correlation varied by SO2 levels.
Copollutant models were analyzed in few studies and for few copollutants. For a Los
Angeles, CA cohort, no SO2-VOC interaction was indicated, and SO2 associations
persisted with adjustment for benzene, xylene, or toluene for some but not all symptoms
(Delfino et al.. 2003a; Delfino et al.. 2003b). Associations for VOCs were attenuated as
well, and copollutant model results were uncertain because of the moderate to high
correlations with SO2 (r = 0.58-0.78) and because exposures were assessed from
monitors 4.8 km from children's homes or schools. Potential exposure error also limits
inference from results showing associations for joint increases in SO2 with PM10, NO2, or
carbon monoxide (CO) that were similar to each single-pollutant association (Schildcrout
5-48
-------
ct al.. 2006). The recent Bronx, NY study analyzed copollutant models for school SO2
and EC, which may have more comparable exposure error. SO2 and EC were moderately
correlated (r = 0.45), consistent with the location in a high diesel truck traffic area (data
were obtained for 2002-2005, before the Diesel Fuel Standard went into effect in 2006,
see Section 2.2.3) (Spira-Cohen et al.. 2011). In the copollutant model, the odds ratio
(OR) for cough was robust for EC but decreased in magnitude and precision for SO2 from
1.60 (95% CI: 1.20, 2.12) to 1.32 (95% CI: 0.93, 1.87) per 40-ppb increase in 1-hmax
S02.
Summary of Respiratory Symptoms in Populations with Asthma
Controlled human exposure studies provide strong evidence for the effects of SO2
exposure on respiratory symptoms in adults with asthma under increased ventilation
conditions. Exposures for 5-10 minutes to 0.2-0.6 ppm SO2 induced respiratory
symptoms in exercising individuals with asthma, with the most consistent evidence from
exposures to 0.4-0.6 ppm SO2 (Table 5-2). Epidemiologic evidence in adults with asthma
is weak, but increases in ambient SO2 concentration are generally associated with
increased risk of asthma symptoms in children (Figure 5-2; Table 5-8). Assessing
coherence specifically with controlled human exposure studies of adolescents with
asthma is difficult because those studies lacked an appropriate control exposure. Limited
findings support associations in children and adults with AHR and elevated IgE.
Epidemiologic results in children are less consistent in recent than previous studies but
support associations for 1-h max SO2 measured at schools or 24-h avg SO2 modeled for
school and home. School or home SO2 measures may better represent exposures than the
concentrations at fixed-site monitors examined in most studies, particularly for 1-h max.
These SO2 metrics are longer than the 5-10 minute SO2 exposures in controlled human
exposure studies, which show transient responses. And, the extent to which confounding
or an interaction with copollutants such as PM2 5, EC, NO2, and VOCs contributed to
epidemiologic associations, including those for populations with asthma and a high
prevalence of atopy or AHR and elevated IgE, and for residents near a coal-fired power
plant, is not fully characterized in the epidemiologic studies. However, an independent
effect of SO2 exposure is indicated by experimental evidence in rodents of allergic
inflammation enhanced by repeated 1-hour exposures to 2 ppm SO2.
Hospital Admission and Emergency Department Visits for Asthma
Since the completion of the 2008 SOx ISA, epidemiologic studies have continued to
examine the association between short-term exposure to ambient SO2 concentrations and
respiratory-related hospital admissions and ED visits, but are primarily limited to
5-49
-------
single-city studies. The sections within this chapter detailing the respiratory-related
hospital admissions and ED visits studies characterize recent studies in the context of the
collective body of evidence evaluated in the 2008 SOx ISA. The 2008 SOx ISA (U.S.
EPA. 2008d) included the first thorough evaluation of respiratory morbidity in the form
of respiratory-related hospital admissions and ED visits, including asthma. These studies
reported generally positive associations with short-term SO2 exposures, with associations
that are often larger in magnitude for children (Figure 5-3). Additionally, SO2
associations with asthma hospital admissions and ED visits were often attenuated, but
remained positive in copollutant models with PM, NO2, or O3.
Within this section focusing on asthma, as well as the rest of the chapter,
respiratory-related hospital admissions and ED visit studies are evaluated separately
because only a small percentage of respiratory-related ED visits result in hospital
admission. Additionally, when evaluating asthma ED visit and hospital admission studies
that focus on children (i.e., defined age ranges <18 years of age), it is important to note
the difficulty of reliably diagnosing asthma in children <5 years of age. Thus, including
children under the age of 5 years in a study population may result in the overestimation of
the number of asthma ED visit and hospital admissions (NAEPP. 2007).
For each of the studies evaluated in this section, Table 5-9 presents the air quality
characteristics of each city, or across all cities, the exposure assignment approach used,
and information on copollutants examined in each asthma hospital admission and ED
visit study. Other recent studies of asthma hospital admissions and ED visits are not the
focus of this evaluation because these studies were conducted in small single-cities,
encompassed a short study duration, had insufficient sample size, or did not examine
potential copollutant confounding. The full list of these studies, as well as study-specific
details, can be found in Supplemental Table 5S-6 (U.S. EPA. 2017c).
5-50
-------
Stuffy
tSon et al. (2013)
TSon et al. (2013)
Lin et al. (2004)
tSamoli et al. (2011)
Sheppard et al. (1999; 2003)
+Son et al. (2013)
Son et al. (2013)
Samoli et al. (2011)
Son et al. (2013)
Samoli et al. (2011)
Wilson et al. (2005)
Ito et al. (2007)
Peel et al. (2005)a
AT SDR (2006)
tStieb et al. (2009)
tByers et al. (2015)a
Villeneuve et al. (2007)
tAlhanti et al. (2015)a
Wilson et al. (2005)
tJalaludin et al. (2008)
fLi et al. (2011)
"("Strickland etal. (2010)a
TByers et al. (2015)a
jAlhanti et al. (2015)a
Wilson et al. (2005)
tAlhanti et al. (2015)a
jByers et al. (2015)a
tAlhanti et al. (2015)a
Wilson et al. (2005)
tAlhanti et al. (2015)a
Ito et al. (2007)
tByers et al. (2015)a
Villeneuve et al. (2007)
din et al. (2008)
"[Strickland etal. (2010)a
Jaffe et al. (2003)
Ito et al. (2007)
tByers et al. (2015)a
Villeneuve et al. (2007)
din et al. (2008)
"[Strickland etal. (2010)a
Location
8 South Korean cities
8 South Korean cities
Bronx County, NY
Athens, Greece
Seattle, WA
8 South Korean cities
8 South Korean cities
Athens, Greece
8 South Korean cities
Athens, Geece
Portland, ME
Manchester, NH
New York, NY
Atlanta, GA
Bronx; NY
Manhattan, NY
7 Canadian cities
Indianapolis, IN
Edmonton, Canada
3 U.S. cities
Portland, ME
Manchester, NH
Syndey, Australia
Detroit, MI
Atlanta, GA
Indianapolis, IN
3 U.S. cities
Portland, ME
Manchester, NH
3 U.S. cities
Indianapolis, IN
3 U.S. cities
Portland, ME
Manchester, NH
3 U.S. cities
New York, NY
Indianapolis, IN
Edmonton, Canada
Syndey, Australia
Atlanta, GA
3 Ohio cities
New York, NY
Indianapolis, IN
Edmonton, Canada
Syndey, Australia
Atlanta, GA
Age
All
0-14
0-14
0-14
5-14
<65
75+
All
0-14
All
0-14
All
All
All
All
All
All
All
All
>2
04
0-14
0-14
1-14
2-18
5-17
5-17
5-18
15-64
15-64
19-39
18-44
>44
40-64
65+
65+
65+
All
All
>2
1-14
5-17
5-34
All
All
>2
1-14
5-17
Hospital Admissions
0-3
0-3
0
0-3
0-1
0-2
04
04
2
0-2
04
0-2
0
0
0-1
0-4b
04c
0-2
04
-10.0 0.0 10.0 20.0
% Increase (95% Confidence Interval)
ED = emergency department; ISA = Integrated Science Assessment.
a = studies that used a 1-h max exposure metric.
b = time-series results.
0 = case-crossover results.
Note: f and red text/symbols = recent studies published since the 2008 ISA for Sulfur Oxides. Black text/symbols = U.S. and Canadian studies evaluated in the 2008 ISA for Sulfur
Oxides; Circle = all-year; diamond = warm/summer months; square = cold/winter months. Gray shading depicts studies that present results for children (i.e., <18 yr of age).
Corresponding quantitative results are reported in Supplemental Table 5S-5 (U.S. EPA. 2017c).
Figure 5-3 Percent increase in asthma hospital admissions and emergency department visits from U.S. and
Canadian studies evaluated in the 2008 SOx ISA and recent studies in all-year and seasonal
analyses for a 10-ppb increase in 24-h avg or 40-ppb increase in 1-h max sulfur dioxide
concentrations.
5-51
-------
Table 5-9 Study-specific details and mean and upper percentile concentrations
from asthma hospital admission and emergency department visit
studies conducted in the U.S. and Canada and evaluated in the 2008
SOx ISA and studies published since the 2008 SOx ISA.
Upper
Mean
Percentile
Location
Exposure
Concentration
Concentrations
Copollutants
Study
years
Assignment
Metric
PPb
PPb
Examination
Hospital admissions
Lin et al. (2004)
Bronx
Avg of SO2
24-h avg
Cases: 16.8
NR
NR
County, NY
concentrations
Controls: 15.6
(1991-1993)
from two
monitoring sites
Sheppard et al.
Seattle, WA
Avg of SO2
24-h avg
8.0
75th: 10.0
Correlation
(1999), Sheppard
(1987-1994)
concentrations
90th: 13.0
(r):
(2003)
from multiple
PM10: 0.31
monitors
PM2.5: 0.22
PM10-2.5: 0.34
Os: 0.07
CO: 0.24
Copollutant
models: none
tSon et al. (2013)
Eight South
Avg of hourly
24-h avg
3.2-7.3
NR
Correlation
Korean cities
ambient SO2
(r):
(2003-2008)
concentrations
PM10: 0.5
from monitors in
each city
O3: -0.1
NO2: 0.6
CO: 0.6
Copollutant
models: none
tZhenq et al. (2015)
Meta-
NR
24-h avg
3.1-45.5®
NR
Correlations
analysis
(r): NR
(1988-2014)
Copollutant
models: none
tSamoli et al. (2011)
Athens,
Avg of SO2
24-h avg
6.4
75th: 8.4
Correlation
Greece
concentrations
(r):
(2001-2004)
across multiple
Os: -0.19
monitors
NO2: 0.55
Copollutant
models:
PM10, SO2,
NO2, O3
5-52
-------
Table 5-9 (Continued): Study specific details and mean and upper percentile
concentrations from asthma hospital admission and
emergency department visit studies conducted in the U.S.
and Canada and evaluated in the 2008 SOx ISA and studies
published since the 2008 SOx ISA.
Upper
Mean Percentile
Location Exposure Concentration Concentrations Copollutants
Study years Assignment Metric ppb ppb Examination
ED visits
Jaffe et al. (2003) Cincinnati, When more than 24-h avg Cincinnati:
Cleveland, one monitoring
and station operating
Columbus, in a day, monitor
OH reporting highest
(1991-1996) 24-h avg SO2
concentration
used
13.7
Cleveland:
15.0
Columbus:
4.2
Max: Correlations
Cincinnati: 50 (r) (range
Cleveland: 64 across cities)
Columbus: 22 N°2:
0.07-0.28
Os:
0.14-0.26
PM10:
0.29-0.42
Copollutant
models: none
Ito et al. (2007) New York, Average SO2 24-h avg 7.8 75th: 10 Correlations
NY concentrations g5^. u (r): NR
(1999-2002) across Copollutant
19 monitors models:
PM2.5, NO2,
O3, CO
ATSDR (2006)
Bronx and
Manhattan,
NY
(1999-2000)
SO2 concentra-
tions from one
monitor in Bronx
and one in
Manhattan
Note: monitors
used in series
not
simultaneously
24-h avg
Manhattan: 12
Bronx: 11
NR
Correlations
(r):
Bronx:
Os: -0.49
NO2: 0.50
PM2.5: 0.39
Max PM10:
0.0.34
Manhattan:
Os: -0.40
NO2: 0.47
PM2.5: 0.26
PM10: 0.24
Copollutant
models: O3,
FRM and
Max PM2.5,
NO2
5-53
-------
Table 5-9 (Continued): Study specific details and mean and upper percentile
concentrations from asthma hospital admission and
emergency department visit studies conducted in the U.S.
and Canada and evaluated in the 2008 SOx ISA and studies
published since the 2008 SOx ISA.
Upper
Mean Percentile
Location Exposure Concentration Concentrations Copollutants
Study years Assignment Metric ppb ppb Examination
Peel et al. (2005) Atlanta, GA Average of SO2 1-h max
(1993-2000) concentrations
from monitors for
several
monitoring
networks
PM2.5 water
soluble
metals: 0.00
PM2.5 sulfate:
0.08
PM2.5 acidity:
-0.03
PM2.5 OC:
0.18
PM2.5 EC:
0.20
Oxygenated
HCs: 0.14
Os: 0.19
CO: 0.26
NO2: 0.34
Copollutant
models: none
16.5 90th: 39.0 Correlations
(r):
PM2.5: 0.17
PM10: 0.20
PM10-2.5: 0.21
UFP: 0.24
Wilson et al. (2005) Portland, SO2 concentra- 24-h avg Portland: 11.1 NR Correlation
ME, and tions from one Manchester' (r) (Range
Manchester, monitor in each -ir c across cities):
NH city 03-
(1996-2000) 0.05-0.24
Copollutant
models: none
tStieb et al. (2009)
Seven
Canadian
cities
(1992-2003)
Average SO2
concentrations
across all
monitors in each
city. Number of
SO2 monitors in
each city ranged
from 1-11.
24-h avg 2.6-10.0
75th: 3.3-13.4
Correlations
(r) only
reported by
city and
season
Copollutant
models: none
5-54
-------
Table 5-9 (Continued): Study specific details and mean and upper percentile
concentrations from asthma hospital admission and
emergency department visit studies conducted in the U.S.
and Canada and evaluated in the 2008 SOx ISA and studies
published since the 2008 SOx ISA.
Upper
Mean
Percentile
Location
Exposure
Concentration
Concentrations
Copollutants
Study
years
Assignment
Metric
PPb
PPb
Examination
tOrazzo et al. (2009)
Six Italian
Average of SO2
24-h avg
All-year:
NR
Correlations
cities
concentrations
2.1-8.1
(r): NR
(1996-2002)
across all
Warm
Copollutant
monitors in each
(Apr-Sep):
models: none
city
1.3-9.0
Cold
(Oct-Mar):
2.6-7.3
tAlhanti et al. (2016)
Three U.S.
Population-
1-h max
Atlanta: 10.7
NR
Correlations
cities
weighted
Dallas: 2.7
(r): NR
Atlanta, GA
average using
St Louis' 10 7
Copollutant
(1993-2009)
data available
w L. LvUlw. 1 W . /
models: none
Dallas, TX
from all monitors
(2006-2009)
measuring SO2
St. Louis,
MO
(2001-2007)
tZhenq et al. (2015)
Meta-
NR
24-h avg
4.6-39.1®
NR
Correlations
analysis
(r): NR
(1988-2014)
Copollutant
models: none
tStrickland et al.
(2010)
Atlanta, GA
(1993-2004)
Population-
weighted
average using
data available
from all monitors
measuring SO2
1-h max
All-year: 10.8
Warm
(May-Oct): 9.6
Cold
(Nov-Apr):
12.0
NR
Correlations
(r): NR
Copollutant
models: none
tLi et al. (2011)
Detroit, Ml
(2004-2006)
Average of SO2 24-h avg
3.8
75th: 5.1
Correlations
concentrations
Max: 27.3
(r), range
across two
across
monitors in
monitors:
Detroit
CO:
metropolitan
0.17-0.31
area that
PM2.5:
measure SO2
0.40-0.53
NO2:
0.42-0.55
Copollutant
models: none
5-55
-------
Table 5-9 (Continued): Study specific details and mean and upper percentile
concentrations from asthma hospital admission and
emergency department visit studies conducted in the U.S.
and Canada and evaluated in the 2008 SOx ISA and studies
published since the 2008 SOx ISA.
Upper
Mean Percentile
Location Exposure Concentration Concentrations Copollutants
Study years Assignment Metric ppb ppb Examination
tBvers et al. (2015) Indianapolis, Double-weighted 1-h max All-year: 10.1
NR
IN
(2007-2011)
average
(distance from
monitor to ZIP
code centroid
and age-specific
census
population) of
two SO2
monitors
Warm: 10.5
Cold: 9.8
Correlations
(r):
All-year:
PM2.5: 0.34
Warm:
1-h max O3:
0.45
8-h max O3:
0.42
PM2.5: 0.38
Cold:
PM2.5: 0.29
tVilleneuve et al. Edmonton, Average of SO2 24-h avg Summer Summer Correlations
(2007) AB concentrations (Apr-Sep) 75th: 3.0 (r): NR
(1992-2002) across three 50th: 2.0 Winter Copollutant
monitoring Winter 75th: 4.0 models: NR
stations (Oct-Mar)
50th: 3.0
tJalaludin et al.
(2008)
Sydney,
Australia
(1997-2001)
Average of SO2
concentrations
across
14 monitoring
stations
24-h avg
All-year: 1.07
Warm: 1.03
Cold: 1.1
Max
All-year: 4.1
Warm: 4.1
Cold: 3.9
Correlations
(r): (warm,
cold)
PM10: 0.37,
0.46
PM2.5: 0.27,
0.46
Os: 0.45,
-0.04
CO: 0.46,
0.51
NO2: 0.52,
0.56
Copollutant
models:
PM10, PM2.5,
Os, CO, NO2
5-56
-------
Table 5-9 (Continued): Study specific details and mean and upper percentile
concentrations from asthma hospital admission and
emergency department visit studies conducted in the U.S.
and Canada and evaluated in the 2008 SOx ISA and studies
published since the 2008 SOx ISA.
Upper
Mean Percentile
Location Exposure Concentration Concentrations Copollutants
Study years Assignment Metric ppb ppb Examination
tSmarqiassi et al.
(2009)
Montreal,
QC
(1996-2004)
SO2 concentra- 24-h avg Regional: 4.3
tions measured
at two monitoring
sites east and
southwest of the
refinery
At-home
estimates of daily
exposure by
estimating SO2
concentrations at
centroid of
residential postal
codes using
AERMOD
East: 6.9
Southwest: 4.4
AERMOD:
East + South-
west: 3.0
East: 3.7
Southwest: 2.4
75th:
Regional: 5.3
East: 9.2
Southwest: 5.9
AERMOD:
East + South-
west: 4.3
East: 5.5
Southwest: 3.0
NR
British
Columbia,
Canada
Average of SO2
24-h avg
24-h avg
concentrations
1-h max
Quebec:
from all monitors
Cases: 2.35
within 7.5 km
from a major
Controls: 2.40
facility (i.e.,
British
refinery, smelter,
Columbia
pulp mill)
Cases: 2.04
Controls: 2.23
Total:
Cases: 2.04
Controls: 2.23
NR
Correlations
(r): NR
Copollutant
models: none
5-57
-------
Table 5-9 (Continued): Study specific details and mean and upper percentile
concentrations from asthma hospital admission and
emergency department visit studies conducted in the U.S.
and Canada and evaluated in the 2008 SOx ISA and studies
published since the 2008 SOx ISA.
Upper
Mean Percentile
Location Exposure Concentration Concentrations Copollutants
Study years Assignment Metric ppb ppb Examination
tWinquist et al.
(2014)
Atlanta, GA,
U.S.
(1998-2004)
Population-
weighted
average using
data available
from all monitors
measuring SO2
1-h max
Warm
(May-Oct): 8.3
Cold
(Nov-April):
10.8
75th:
Warm: 11.4
Cold: 14.6
Correlations
(r):
Warm:
Os: 0.27
CO: 0.32
NO2: 0.44
PM2.5: 0.28
EC: 0.31
Sulfate: 0.24
Secondary
PM2.5: 0.24
Cold:
Os: 0.05
CO: 0.22
NO2: 0.41
PM2.5: 0.07
EC: 0.18
Sulfate: 0.02
Secondary
PM2.5: 0.08
Copollutant
models: none
tPearce et al. (2015) Atlanta, GA SO2 1-h max 14.6 NR Correlations
concentrations (r):
from one monitor NP
Copollutant
models: none
Outpatient and physician visits
tBurra et al. (2009) Toronto. ON
Average of SO2 1-h max
9.7
75th: 12.0
Correlations
(1992-2001)
concentrations
95th: 35.0
(r): NR
across six
monitors
Max: 62.0
Copollutant
models: none
5-58
-------
Table 5-9 (Continued): Study specific details and mean and upper percentile
concentrations from asthma hospital admission and
emergency department visit studies conducted in the U.S.
and Canada and evaluated in the 2008 SOx ISA and studies
published since the 2008 SOx ISA.
Upper
Mean Percentile
Location Exposure Concentration Concentrations Copollutants
Study years Assignment Metric ppb ppb Examination
tSinclair et al. (2010) Atlanta, GA, SO2 concentra- 1-h max 1998-2000: NR Correlations
U.S. tions collected as 19.3 (r): NR
(1998-2002) part of AIRES at 2000-2002: Copollutant
SEARCH 17g models: none
Jefferson street „„„„
site 1998-2002:
18.3
AERMOD = American Meteorological Society/U.S. EPA Regulatory Model; AIRES = Aerosol Research Inhalation Epidemiology
Study; avg = average; CO = carbon monoxide; EC = elemental carbon; FRM = Federal Reference Method; HCs = hydrocarbons;
ISA = Integrated Science Assessment; N02 = nitrogen dioxide; NR = not reported; 03 = ozone; OC = organic carbon;
PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm; PM2.5 = particulate matter with a
nominal aerodynamic diameter less than or equal to 2.5 |jm; PM10-2 5 = particulate matter with a nominal aerodynamic diameter
less than or equal to 10 |jm and greater than 2.5 |jm; r = Pearson correlation coefficient; SEARCH = Southeast Aerosol Research
Characterization; S02 = sulfur dioxide; UFP = ultrafine particle.
aRange of mean concentrations across all studies included in the meta-analysis.
f = studies published since the 2008 ISA for Sulfur Oxides.
Hospital Admissions
The 2008 SOx ISA identified only two U.S.-based studies and no Canadian studies that
examined the association between short-term SO2 exposures and asthma hospital
admissions. These studies reported positive associations; however, they were limited to
studies of individual cities (Figure 5-3). The asthma hospital admission studies averaged
SO2 concentrations over multiple monitors and only examined 24-h avg exposure
metrics, which may not adequately capture the spatial and temporal variability in SO2
concentrations (Sections 3.4.2.2 and 3.4.2.3). While correlations between 24-h avg and
1-h max SO2 concentrations are high (r > 0.75) at most monitors, lower correlations may
occur at some monitors and in individual studies, adding uncertainty to the ability of 24-h
avg metrics to capture peak SO2 concentrations. Additionally, relatively few studies have
examined the potential confounding effects of other pollutants on the S02-asthma
hospital admissions relationship.
To date, a limited number of studies have been published since the 2008 SOx ISA that
focus on the relationship between short-term SO2 exposures and asthma hospital
admissions. In a time-series study conducted in Athens, Greece, Samoli et al. (2011)
evaluated the association between multiple ambient air pollutants and pediatric asthma
hospital admissions for ages 0-14 years. In an all-year analysis, the authors reported a
positive association with SO2 [16.5 % (95% CI: 2.3, 32.6); lag 0 increase for a 10-ppb
5-59
-------
increase in 24-h avg SO2 concentrations]. In copollutant analyses, the authors found SO2
risk estimates to be robust in models with PM10 [13.0% (95% CI: -1.5, 29.7)] and O3
[16.5% (95% CI: 2.3, 32.6)]. However, models with NO2 showed an increase in the SO2
risk estimate [21.3% (95% CI: 1.1, 45.5)]. SO2 was low (r < 0.4) to moderately (r ranging
from 0.4-0.7) correlated with other pollutants examined in the study, with the highest
correlation with NO2 (r = 0.55).
The association between short-term SO2 exposures and asthma hospital admissions was
also examined by Son et al. (2013) in a study of eight South Korean cities. In addition to
focusing on asthma, the authors examined allergic disease hospital admissions, which
encompass asthma. For all ages, the authors reported a 5.3% increase (95% CI: -2.4,
13.0) in asthma hospital admissions for a 10-ppb increase in 24-h avg SO2 concentrations
and a 3.1% increase (95% CI: -3.7, 10.7) in allergic diseases hospital admissions. In
analyses focusing on children (ages 0-14 years) and older adults (>75 years of age), the
authors reported associations that were larger in magnitude, compared to all ages for both
asthma and allergic diseases hospital admissions (Figure 5-3).
The evidence from studies evaluated in the 2008 SOx ISA, as well as recent studies
indicating a positive association between short-term SO2 exposure and asthma hospital
admissions, is supported by a meta-analysis conducted by Zheng et al. (2015) that
focused on all studies examining air pollution and asthma hospital admissions and ED
visits published between 1988 and 2014. For SO2, the authors reported a 2.1% increase
(95% CI: 0.5, 3.70) in asthma hospital admissions for a 10-ppb increase in 24-h avg SO2
concentrations based on estimates from 31 studies. The results from Zheng et al. (2015)
are smaller in magnitude compared to the other asthma hospital admission studies
summarized in Figure 5-3. perhaps due to the meta-analysis only including single-day lag
estimates from each of the studies. The results of the meta-analysis were found to be
robust in sensitivity analyses examining publication bias; however, the publication bias
analysis was not conducted separately for asthma hospital admissions and ED visits
results.
Emergency Department Visits
The majority of studies, examining respiratory-related hospital admissions and ED visits,
have focused on asthma ED visits. Studies evaluated in the 2008 SOx ISA were primarily
limited to single-city studies that provided generally positive associations between SO2
and asthma ED visits, with positive associations being reported in some study locations
and evidence of no association in other locations (Figure 5-3). Additionally, there was
limited evidence for potential seasonal differences in SO2 associations with asthma ED
visits. As with the hospital admission studies, there has been limited analyses examining
the potential confounding effects of copollutants on the SCh-asthma ED visit relationship.
5-60
-------
Recent studies that examined the association between short-term SO2 exposures and
asthma ED visits have primarily focused on either children or the entire population, with
a few studies examining whether effects differ by lifestage. Additionally, unlike the
hospital admission studies, the ED visit studies examined both 24-h avg and 1-h max
exposure metrics, which can provide some additional insight, on a population level, into
the short-term exposures that result in respiratory effects in controlled human exposure
and animal toxicological studies (see previous subsections of Section 5.2.1.2).
Strickland et al. (2010) examined the association between SO2 exposure and pediatric
asthma ED visits (ages 5-17 years) in Atlanta, GA, using air quality data over the same
years as Tolbert et al. (2007). who examined all respiratory ED visits. However, unlike
Tolbert et al. (2007). who used a single-site monitor, Strickland et al. (2010) used
population-weighting, a more refined exposure assignment approach, to combine daily
pollutant concentrations across monitors. As discussed in Section 3.4.4.2. a study by
Goldman et al. (2012) shows that the bias in health effect estimates decreases when using
population-weighted averages instead of the values from a fixed-site monitor for
assigning exposure. In Strickland et al. (2010). the authors developed a statistical model
using hospital-specific, time-series data that is essentially equivalent to a time-stratified,
case-crossover analysis (i.e., using interaction terms between year, month, and
day-of-week to mimic the approach of selecting referent days within the same month and
year as the case day). Strickland et al. (2010) observed a 4.2% (95% CI: -2.1, 10.8)
increase in ED visits for a 40-ppb increase in 1-h max SO2 concentrations at lag 0-2 days
in an all-year analysis. The potential confounding effects of other pollutants on the
S02-asthma ED visit relationship was not assessed in this study, and correlations
between pollutants were not presented. However, when evaluating the correlation of
pollutants examined over the same study years in Tolbert et al. (2007). SO2 had a low
correlation with all pollutants (r < 0.36).
Positive associations between short-term SO2 exposures and pediatric asthma ED visits
were also observed in a study conducted by Li et al. (2011) in Detroit, MI that focused on
whether there was evidence of a threshold in the air pollution-asthma ED visit
relationship. In the main nonthreshold analysis, the authors conducted both time-series
and time-stratified case-crossover analyses. Li etal. (2011) observed similar results in
both analyses, which indicated an association between SO2 and asthma ED visits, [time
series: 20.5% (95% CI: 8.9, 33.2); lag 0-4 for a 10-ppb increase in 24-h avg SO2
concentrations; case-crossover: 22.8% (95% CI: 12.6, 33.7); lag 0-4], The results of the
U.S.-based studies focusing on children conducted by Strickland et al. (2010) and Li et al.
(2011) are consistent with those of Jalaludin et al. (2008) in a study of children
1-14 years of age conducted in Sydney, Australia. In addition to conducting the analysis
focusing on ages 1-14 years, the authors also examined whether risks varied among age
5-61
-------
ranges within this study population (Chapter 6). Jalaludin et al. (2008) examined single-
day lags ranging from 0 to 3 days as well as the average of 0-1 days. In the 1-14 years of
age analysis, the authors observed slightly larger associations at lag 0-1 days [29.7%
(95% CI: 14.7, 46.5)] compared to lag 0 [22.0% (95% CI: 9.1, 34.5)] for a 10-ppb
increase in 24-h avg SO2 concentrations. An examination of the potential confounding
effects of other pollutants was assessed in copollutant models with PM10, PM2.5, O3, CO,
or NO2 at lag 0. SO2 was found to be weakly to moderately correlated with these
pollutants, r = 0.27-0.52. Jalaludin et al. (2008) reported that the SCh-asthma ED visit
association was slightly attenuated, but remained positive in all copollutant models, with
the magnitude of the association ranging from a 13.2-16.1% increase in asthma ED
visits.
Bvers et al. (2015). in a study conducted in Indianapolis, IN, examined asthma ED visits
across all ages as well as various lifestages (i.e., 5-17, 18-44, and >45 years of age). The
authors used a double-weighted approach to assign exposure by first weighting air
pollution concentrations by distance from a monitor to the ZIP code centroid and then
weighting the concentrations by the age-specific census population. In an all-year
analysis for all ages, the authors reported a 0.4% increase in asthma ED visits (95% CI:
-3.6, 4.5) at lag 0-2 for a 40-ppb increase in 1-h max SO2 concentrations, with evidence
of a larger association when focusing on pediatric asthma ED visits [5.4% (95% CI: -3.2,
14.5); lag 0-2], which is consistent with Strickland et al. (2010). Li etal. (2011). and
Jalaludin et al. (2008). Although copollutant analyses were not conducted, SO2 was found
to have a low correlation with PM2 5 (/' < 0.4) in all-year and seasonal analyses, and
moderate correlation with 1-h max and 8-h max O3 in warm season analyses
(r = 0.42-0.45). Additionally, when examining SO2 concentrations across the entire study
period, the authors noted that only 36 days (i.e., 2.1% of days) had 1-h max SO2
concentrations that exceeded the NAAQS.
Alhanti et al. (2016) also used the approach of assigning exposure using
population-weighting similar to Strickland et al. (2010). but expanded the study area from
Atlanta, GA to include two additional cities: Dallas, TX and St. Louis, MO. The analysis
focused on examining whether there was evidence of differential risk across lifestages
(i.e., 0-4, 5-18, 19-39, 40-64, and 65+ years of age) for asthma ED visits across a
number of air pollutants, including SO2. Analyses were conducted for each individual
city, and an overall estimate across all three cities was calculated by taking the
inverse-variance weighted average of the city-specific risk estimate. Across the
individual cities, there was evidence of positive and negative associations for all age
categories examined except ages 5-18 years for which positive associations were
observed across all cities, which is consistent with the single-city studies detailed above.
In the combined analysis across the three cities, Alhanti et al. (2016) reported positive
5-62
-------
associations for ages 0-4 years [4.1% (95% CI: -0.8, 9.2); lag 0-2 for 40-ppb increase in
1-h max SO2 concentrations] and 5-18 years [5.7% (95% CI: -0.8, 11.8); lag 0-2]
(Sarnat. 2016). In sensitivity analyses, the results were found to be robust to alternative
model specifications for both control for temporal trends and weather covariates.
As detailed in the asthma hospital admissions section, Zheng et al. (2015) conducted a
meta-analysis of asthma hospital admission and ED visit studies. In the analysis focusing
on ED visit studies, the authors reported a 3.5% increase (95% CI: 1.9, 5.1) in asthma ED
visits for a 10-ppb increase in 24-h avg SO2 concentrations based on single-day lag
estimates from 34 studies. This result is in the range of risk estimates reported in studies
that observed positive associations between short-term SO2 exposure and asthma ED
visits (Figure 5-3).
Although a number of recent studies add to the evidence from the 2008 SOx ISA
indicating a positive association between asthma ED visits and short-term SO2 exposures,
not all studies have reported positive associations. Both Stieb et al. (2009) and Villeneuve
et al. (2007). in studies conducted in seven Canadian cities and Edmonton, AB,
respectively, did not observe evidence of a positive association between short-term SO2
exposures and asthma ED visits (Figure 5-3 and Table 5-9). The evidence of no
association was observed over multiple lag structures (i.e., both single and multiday lags)
(Stieb et al.. 2009; Villeneuve et al.. 2007) as well as subdaily exposure metrics (i.e., 3-h
avg pollutant concentrations) (Stieb et al.. 2009).
Hospital Admissions and Emergency Department Visits for Respiratory
Conditions Associated with Asthma
As stated previously, asthma is difficult to diagnose in children less than 5 years of age
(NAEPP. 2007); however, asthma-like symptoms in children within this age range are
often presented in the form of transient wheeze. Although studies that examine ED visits
for wheeze do not directly inform upon the relationship between short-term SO2
exposures and asthma, they can add supporting evidence. Orazzo et al. (2009) examined
the association between short-term SO2 exposures and wheeze ED visits, in children
(ages 0-2 years) in six Italian cities. In a time-stratified case-crossover analysis, Orazzo
et al. (2009) examined associations for multiday lags ranging from 0-1 to 0-6 days. The
authors reported the strongest evidence for an association between short-term SO2
exposures and wheeze ED visits at lags of 0-3 to 0-6 days with estimates ranging from
2.1 to 4.3%, respectively, for a 10-ppb increase in 24-h avg SO2 concentrations. Within
this study, copollutant analyses or correlations with other pollutants were not presented.
Smargiassi et al. (2009) and Brand et al. (2016) also provided additional information on
whether there is an association between short-term SO2 exposures and health effects that
may be closely related to asthma. The distinction between asthma and asthma-related
5-63
-------
outcomes is made in this case because of the focus on children 2-4 years of age in
Smargiassi et al. (2009). which examined asthma hospital admissions and ED visits, and
Brand et al. (2016). which examined hospital admissions for the combination of asthma
and bronchiolitis. An asthma exacerbation for children 2-4 years of age may not
necessarily represent the same health outcome as those studies discussed earlier in this
section which include older individuals in whom asthma is more easily diagnosed.
In both Smargiassi et al. (2009) and Brand et al. (2016). the authors aimed to examine
whether industrial sources of air pollution result in higher exposures to air pollutants,
including SO2, and subsequently an increase in asthma-related hospital admissions and
ED visits. Within Smargiassi et al. (2009). the authors examined the influence of a point
source of SO2 (i.e., stack emissions from a refinery) in Montreal on asthma hospital
admissions and ED visits using data from two fixed-site monitors as well as estimates of
SO2 concentrations from a dispersion model [American Meteorological
Society/Environmental Protection Agency Regulatory Model (AERMOD)]. The authors
examined both daily mean and daily peak SO2 concentrations. When comparing SO2
concentrations at one monitoring site east of the refinery with those obtained via
AERMOD, the authors observed a modest correlation (daily mean SO2, r = 0.43; daily
peak SO2, r = 0.36). An examination of hospital admissions and ED visits for locations
east and southwest of the refinery found that associations with SO2 estimates from
AERMOD were slightly larger in magnitude for the same-day daily peak [hospital
admissions: 1.46 (95% CI: 1.10, 1.93); ED visits: 1.18 (95% CI: 1.05, 1.33) fora40-ppb
increase in 1-h max SO2 concentrations] compared to daily mean concentrations [hospital
admissions: 1.36 (95% CI: 1.05, 1.81); ED visits: 1.15 (95% CI: 1.02, 1.27) for a 10-ppb
increase in 24-h avg SO2 concentrations] in an unadjusted model. When examining
associations using SO2 concentrations from ambient monitors, Smargiassi et al. (2009)
did not find consistent evidence of an increase in asthma hospital admissions or ED
visits. The authors also examined an adjusted model to control for daily weather variables
and all other regional pollutants (i.e., PM2 5, SO2, NO2, and O3), but these results are not
presented because, as discussed within this ISA, the evaluation of potential copollutant
confounding is limited to two-pollutant models because the results from multipollutant
models are difficult to interpret due to multicollinearity between pollutants. However, the
results from the unadjusted (i.e., single-pollutant model) and adjusted models were
generally similar.
Whereas Smargiassi et al. (2009) focused on a population residing near two refineries in
Montreal, Brand et al. (2016) examined the association between air pollutant emissions
and concentrations and asthma-related hospital admissions from a number of industrial
facilities (i.e., metal smelters, pulp mills, and oil refineries) in both Montreal and British
Columbia. To capture the potential influence of air pollutants, including SO2, from
5-64
-------
industrial facilities on asthma-related hospital admissions, the authors limited the analysis
to air quality monitors and the population residing within 7.5 km of a facility. In a
time-stratified, case-crossover analysis, the authors reported no evidence of an
association when examining the relationship between 24-h avg and 1-h max SO2
concentrations at lag 0 from "any industry" in individual city analyses as well as a pooled
analysis across both cities with ORs ranging from 0.79-0.88 for a 10 ppb increase in 24-h
avg SO2 concentrations and 0.79-0.86 for a 40 ppb increase in 1-h max SO2
concentrations. Only the analysis in Quebec examined associations between SO2 from
smelters and asthma-related hospital admissions, and as a result was excluded from the
"any industry" pooled analysis. Brand et al. (2016) reported no evidence of an association
in analyses using 24-h avg [OR = 0.88 (95% CI: 0.50, 1.50)] and 1-h max [OR = 0.79
(95% CI: 0.48, 1.25)] SO2 concentrations that were identified as being from smelters.
Collectively the results from Smargiassi et al. (2009) and Brand et al. (2016) provide
initial evidence that SO2 emitted from industrial facilities, as captured by measurements
from ambient monitors, is not associated with an increase in asthma-related hospital
admissions and ED visits. However, these results may reflect the fact that a monitor may
not adequately capture spatial and temporal variability in SO2 concentrations, including
peak exposures of residents (see Section 3.4.2).
Outpatient and Physician Visits Studies of Asthma
Several recent studies examined the association between ambient SO2 concentrations and
physician or outpatient (nonhospital, non-ED) visits for asthma. In Toronto, Burra et al.
(2009) examined asthma physician visits among patients aged 1-17 and 18-64 years in a
study focusing on differences by sex and income within each age category. For children,
the authors reported evidence of consistent positive associations between short-term
increases in SO2 concentrations and asthma physician visits for most of the single and
multiday lags examined (i.e., 0, 0-1, 0-2, 0-3), with no evidence of an association for a
0-4-day lag. In the analysis of adults, a similar pattern of associations was observed;
however, there was no evidence of an association at the two longest lags examined, 0-3
and 0-4 days.
In a study conducted in Atlanta, GA, Sinclair etal. (2010) examined the association
between multiple respiratory outcomes, including asthma and outpatient visits, from a
managed care organization. The authors separated the analysis into two time periods (the
first 25 months of the study period and the second 28 months of the study period) to
compare the air pollutant concentrations and relationships between air pollutants and
acute respiratory visits for the 25-month time period examined in Sinclair and Tolsma
(2004) (i.e., August 1998-August 2000), and an additional 28-month time period of
available data from the Atlanta Aerosol Research and Inhalation Epidemiology Study
5-65
-------
(ARIES) (i.e., September 2000-December 2002). As detailed in Table 5-9. SO2
concentrations were relatively similar between periods, differing by less than 2 ppb. A
comparison of the two time periods indicated that risk estimates across outcomes tended
to be larger in the earlier 25-month period compared to the later 28-month period, with
evidence of consistent positive associations across the lags examined for asthma (both
child and adult), but confidence intervals were relatively large.
Examination of Seasonal Differences
In addition to examining the association between short-term SO2 exposures and asthma
hospital admissions and ED visits in all-year analyses, some studies also conducted
seasonal analyses. When evaluating these studies, it is important to note that the
difference in the geographic locations examined across studies complicates the ability to
draw overall conclusions regarding the seasonal patterns of associations.
In the study of eight South Korean cities, Son et al. (2013) examined potential seasonal
differences across respiratory hospital admission outcomes. For asthma and allergic
disease hospital admissions, the association with SO2 was largest in magnitude during the
summer, although confidence intervals were quite large [asthma: 19.1% (95% CI: -18.3,
73.9), lag 0-3; allergic disease: 21.9% (95% CI: -6.7, 58.6), lag 0-3 for a 10-ppb
increase in 24-h avg SO2 concentrations]. Across the eight cities, mean 24-h avg SO2
concentrations were lowest during the summer season (4.4 ppb compared to a range of
4.8 to 7.0 in the other seasons), which was also observed for NO2, PM10, and CO. The
seasonal asthma hospital admission results of Son et al. (2013) are similar to those
reported in Samoli et al. (2011) in a study conducted in Athens, Greece. Samoli et al.
(2011) observed the largest magnitude of an association during the summer months
[46.6% (95% CI: -13.8, 149.3); lag 0 for a 10-ppb increase in 24-h avg SO2
concentrations], but also reported a similar association in the autumn months [42.6 %
(95% CI: -0.5, 104.4); lag 0], Although positive, associations for the winter and spring
months were smaller in magnitude, 20.2 and 31.8%, respectively.
The initial indication of larger associations during the summer for asthma hospital
admissions is further supported by the analysis of Strickland et al. (2010) examining
short-term SO2 exposures and pediatric asthma ED visits in Atlanta. The authors reported
evidence of asthma ED visit associations larger in magnitude during the summer [10.8%
(95% CI: 0.7, 21.7); lag 0-2 for a 40-ppb increase in 1-h max SO2 concentrations], with
no evidence of an association during the winter [0.4% (95% CI: -7.5, 9.0)]. These results
are consistent with Bvers et al. (2015). who reported associations larger in magnitude in
the summer for all ages [3.1% (95% CI: -2.6, 8.6); lag 0-2 for a 40-ppb increase in 1-h
max SO2 concentrations], and particularly children 5-17 years of age [13.0% (95% CI:
0.8, 26.8); lag 0-2], and no evidence of an association in the cold season across all ages
5-66
-------
examined. However, in another study focusing on asthma physician visits in Atlanta,
Sinclair et al. (2010) reported inconsistent evidence of seasonal differences in risk
estimates, with the pattern of associations being different in each of the time periods
examined in the study. It is important to note that the results of Sinclair et al. (2010) may
be a reflection of the severity of asthma exacerbations requiring medical attention and
people proceeding directly to a hospital for treatment instead of first visiting a physician.
Therefore, the study may not be able to adequately capture associations, and specifically,
any potential seasonal differences.
The meta-analysis conducted by Zheng et al. (2015) provides some additional supporting
evidence for potential seasonal differences in SCh-asthma hospital admission and ED
visit associations. In a combined analysis including both asthma hospital admission and
ED visit studies that reported seasonal results, Zheng et al. (2015) reported slightly larger
associations in the warm [4.8% (95% CI: 2.7, 7.0) for a 10-ppb increase in 24-h avg SO2
concentrations] compared to the cold season [3.2% (95% CI: 0.5, 5.9)], but confidence
intervals did overlap.
Although there is some evidence for larger associations during the summer, studies
conducted by Villeneuve et al. (2007) in Edmonton, AB and Jalaludin et al. (2008) in
Sydney, Australia present conflicting results. As stated above, Villeneuve et al. (2007)
did not find evidence of an association between short-term SO2 exposures and asthma ED
visits, including in the seasonal analysis, while Jalaludin et al. (2008) reported evidence
of larger associations during the cold months (May-October) compared to the warm
months (November-April) (Figure 5-3).
Overall, the results of Samoli etal. (2011). Son et al. (2013). Strickland et al. (2010). and
Bvers et al. (2015) suggest that associations are larger in magnitude during the summer
season, but this conclusion should be viewed with caution because the results of each
study are highly imprecise, as reflected by the wide confidence intervals for each
seasonal result. Additionally, the interpretation of results from these studies is
complicated by the lack of copollutant analyses and by the results from Villeneuve et al.
(2007) and Jalaludin et al. (2008) which do not show evidence of larger associations
during the summer or warm season.
Lag Structure of Associations
When examining associations between air pollution and a specific health outcome, such
as respiratory-related hospital admissions, it is informative to assess whether exposure to
an air pollutant results in an immediate, delayed, or prolonged effect on health. Recent
studies that examine both multiple single- and multiday lags can help provide information
5-67
-------
on whether there is a specific exposure window(s) that contribute to SCh-related asthma
hospital admissions and ED visits.
Son et al. (2013) examined the lag structure of associations for multiple
respiratory-related hospital admissions, including asthma and allergic disease, by
analyzing both single- and multiday lags. Across single-day lags of 0 to 3 days, positive
associations were observed across each lag, but the magnitude of the association varied
across single-day lags for each outcome. For both asthma and allergic disease hospital
admissions, the largest association, in terms of magnitude, for SO2 was observed for each
of the multiday lags examined, with the largest occurring at lag 0-3 days [asthma: 5.3%
(95% CI: -2.4, 13.0); allergic disease: 3.1% (95% CI: -3.7, 10.7) for a 10-ppb increase in
24-h avg SO2 concentrations].
Studies conducted by Samoli etal. (2011) and Jalaludin et al. (2008) report evidence for
the strongest SCh-asthma hospital admission and ED visit associations occurring rather
immediately (lag 0) as well as over the first few days after exposure, average of lags from
0 up to 2 days. Samoli et al. (2011) in the examination of single- and multiday lags for
associations between SO2 and asthma hospital admissions in Athens, Greece found
associations of similar magnitude at lag 0 and a 0-2 day distributed lag, but the
confidence interval around the association from the distributed lag model was wide
(quantitative results not presented). The associations reported for single-day lags of 1 and
2 days were small and close to null. Jalaludin et al. (2008). in a study in Sydney,
Australia, found when examining single-day lags of 0 to 3 days that asthma ED visit
associations were largest for lag 0 [22.0% (95% CI: 9.1, 34.5) for a 10-ppb increase in
24-h avg SO2 concentrations] and 1 day [16.1% (95% CI: 5.1, 26.5)]. This result is
further reflected in the largest SO2 association being observed for the multiday lag of
0-1 days [29.7% (95% CI: 14.7, 46.5)].
Only a limited number of studies have examined the lag structure of associations, and the
results across studies are not fully supported by the rest of the literature base. Villeneuve
et al. (2007). when studying asthma ED visits in seven Canadian cities, examined
single-day lags (0 and 1 day) as well as multiday lags of 0-2 and 0-4 days. The authors
reported no evidence of an association between short-term SO2 exposures and asthma ED
visits at any lag. Additionally, Orazzo et al. (2009) in the study of wheeze ED visits in six
Italian cities, examined multiday lags ranging from 0-1 to 0-6 days. Across the lags
examined, the authors reported evidence of increasing magnitude of the association as the
length of the multiday lag increased, with lag 0-6 days showing the largest association.
5-68
-------
Exposure Assignment
Questions often arise in air pollution epidemiologic studies about the method used to
assign exposure (see Section 3.3.3). Strickland et al. (2011). using ED visit data from
Atlanta, GA, assessed the effect of various exposure assignment approaches on the
relationship between short-term air pollution exposures and asthma ED visits. The
authors used warm season data from Strickland et al. (2010) to examine the relative
influence of different exposure assignment approaches (i.e., central monitor, unweighted
average across available monitors, and population-weighted average) on the magnitude
and direction of associations between SO2 and pediatric asthma ED visits. SO2 exhibited
a relatively low chi-square goodness-of-fit statistic compared with other pollutants, which
the authors attributed to spatial heterogeneity in SO2 concentrations (Section 3.4.2.2).
Strickland et al. (2011) reported that effect estimates per interquartile range (IQR)
increase in SO2 were similar across the metrics; however, based on a standardized
increment (i.e., 20 ppb in the study), the magnitude of the association between SO2 and
pediatric asthma ED visits varied [central monitor 3.0% (95% CI: -0.4, 8.4); unweighted
average 12.8% (95% CI: 2.8, 23.4); population-weighted average 10.9% (95% CI: 0.8,
21.9) for a 40-ppb increase in 1-h max SO2 concentrations at lag 0-2 days]. The
difference in associations observed across the various exposure assignment approaches
when using the standardized increment can be attributed to the value (i.e., a 1-h max SO2
concentration of 20 ppb) not reflecting an increase in SO2 concentrations that is reflective
of the SO2 distribution in Atlanta (e.g., in the study, the standardized increment for 1-h
max SO2 is 20 ppb, but the IQR, which is often used to calculate the relative risk (RR),
differs across the exposure assignment approaches, varying from 9.6 to 13.9 ppb).
Concentration-Response Relationship
To date, few studies have examined the concentration-response (C-R) relationship
between SO2 concentration in ambient air and respiratory morbidity. In recent
epidemiologic studies, Strickland et al. (2010) and Li etal. (2011) examined the shape of
the SC>2-pediatric asthma ED visit relationship using different analytical approaches.
Strickland et al. (2010) examined the C-R relationship by conducting quintile and locally
weighted scatterplot smoothing (LOESS) C-R analyses. In the quintile analysis, SO2
associations were examined in both the warm and cold seasons; however, no associations
were observed for the cold season for any quintile. Focusing on the warm season, the
authors found evidence of an increase in the magnitude of the association for 1-h max
concentrations within the range of 7 to <24.2 ppb, relative to the first quintile (i.e., SO2
concentrations <3.1 ppb). The smallest associations were observed for the 5th quintile,
which represented 1-h max concentrations ranging from 24.2 to <149 ppb; however, this
quintile represented the extreme end of the distribution of SO2 concentrations where data
5-69
-------
density was low. Additionally, the LOESS C-R relationship analysis provides evidence of
a linear relationship between short-term SO2 exposures and asthma ED visits along the
distribution of 1-h max concentrations from the 5th (2.1 ppb) to 95th (21.5 ppb)
percentile (Sacks. 2015) (Figure 5-4). Collectively, these analyses do not provide
evidence of a threshold.
In a study conducted in Detroit, MI, Li et al. (2011) examined whether there is evidence
of a nonlinear C-R relationship for air pollutants and pediatric asthma ED visits.
Associations with SO2 were examined in both a time-series and time-stratified,
case-crossover study design assuming (1) a linear relationship and (2) a nonlinear
relationship starting at a 24-h avg concentration of 8 ppb [i.e., the maximum likelihood
estimate within the 10th to 95th percentile concentration where a change in linearity may
occur (~91 st percentile)]. It is important to note the analysis that assumed a nonlinear
relationship did not assume zero risk below the inflection point. The focus of the analysis
was on identifying whether risk increased above that observed in the linear models at 24-
h avg SO2 concentrations above 8 ppb. In the analyses assuming linearity, the authors
examined single-day lags of 3 and 5 days and multiday lags of 0-2 and 0-4 days.
Positive associations were observed for all lags examined and were relatively consistent
across models, with the strongest association for a 0-4-day lag [time series: 20.5% (95%
CI: 8.9, 33.2); case-crossover: 22.8% (95% CI: 12.6, 33.7) for a 10-ppb increase in 24-h
avg SO2 concentrations]. In the models that assumed a nonlinear relationship, the authors
did not observe evidence of increased risk above ~8 ppb. However, it is important to note
that the data density is low at 24-h avg concentrations greater than 8 ppb, as reflected by
this value representing the ~91st percentile of SO2 concentrations.
Collectively, Strickland et al. (2010) and Li etal. (2011) provide initial evidence of a log-
linear, no-threshold relationship between short-term SO2 concentrations and asthma ED
visits. However, it is important to note that these studies have not fully explored potential
alternatives to linearity when examining the shape of the C-R relationship, which in
combination with the potential measurement error due to lack of characterization of the
spatial and temporal variability in SO2 exposure concentrations, complicates the
interpretation of the S02-asthma ED visit C-R relationship (Sections 3.4.2.2 and 3.4.2.3).
5-70
-------
•r?
OJ
to
nd
K
0>
OH
8
8
o
Sulfur Dioxide Warm Season
10 15
Concentration (ppb)
20
Note: solid line = smoothed concentration-response estimate. Dashed line = twice-standard error estimates.
Source: Strickland et al. (2010). Reprinted with permission of the American Thoracic Society.
Figure 5-4 Concentration-response for associations between 3-day average
(lag 0-2) sulfur dioxide concentrations and emergency
department visits for pediatric asthma at the 5th to 95th percentile
of sulfur dioxide concentrations in the Atlanta, GA area.
Sulfur Dioxide within the Muitipoiiutant Mixture
An important question during the review of any criteria air pollutant, is whether the
pollutant has an independent effect on human health. Ambient exposures to criteria air
pollutants are in the form of mixtures, however, making this question difficult to answer
in epidemiologic studies, especially when the pollutant of interest is highly correlated
with other pollutants in the mixture. Epidemiologic studies traditionally try to identify the
independent effect of a criteria air pollutant through the use of copollutant models, but
these methods do not consider the broader air pollution mixture. Recent studies
conducted by Winquist et al. (2014) and Pearce et al. (2015) using pediatric asthma ED
visits data from Atlanta assessed whether specific mixtures are more strongly associated
with health effects compared to others. Although the primary objective of these types of
studies is not to directly assess the independent effects of a pollutant, they can inform the
5-71
-------
understanding of the role of SO2 in the air pollution mixture (e.g., contributing to an
additive or synergistic effect).
Winquist et al. (2014) examined multipollutant mixtures by focusing on the joint effect
(i.e., the combined effect of multiple pollutants) of pollutants often associated with
specific air pollution sources. Associations between short-term SO2 exposures and
pediatric asthma ED visits (i.e., ages 5-17 years) were examined in single-pollutant
models and also in a multipollutant context in joint models for pollutant combinations
representative of irritant gases (i.e., O3, NO2, and SO2), power plants (i.e., SO2 and
SO42 ), and NAAQS pollutants (i.e., O3, CO, NO2, SO2, and PM2 5). It is important to note
that the pollutant combination analyses attempt to address a different question (i.e., what
is the risk associated with exposure to a combination of pollutants?) than a traditional
copollutant analysis, which focuses on identifying the independent effect of a pollutant.
Using the model detailed in Strickland et al. (2010). the authors examined the
relationship between each combination and pediatric asthma ED visits using a Poisson
model in the context of a time-referent case-crossover analysis. The authors reported
results for an IQR increase for lag 0-2 days in single-pollutant analyses as well as three
types of joint effect models [i.e., no interaction terms (primary), first-order multiplicative
interactions between pollutants (interactions), and nonlinear pollutant terms (nonlinear)]
(Figure 5-5).
In single-pollutant analyses, SO2 associations were smaller in magnitude compared to the
other pollutants that comprised each pollutant combination, but the confidence interval
surrounding each SO2 estimate was relatively narrow. Across pollutant combinations that
contained SO2, joint effect models reported consistent positive associations with pediatric
asthma ED visits in the warm season. Additionally, for each pollutant combination the
association observed was larger in magnitude than any single-pollutant association,
including SO2, but not equivalent to the sum of each individual pollutant association for a
specific combination. In the warm season analyses, associations across the different joint
effect models were relatively similar. Overall, the results during the cold season were
more variable.
5-72
-------
1.35
1.30
1.25
1.20
O 1.15
'5
ro
1.10
a;
4->
ro
0C 1.05
1.00
0.95
0.90
0.85
T
SH
11
»I
i»
Oxidant
Gases
Combination
"I c ¦ s ¦
Mil
U fc o
O ** W
ij g «"
11
S*
8 8 J
«l
Secondary
Combination
SM K
Traffic
Combination
sec *
Power Plant
Combination
S I
5 !
i *
Criteria Pollutants
Combination
3 J
i s 1
5 f
M
! a r
¦ ! ]
I s t
i*
Oxidant
Gases
Combination
Secondary
Combination
Warm Season
SPC X
Traffic
Combination
Cold Season
§ s i 11
i *
*
Power Plant
Combination
O w fj k» f
* s SI f
SP€ *
Criteria Pollutants
Combination
CO = carbon monoxide; EC = elemental carbon; JE = joint model estimate; N02 = nitrogen dioxide; 03 = ozone; PM25 = particulate matter with a nominal aerodynamic diameter less
than or equal to 2.5 pm; S02 = sulfur dioxide; S04 = sulfate; SPE = single-pollutant model estimate.
Note: Interquartile range for 1-h max S02 concentrations = 10.51 ppb.
Source: fWinouist et aL 2014V Reprinted with permission of Wolters Kluwer Health.
Figure 5-5 Rate ratio and 95% confidence intervals for single-pollutant and joint effect models for each
pollutant combination in warm and cold season analyses for an interquartile range increase in
each pollutant at lag 0-2 days.
5-73
-------
Pearce et al. (2015) took a different approach to examining multipollutant mixtures by
using an unsupervised learning tool, the self-organizing map (SOM), which is similar to
cluster analysis. Using air pollution concentrations for 10 pollutants from a single
monitor, the authors identified nine distinct day types representative of air quality in
Atlanta during the study period. These unique days were then used as indicator variables
to examine associations with pediatric asthma ED visits using the same statistical
approach as Strickland et al. (2010) and Winquist et al. (2014). Across the nine SOMs,
some pollutant combinations represented days consisting of high single pollutant
extremes, which included a day with high 1-h max SO2 concentrations (i.e., mean
concentration of 48.8 ppb and concentrations ranging from 8.5-23.7 ppb for all other
SOMs). In analyses of all SOMs focusing on lag 1, the strongest associations were
observed for days representing above average concentrations for all pollutants, and for
days representing a collection of primary (i.e., CO, NO2, NOx, EC, and OC) or secondary
pollutants (i.e., O3, ammonium ion (NFL/), and SO4) (Figure 5-6). Additional evidence of
associations with pediatric asthma ED visits was observed for days with single pollutant
extremes, including days with high SO2 concentrations and generally lower
concentrations for all other pollutants (Figure 5-6). Interestingly, when comparing SOMs
results with single-pollutant results in sensitivity analyses, the authors reported a null
association with SO2 at lag 1. This result differs from that observed in Strickland et al.
(2010) and Winquist et al. (2014). but the difference could be due to the fact that Pearce
et al. (2015) focused only on lag 1 because they were examining distinct pollution
profiles that often do not occur on multiple days in a row. In contrast, Strickland et al.
(2010) and Winquist et al. (2014) examined associations over a multiday average of
0-2 days. Additionally, the difference between the SOM and single-pollutant SO2 result
could be because the SOM with high SO2 concentrations was better able to capture the
immediate respiratory response due to higher peak concentrations, which would be
consistent with the effects observed in controlled human exposure and animal
toxicological studies.
Although the single-pollutant results of Winquist et al. (2014) and Pearce et al. (2015)
differ due to the lags examined, the studies contribute to evidence that SO2 alone and in
combination with other pollutants is associated with asthma ED visits. The studies also
highlight the difficulty in separating out the independent effect of a pollutant that is part
of a mixture because multiple pollutants are often highly correlated.
5-74
-------
Oi
cd
p4
-------
were largest in magnitude for multiday lags that encompassed the first few days after
exposure (i.e., average of 0-2 and 0-3 day lags). This evidence generally supports the
timing of SO2 effects observed in the controlled human exposure and animal
toxicological studies (Section 5.2.1.2). The examination of potential copollutant
confounding was rather limited in the body of studies that focused on asthma hospital
admissions and ED visits (Samoli etal.. 2011; Jalaludin et al.. 2008). Across studies, SO2
was found to be low (r < 0.4) to moderately (r = 0.4-0.7) correlated with other pollutants
examined. Evidence from these studies is consistent with those studies evaluated in the
2008 SOx ISA and adds to the body of evidence indicating that SCh-asthma hospital
admission and ED visit associations remain positive, although in some instances
attenuated in magnitude, in copollutant models.
A number of recent studies also examined whether there was evidence that the
association between short-term SO2 exposures and asthma hospital admissions and ED
visits was modified by season or some other individual- or population-level factor
(Chapter 6). An examination of seasonal differences in S02-asthma hospital admission
and ED visit associations provide some evidence of SO2 effects being larger in magnitude
in the summer or warm season, but the lack of this pattern across all studies that
conducted seasonal analyses suggests that seasonal associations may vary by geographic
location. Studies of individual- and population-level factors provide some evidence of
differences in associations by lifestage, for children and older adults, and more limited
evidence for differences by sex (Chapter 6).
Additionally, some recent studies examined various study design issues, including model
specification and exposure assignment. An examination of model specification, as
detailed in Section 5.2.1.6. indicates that the relationship between short-term SO2
exposures and respiratory-related hospital admissions, including those for asthma and
allergic disease, are sensitive to using less than 7 degrees of freedom (df) per year to
account for temporal trends, but robust to alternative lags and df ranging from 3 to 6 for
weather covariates (Son et al.. 2013). The results of Son et al. (2013) are supported by the
sensitivity analyses examining model specification conducted by Alhanti et al. (2016) for
asthma ED visits where the results were relatively consistent when the number of df for
temporal trends was increased and alternative covariates for weather used. An
examination of various exposure assignment approaches, including single fixed-site,
average of multiple monitors, and population-weighted average, suggests that each
approach may influence the magnitude, but not direction, of the SCh-asthma ED visit risk
estimate (Strickland et al.. 2011).
Finally, recent studies examined whether the shape of the S02-asthma ED visits C-R
relationship is linear or provides evidence of a threshold. These studies provide initial
5-76
-------
evidence of a log-linear, no-threshold relationship between short-term SO2 exposures and
asthma ED visits (Li et al.. 2011; Strickland et al.. 2010). but a thorough empirical
exploration of alternatives to linearity has not been conducted. An examination of
seasonal differences in SCh-asthma hospital admission and ED visit associations provide
some evidence of SO2 effects being larger in magnitude in the summer or warm season,
but the lack of this pattern across all studies that conducted seasonal analyses suggests
that seasonal associations may vary by geographic location.
Subclinical Effects Underlying Asthma Exacerbation: Pulmonary
Inflammation and Oxidative Stress
Pulmonary inflammation is a key subclinical effect in the pathogenesis of asthma. It
consists of both acute and chronic responses and involves the orchestrated interplay of
the respiratory epithelium and both the innate and adaptive immune system. The
immunohistopathologic features of chronic inflammation involve infiltration of airway
epithelium by inflammatory cells such as eosinophils, lymphocytes, mast cells, and
macrophages and the release of inflammatory mediators such as cytokines and
leukotrienes. Oxidative stress is also relevant to asthma exacerbation. Many transcription
factors regulating the expression of pro-inflammatory cytokines are redox sensitive, and
inflammatory cells that respond to cytokine signaling can generate reactive oxygen
species leading to oxidative stress.
This section characterizes the evidence on SO2 exposure effects on pulmonary
inflammation and oxidative stress in humans with asthma and in animal models of
allergic airway disease (see Section 5.2.1.7 for healthy humans and animal models). The
2008 SOx ISA (U.S. EPA. 2008d') concluded that evidence from the limited number of
controlled human exposure, epidemiologic, and animal toxicological studies was
insufficient to determine that exposure to SO2 at current ambient concentrations was
associated with inflammation in the airway. However, several studies provided evidence
for subclinical effects related to allergic inflammation. There are no recent controlled
human exposure studies, but there is additional investigation in epidemiologic and animal
toxicological studies. Epidemiologic results are inconsistent for pulmonary inflammation
and oxidative stress, including those for SO2 measured at or near children's schools.
However, recent findings in rats link short-term SO2 exposure to allergic inflammation.
Controlled Human Exposure Studies
Pulmonary inflammation following 5-10-minute exposure to SO2 was discussed in the
previous ISA; no new studies were available for review. Briefly, Tunnicliffe et al. (2003)
measured levels of exhaled nitric oxide (eNO), an indirect marker for pulmonary
inflammation, in individuals with asthma before and after a 1-hour exposure to 0.2 ppm
5-77
-------
SO2 under resting conditions. Nasal lavage fluid levels of the antioxidants, ascorbic and
uric acid, were also measured pre- and post-exposure. No statistically significant
differences were observed between pre- and post-exposure for any of these indicators.
Because subjects were exposed at rest and exposed to low concentrations, it is unlikely
that enough SO2 reached the airways to cause an effect. Gong et al. (2001) evaluated the
response of individuals with asthma to 0.75 ppm SO2 during exercise. In addition to
changes in lung function and symptoms, there was a statistically significant increase in
eosinophil count in induced sputum 2 hours after a 10-minute exposure. This response
was significantly dampened by pretreatment with a leukotriene receptor antagonist. These
results provided some evidence that SO2 elicits an inflammatory response in the airways
of individuals with asthma that extends beyond the immediate bronchoconstriction
response typically associated with SO2 exposure. Additionally, this study provides further
evidence that the bronchoconstriction response is only partially due to neural reflexes and
that inflammatory mediators play an important role (Section 4.3.1).
Epidemiologic Studies
Recent epidemiologic evidence is inconsistent for associations of short-term increases in
ambient SO2 concentration with pulmonary inflammation and oxidative stress in adults
and children with asthma (Table 5-10). Outcomes were assessed at varying frequency:
daily, weekly, or seasonally. All studies examined eNO. Higher eNO has been linked to
higher eosinophil counts (Brodv et al.. 2013) as well as prevalence and exacerbation of
asthma (Soto-Ramos et al.. 2013; Carraro et al.. 2007; Jones et al.. 2001; Kharitonov and
Barnes. 2000). An SCh-associated increase in eNO was observed in a population of adults
with asthma with high prevalence of atopy (90%) (Maestrelli et al.. 2011) (Table 5-10).
although associations with lung function or asthma control score were not observed. The
results of Maestrelli et al. (2011) are also supported by findings that allergic
inflammation in rats persists 24 hours after SO2 exposures repeated over many days. The
multicity U.S. asthma medication trial observed imprecise associations for eNO with
wide 95% CIs in the ICS, beta-agonist, and placebo groups (Qian et al.. 2009a). Both
studies of adults with asthma estimated SO2 exposure from fixed-site monitors. Neither
indicated whether the measurements adequately represented the spatiotemporal
variability in SO2 concentrations in the study area, and the U.S. study averaged
concentrations from monitors within 32 km of each subject's ZIP code centroid.
Two recent studies measured SO2 at or 0.65 km from children's schools (Greenwald et
al.. 2013; Lin et al.. 2011b). which may better represent some component of subjects'
exposure. Results are inconsistent. Percent changes in eNO were 31 (95% CI: -24, 119)
per 10-ppb increase in SO2 measured at a school in El Paso, TX (Greenwald et al.. 2013)
and 5.5 (95% CI: 2.7, 8.3) per 10-ppb increase in SO2 measured near a school in Beijing,
5-78
-------
China before and after the 2008 Olympics (Lin et al.. 201 lb). Among children with
asthma not using ICS in Windsor, ON, SO2 concentrations at a monitor within 10 km of
homes were not associated with eNO but were associated with markers of oxidative stress
in exhaled breath condensate (EBC) (Liu et al.. 2009b). The school-based studies differed
in lags examined, and an association was observed with lag 0 SO2 (Lin et al.. 2011b) but
not lag 0-3 avg SO2 (Greenwald et al.. 2013). For SO2 measured at fixed-site monitors,
associations were observed with both lag 0 and lag 0-2 avg concentrations (Liu et al..
2009b). Prevalence of atopy was not reported for the study populations of children.
Findings are inconsistent for SO2 associations with pulmonary inflammation and
oxidative stress in children and adults with asthma. Copollutant confounding is an
additional uncertainty. Associations were observed with PM2 5, BC, CO, O3, and NO2
(Lin etal.. 2011b; Maestrelli et al.. 2011; Liu et al.. 2009b). Only Liu et al. (2009b)
reported S02-copollutant correlations, indicating the potential for confounding with PM2 5
(r = 0.56), less so with NO2 (r = 0.18), and likely not with O3 (r = -0.02). Maestrelli et al.
(2011) did not examine copollutant models, and results in children with asthma are
conflicting. For pollutants measured 0.65 km from school, SO2 associations with eNO
persisted but decreased with adjustment for PM2 5 or BC (Lin etal.. 2011b). Based on
pollutants measured up to 10 km from home, the SO2 association with oxidative stress
decreased with adjustment for NO2 and did not persist with adjustment for PM2 5 (Liu et
al.. 2009b) (Table 5-10).
Animal Toxicological Studies
The 2008 SOx ISA (U.S. EPA. 2008d) discussed several studies that investigated the
effects of exposure to SO2 on inflammatory responses. While one study failed to
demonstrate inflammation following a single subacute exposure to 1 ppm SO2 (U.S.
EPA. 2008d). other studies found that repeated SO2 exposure enhanced the development
of an allergic phenotype and altered physiologic responses in animal models of allergic
airway disease. These studies demonstrating effects of repeated SO2 exposures in models
of allergic airway disease are listed in Table 5-11 and described here. In addition, other
studies involving repeated SO2 exposures in naive rats, including studies that demonstrate
increased sensitivity to allergens, have been conducted and are described below in
Section 5.2.1.7
5-79
-------
Table 5-10 Epidemiologic studies of pulmonary inflammation and oxidative stress in populations with asthma
published since the 2008 ISA for Sulfur Oxides.
SO2 Averaging
Study Population and SO2 Exposure Time and Effect Estimate (95% CI) SO2 Copollutant Model Results and
Methodological Details Estimates (ppb) Lag Day Single-Pollutant Modela Correlations
Adults with asthma
tQian et al. (2009b)
Boston, MA; New York City, NY;
Philadelphia, PA; Madison, Wl;
Denver, CO; San Francisco, CA;
1997-1999
N = 119, ages 12-65 yr. 100%
persistent asthma. 1/3 ICS use,
1/3 beta-agonist use, 1/3 placebo use.
Examined every 2-4 wk for 16 wk.
Recruited from clinics as part of an
asthma medication trial. Multiple
comparisons—many pollutants, lags,
medication use analyzed.
Monitors averaged
within 32 km of subject
ZIP code centroid.
Mean (SD): 5.3 (4.4)
75th percentile: 7.6
Max: 27
24-h avg Change in eNO (ppb)
0 All subjects: 0.09 (-0.07, 0.25)
ICS: 0.17 (-0.11, 0.44)
Beta-agonist: 0.04 (-0.18, 0.27)
0-3 avg All subjects: 0.07 (-0.12, 0.26)
ICS: 0.15 (-0.13, 0.43)
Beta-agonist: 0.10 (-0.19, 0.38)
Remains null with: PM10, NO2, or O3 (all
subjects, lag 0)
r= 0.58 NO2, NR PM10.
tMaestrelli et al. (2011)
Padua, Italy, 2004-2005
N = 32, mean (SD) age 40 (7.5) yr.
81% persistent asthma. 69% ICS use.
90% atopy.
Six measures over 2 yr. Recruited
from database of beta-agonist users
(>6 times per yr for 3 yr).
Two monitors in city 24-h avg
Medians across 0
seasons: 0.87-2.7
75th percentiles across
seasons: 1.3-4.1
Change in eNO (ppb)
All subjects: 55 (-2.3, 113)
Nonsmokers: 82 (3.1, 161)
Change in EBC pH
Decrease = more inflammation
All subjects: 0.46 (-0.20, 1.1)
Nonsmokers: 0.18 (-0.34, 0.69)
n =22
No copollutant model
No association with personal or fixed-site
PM2.5 or PM10.
Copollutant correlations NR.
5-80
-------
Table 5-10 (Continued): Epidemiologic studies of pulmonary inflammation and oxidative stress in populations
with asthma published since the 2008 ISA for Sulfur Oxides.
Study Population and
Methodological Details
SO2 Exposure
Estimates (ppb)
SO2 Averaging
Time and
Lag Day
Effect Estimate (95% CI)
Single-Pollutant Modela
SO2 Copollutant Model Results and
Correlations
Children with asthma
tGreenwald et al. (2013) Monitor at school 24-h avg
El Paso, TX, Mar-Jun 2010 A: residential area 0-3 avg
N = 38, mean age 10 yr. 47% daily B: 91 m from major
asthma medication use. road
Weekly measures for 13 wk. Mean (SD): 1.2 (0.44)
Recruited from schools. and 0.84 (0.54)
Upper percentiles NR.
Percent change in eNO No copollutant model
A: -59 (-89, 36) No association with PM2.5.
B: 31 (-24, 119) Pearson r= -0.14 BC, -0.22 NO2, -0.07
BTEX, 0.14 cleaning product VOCs.
tLinetal. (2011b)
Beijing, China
N = 8, ages 9-12 yr
Daily measures for five 2-wk periods
before and after Olympics.
Recruitment from school.
Monitor 0.65 km from
school
Means across five
periods before and
after Olympics: 3.7-45
24-h avg Percent change in eNO
0 5.5(2.7,8.3)
1
Persist with: BC or PM2.5
Copollutant correlations NR.
3.4 (1.4, 5.4)
tLiuetal. (2009b). Liu (2013)
Windsor, ON
Oct-Dec 2005
N = 182, ages 9-14 yr. 37% ICS use.
35% beta-agonist use.
Weekly measures for 4 wk. Recruited
from schools. Mean 1.6 and 2.2 h/day
spent outdoors for two study groups.
Two monitors
averaged
99% homes within
10 km of sites
Median: 4.5
95th percentile: 16
24-h avg Percent change
0 eNO: 9.0 (-7.6, 29)
TBARS: 28 (0.46, 63)
8-lsoprostane: 23 (3.9, 44)
0-2 avg eNO: -5.6 (-28, 24)
TBARS: 77 (31, 131)
8-lsoprostane: -0.55 (-28, 38)
TBARS, lag 0-2 avg
Persists with: NO2 or O3
Does not persist with: PM2.5
Spearman r= 0.56 PM2.5, 0.18 NO2,
¦ -0.02 O3.
Avg = average; BC = black carbon; BTEX = benzene, toluene, ethylbenzene, xylene; CI = confidence interval; EBC = exhaled breath condensate; eNO = exhaled nitric oxide;
ICS = inhaled corticosteroid; ISA = Integrated Science Assessment; max = maximum; n = sample size; N = population number; N02 = nitrogen dioxide; NR = not reported;
03 = ozone; PM2.5 = particulate matter with nominal aerodynamic diameter less than or equal to 2.5 |jm; PM10 = particulate matter with nominal aerodynamic diameter less than or
equal to 10 |jm; r = correlation coefficient; SD = standard deviation; S02 = sulfur dioxide; TBARS = thiobarbituric acid reactive substances; VOCs = volatile organic compound.
aEffect estimates are standardized to a 10-ppb increase in 24-h avg S02.
fStudies published since the 2008 ISA for Sulfur Oxides.
5-81
-------
Table 5-11 Study-specific details from animal toxicological studies of
subclinical effects underlying asthma.
Study
Species (Strain); n;
Sex; Lifestage/Age
or Weight
Exposure Details
(Concentration; Duration)
Endpoints Examined
Li etal. (2007)
Rats (Wistar);
n = 6/group; M; age
NR
Sensitization by i.p. injection of 100 mg
ovalbumin followed by booster injection
of 10 mg ovalbumin after 7 days followed
by:
(1) Challenge with 1% ovalbumin aerosol
for 30 min for 7 days beginning at
15 days,
(2) Exposure to 2 ppm SO2 for 1 h/day
for 7 days, or
(3) SO2 exposure followed by ovalbumin
aerosol challenge for 7 days
Endpoints examined 24 h
following the last challenge
BALF—inflammatory cell
counts
Lung—histopathology,
immunohistochemistry
Lung and tracheal
tissue—mRNA and protein
levels of MUC5AC and
ICAM-1
Li etal. (2008)
Rats (Wistar);
n = 6/group; M; age
NR; 180-200 g
Sensitization by i.p. injection of 100 mg
ovalbumin followed by booster injection
of 10 mg ovalbumin after 7 days followed
by:
(1) Challenge with 1% ovalbumin aerosol
for 30 min for 7 days beginning at
15 days,
(2) Exposure to 2 ppm SO2 for 1 h/day
for 7 days, or
(3) SO2 exposure followed by ovalbumin
aerosol challenge for 7 days
Endpoints examined 24 h
following the last challenge
BALF—total and differential
cell counts, EGF
Lung tissue—histopathology
Lung and tracheal
tissue—mRNA levels of
EGF, EGFR, COX-2
Lung tissue—protein levels
of EGFR, COX-2
Xie et al. (2009) Rats (Wistar); Sensitization by i.p. injection of 100 mg
n = 6/group; M; age ovalbumin followed by booster injection
NR of 10 mg ovalbumin after 7 days followed
by:
(1) Challenge with 1% ovalbumin aerosol
for 30 min for 7 days beginning at
15 days,
(2) Exposure to 2 ppm SO2 for 1 h/day
for 7 days, or
(3) SO2 exposure followed by ovalbumin
aerosol challenge for 7 days
Endpoints examined 24 h
following the last challenge
Lung tissue—mRNA levels
of p53, bax, bcl-2
Lung—protein levels of p53,
bax, bcl-2
5-82
-------
Table 5-11 (Continued): Study-specific details from animal toxicological studies
of subclinical effects underlying asthma.
Study
Species (Strain); n;
Sex; Lifestage/Age
or Weight
Exposure Details
(Concentration; Duration)
Endpoints Examined
Li etal. (2014)
Rats (Wistar);
n = 6/group; M; age
NR; 180-200 g
Sensitization by i.p. injection of 100 mg
ovalbumin followed by booster injection
of 10 mg ovalbumin after 7 days followed
by:
(1) Challenge with 1% ovalbumin aerosol
for 30 min for 7 days beginning at
15 days,
(2) Exposure to 2 ppm SO2 for 1 h/day
for 7 days, or
(3) SO2 exposure followed by ovalbumin
aerosol challenge for 7 days
Endpoints examined
BALF—inflammatory cell
counts and cytokines IL-4,
IFN-y, TNF-a, IL-6
Serum—IgE
Lung—histopathology
Lung and tracheal
tissue—mRNA and protein
levels of NFkB, IkBq, IKK(3,
IL-6, IL-4, TNF-a, FOXp3
EMSA NFkB binding activity
BALF = bronchoalveolar lavage fluid; bax = B-cell lymphoma 2-like protein 4; bcl-2 = B-cell lymphoma 2;
COX-2 = cyclooxygenase-2; EGF = epidermal growth factor; EGFR = epidermal growth factor receptor; EMSA = electrophoretic
mobility shift assay; FOXp3 = forkhead box p3; ICAM-1 = intercellular adhesion molecule 1; IFN-y = interferon gamma;
IgE = immunoglobulin E; IKK(B = inhibitor of nuclear factor kappa-B kinase subunit beta; IL-4 = interleukin-4; IL-6 = interleukin-6;
IkBo = nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, alpha; i.p. = intraperitoneal; M = male;
MUC5AC = mucin 5AC glycoprotein; n = sample size; NFkB = nuclear factor kappa-light-chain-enhancer of activated B cells;
NR = not reported; p53 = tumor protein p53; S02 = sulfur dioxide; TNF-a = tumor necrosis factor alpha.
Repeated exposure to SO2 promoted an allergic phenotype when ovalbumin sensitization
and challenge preceded SO2 exposure. As described in the 2008 SOx ISA (U.S. EPA.
2008d), Li et al. (2007) demonstrated that rats, which were first sensitized and challenged
with ovalbumin and subsequently exposed to 2 ppm SO2 for 1 hour/day for 7 days, had
an increased number of inflammatory cells in BALF and an enhanced histopathological
response compared with those treated with ovalbumin or SO2 alone. Similarly,
intercellular adhesion molecule l(ICAM-l), a protein involved in regulating
inflammation, and MUC5AC, a mucin protein, were upregulated in lungs and trachea to a
greater extent in rats treated with ovalbumin and SO2 than in those treated with
ovalbumin or SO2 alone. A follow-up study involving the same exposure regimen (2 ppm
SO2 for 1 hour) in the same allergic animal model (rats sensitized and challenge with
ovalbumin) also found that repeated SO2 exposure enhanced inflammatory and allergic
responses to ovalbumin (Li et al.. 2014). Numbers of eosinophils, lymphocytes, and
macrophages were greater in BALF of SC>2-exposed and ovalbumin-treated animals than
in animals treated only with ovalbumin. In addition, SO2 exposure enhanced upregulation
and activation of NFkB, a transcription factor involved in inflammation and upregulation
of the cytokines interleukin-6 (IL-6) and interleukin-4 (IL-4) in lung tissue in this model
of allergic airway disease. Furthermore, BALF levels of IL-6 and IL-4 were increased to
a greater extent in SC>2-exposed and ovalbumin-treated animals compared with
ovalbumin treatment alone. These results indicate that repeated SO2 exposure enhanced
activation of the nuclear factor kappa-light-chain-enhancer of activated B cells (NFkB)
5-83
-------
inflammatory pathway and upregulation of inflammatory cytokines in ovalbumin-treated
animals. Furthermore, SO2 exposure enhanced the effects of ovalbumin on levels of
interferon gamma (IFN-y, decreased) and IL-4 (increased) in BALF and on IgE levels in
serum (increased). Because levels of IL-4 are indicative of T-derived lymphocyte
helper 2 (Th2) status and levels of IFN-y are indicative of T helper 1 (Thl) status, these
results suggest a shift in Thl/Th2 balance towards Th2 in rats made allergic to
ovalbumin, an effect exacerbated by SO2 exposure. These Th2-related changes are
consistent with the observed increases in serum IgE and BALF eosinophils in
ovalbumin-treated animals, effects which were also enhanced by SO2 exposure.
Alternatively, Th2-related changes may reflect a Type 2 immune response mediated by
Group 2 innate lymphoid cells. Taken together, these results indicate that repeated
exposure to SO2 exacerbated inflammatory and allergic responses in this animal model.
Two other follow-up studies by the same laboratory examined the effects of inhaled SO2
on the asthma-related genes encoding epidermal growth factor (EGF), epidermal growth
factor receptor (EGFR), and cyclooxygenase-2 (COX-2) and on apoptosis-related genes
and proteins in this same model based on sensitization with ovalbumin (Xie et al.. 2009;
Li et al.. 2008). While EGF and EGFR are related to mucus production and airway
remodeling, COX-2 is related to inflammation and apoptosis and may play a role in
regulating airway inflammation. SO2 exposure enhanced the effects of ovalbumin
challenge in this model, resulting in greater increases in mRNA and protein levels of
EGF, EGFR, and COX-2 in the trachea compared with ovalbumin challenge alone. SO2
exposure enhanced other effects of ovalbumin in this model, resulting in a greater decline
in mRNA and protein levels of p53 and B cell lymphoma 2 like protein 4 (bax) and a
greater increase in mRNA and protein levels of B cell lymphoma 2 (bcl-2) in the lungs
compared with ovalbumin challenge alone. The increased ratio of bcl-2:bax, an indicator
of susceptibility to apoptosis, observed following ovalbumin challenge, was similarly
enhanced by SO2. Thus, repeated exposure to SO2 may impact numerous processes that
may be involved in inflammation and/or airway remodeling in allergic airway disease.
Summary of Subclinical Effects Underlying Asthma Exacerbation
Whereas previous evidence was limited and inconsistent, recent evidence from
experimental studies supports a relationship between short-term exposure to SO2 and
allergic responses related to asthma. This includes a finding of eosinophilic inflammation
in individuals with asthma exposed acutely to SO2. In addition, enhanced inflammation
and allergic responses were demonstrated in animals made allergic to ovalbumin and
exposed repeatedly to SO2. Epidemiologic findings are inconsistent overall, including
recent results based on SO2 measured at or near children's schools. However, coherent
with experimental studies, an S02-associated increase in pulmonary inflammation was
5-84
-------
observed in a population of adults with asthma and a high prevalence of atopy.
Copollutant confounding is not addressed in these results, but the evidence from animal
toxicological studies provides some biological plausibility for an effect of SO2 exposure,
particularly because effects in rats were shown to occur with repeated exposures and
24 hours after exposure ended. The evidence for SCh-related allergic inflammation also
supports evidence across disciplines for SO2 effects on asthma symptoms, hospital
admissions, and ED visits, as well as lung function decrements in people with asthma.
Summary of Asthma Exacerbation
The 2008 ISA for Sulfur Oxides did not explicitly draw a conclusion about a relationship
between short-term SO2 exposure and asthma exacerbation but described strong support
from controlled human exposure studies for SC>2-induced lung function decrements and
increases in respiratory symptoms in adults with asthma when ventilation rates were
increased. Such effects in adolescents with asthma are less clear due to a paucity of data,
but effects appear similar to adults. There are no laboratory studies of children exposed to
SO2; however, a number of studies have assessed airway responsiveness of children and
adults exposed to the bronchoconstrictive stimuli methacholine. Based largely on those
studies, school-aged children (-5-11 years of age), particularly boys and perhaps obese
children, might be expected to have greater responses (i.e., larger decrements in lung
function) following exposure to SO2 than adolescents and adults.
In adults with asthma, short-term exposures for 5-10 minutes to 0.2-0.3 ppm SO2
resulted in 5-30% of exercising individuals experiencing moderate or greater decrements
(i.e., >15% decrease in FEVi or >100% increase in sRaw; Table 5-2). Decrements in
FEVi at 0.3 ppm SO2 were statistically significant in responsive individuals (defined as
those having an FEVi decrease of >15% after exposure to 0.6 or 1.0 ppm SO2;
Table 5-3). At concentrations greater than or equal to 0.4 ppm, 20-60% of people with
asthma experienced S02-induced decrements in lung function, which were frequently
accompanied by respiratory symptoms. There is a clear concentration-response
relationship for exposures to SO2 between 0.2 and 1.0 ppm, both in terms of increasing
severity of effect and percentage of asthmatics affected. These concentrations are in the
range of the highest 5-minute ambient SO2 concentrations in some U.S. cities during
2013-2015 (Section 2.5.2.2).
Epidemiologic evidence generally supports SCh-associated increases in asthma hospital
admissions and ED visits, particularly in children (Figure 5-3). and respiratory symptoms
in children with asthma (Figure 5-2; Table 5-8). Epidemiologic evidence is inconsistent
for SO2 associations with lung function decrements in adults and children with asthma
(Tables 5-6 and 5-7). For the limited results from previous epidemiologic and controlled
5-85
-------
human exposure studies on airway responsiveness (i.e., response to methacholine), an
independent effect of SO2 is unclear. Two controlled human exposure studies
demonstrated increased airway responsiveness to subsequent allergen challenge for at
least 48 hours following SO2 exposure in combination with a copollutant (i.e., NO2).
Most epidemiologic studies estimated SO2 exposure from fixed-site monitors without
fully characterizing the spatial and temporal variation in SO2 exposure. A few recent
studies aimed to address the uncertainty in exposure estimates and observed
asthma-related effects in association with S02 measured or modeled at or near schools or
homes. Studies did not statistically correct for exposure measurement error. As in the
2008 ISA for Sulfur Oxides, copollutant confounding is unresolved in the epidemiologic
evidence. Many recent studies continue to indicate that SO2 associations with asthma
hospital admissions and ED visits remain relatively unchanged in magnitude in
copollutant models, but SO2 associations with asthma symptoms and pulmonary
inflammation often did not persist after adjustment for PM2 5, EC/BC, or NO2. The role of
SO2 in ambient multipollutant mixtures is not clearly elucidated. Controlled human
exposure studies show asthma-related effects when SO2 exposure occurs with O3 or NO2,
and limited epidemiologic examination shows associations for multipollutant mixtures
that contain SO2. However, associations for mixtures containing SO2 are similar to or less
than the sum of single-pollutant effect estimates for SO2, CO, NO2, PM10, or PM2 5,
indicating an overlap in associations for copollutants.
Expanded evidence for S02-induced allergic inflammation supports an effect of SO2
exposure on asthma exacerbation. Epidemiologic findings of S02-associated increases in
pulmonary inflammation are inconsistent, but enhanced allergic inflammation and
allergic responses are demonstrated in a previous controlled human exposure study of
adults with asthma and in multiple recent studies from a single laboratory in rats made
allergic to ovalbumin and exposed repeatedly to 2 ppm SO2. These findings provide some
support for the epidemiologic associations for SO2 with decreased lung function as well
as increased airway responsiveness, respiratory symptoms, and pulmonary inflammation
observed in most studies in populations of children and adults with asthma and a high
prevalence of atopy, or with AHR and elevated IgE. However, it is not clear whether risk
of S02-related respiratory effects differs in individuals with atopic asthma. Differential
effects by asthma severity or other asthma phenotypes are not well characterized.
Much of the epidemiologic evidence for S02-associated asthma exacerbation is for 24-h
avg SO2 concentrations. Although 24-h avg and 1-h max SO2 concentrations are
correlated at the same monitor, it is not clear whether this correlation applies across a
community. Some recent studies add evidence of association for asthma symptoms and
ED visits with increases in 1-h max SO2 concentrations, including SO2 measured at
schools. For lung function decrements, pulmonary inflammation, and asthma hospital
5-86
-------
admission and ED visit studies, several results indicate associations for 3- or 4-day avg
SO2 concentrations. The evidence for enhanced allergic inflammation, which is seen after
repeated 2 ppm SO2 exposures and 24 hours after exposure ended, somewhat supports the
biological plausibility of epidemiologic associations with asthma-related outcomes.
Moreover, controlled human exposure studies clearly demonstrate that SO2 exposures of
0.2-0.6 ppm can induce effects related to asthma exacerbation.
5.2.1.3 Allergy Exacerbation
The evidence described in the preceding section for SO2 and allergen coexposure
enhancing inflammation in rodent models of allergic airway disease indicates that SO2
exposure may increase the sensitivity of people with allergic asthma to an allergen. This
evidence also suggests the potential for SO2 exposure to affect respiratory responses in
people with allergy but not asthma. The 2008 ISA for Sulfur Oxides did not make distinct
statements about a relationship with SO2 exposure, but relevant epidemiologic studies
had inconsistent findings. Recent epidemiologic evidence is also uncertain, including that
for school SO2 measurements.
Lung Function in Populations with Allergy
Previous epidemiologic studies examining children or adults with high serum IgE levels,
but without AHR, did not find associations between short-term increases in ambient SO2
concentration and decreases in lung function (Boezen et al.. 2005; Boezen et al.. 1999).
The same studies observed associations for groups with high serum IgE and AHR.
Previous findings were based on 24-h avg SO2 measured at a single site in each city. The
recent study improves on previous studies, measuring SO2 at children's schools (Corrcia-
Deur et al.. 2012V Also, the 2-h avg metric used in this study is more comparable to the
exposure durations examined in experimental studies. In this group of children with
allergy in Sao Paolo, Brazil, SO2 was not associated with PEF [-0.82% (95% CI: -1.9,
0.31) per 10-ppb increase in 2-h avg SO2]. Results were similar for allergy defined by
high serum IgE levels alone like previous studies and by multiple criteria (i.e., high IgE
levels, positive skin prick test, and high blood eosinophil levels). There was evidence for
an association among all children (with and without allergy), but that was attenuated in
copollutant models with PM10, NO2, or CO. Correlations with SO2 were not reported.
5-87
-------
Respiratory Symptoms and Physician Visits in Populations with Allergy
Limited to epidemiologic studies, evidence for an association between short-term SO2
exposure and allergy symptoms is inconsistent. Nonspecific upper and lower respiratory
symptoms were examined in children and adults with high IgE levels but without AHR,
and there were no associations with SO2 exposure (Boezen et al.. 2005; Boezen et al..
1999). For symptoms specific to allergy, Villeneuve et al. (2006b) observed an
SCh-associated increase in physician visits for allergic rhinitis in older adults. Recent
findings for allergic rhinitis or eczema in children are mixed. However, inference about
an SO2 effect is weak both for results indicating an association (Kim et al.. 2016a) and
results not indicating an association (Annesi-Maesano et al.. 2012; Linares et al.. 2010).
Limitations include cross-sectional design (Annesi-Maesano et al.. 2012; Linares et al..
2010). analysis of a multipollutant model with NO2, O3, PM10, and pollen (Kim et al..
2016a; Annesi-Maesano et al.. 2012). lack of consideration of confounding by
meteorological factors (Kim et al.. 2016a). or inclusion of children with and without
allergy in analysis of eczema (Linares et al.. 2010). For results supporting a relationship
with allergy symptoms, associations were observed with same-day (lag 0) 24-h avg SO2
concentrations. These concentrations were from a single monitor in the city, and
information was not reported on the extent to which the measurements represented the
spatiotemporal variability in SO2 concentrations in the study area. Associations were
observed with copollutants such as NO2, PM10, and BS, although these results were
inconsistent as well (Villeneuve et al.. 2006b; Boezen et al.. 2005; Boezen et al.. 1999).
Correlations with SO2 concentrations were not reported, and copollutant models were not
analyzed. Thus, the extent to which the supporting findings may indicate an independent
association for SO2 is unclear.
Subclinical Effects Underlying Allergy Exacerbation
In addition to the animal toxicological evidence for SCh-enhanced allergic inflammation,
a previous epidemiologic study of a population of children in which 8% had asthma and
18% had atopy found an S02-associated decrease in blood eosinophil number, which was
presumed to reflect increased recruitment to the airways (Sovseth et al.. 1995). Exposure
assessment from a monitor 2 km from most subjects' homes is an uncertainty, as is
confounding by PM. The study was conducted in a European city with an aluminum
smelter that emitted SO2 and PM, and PM was not examined for association with
eosinophils.
5-88
-------
Summary of Allergy Exacerbation
Epidemiology studies found little evidence of a relationship between short-term exposure
to SO2 and lung function, respiratory symptoms, or physician visits in populations with
allergy. Animal toxicological studies reported that SO2 exposure enhanced allergic
inflammation (Section 5.2.1.2).
5.2.1.4 Chronic Obstructive Pulmonary Disease Exacerbation
COPD is a lung disease characterized by destruction of alveolar tissue, airway
remodeling, and airflow limitation. Reduced airflow can decrease lung function, and
clinical symptoms demonstrating exacerbation of COPD include cough, dyspnea, sputum
production, and shortness of breath. Severe exacerbation can lead to hospital admissions
or ED visits. This spectrum of outcomes has been evaluated in relation to short-term SO2
exposure, and evidence across outcomes and disciplines is inconsistent. This applies to
the small body of studies available for the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d)
as well as the few available recent studies. Recent findings come from epidemiologic
studies, and most are for hospital admissions and ED visits.
Lung Function and Respiratory Symptoms
Evidence from a controlled human exposure study and epidemiologic studies does not
support an effect of SO2 exposure on lung function in adults with COPD. Recent
epidemiologic studies add information on respiratory symptoms and mostly do not
indicate an association with ambient SO2 concentrations.
Linn et al. (1985a) reported that a 15-minute exposure to 0.4 and 0.8 ppm SO2 had no
effect on lung function in older adults with physician-diagnosed COPD. These adults
were much older than the adults with asthma (Table 5-2) or healthy adults (Table 5-15)
examined in controlled human exposure studies. Also, the level of exercise in adults with
COPD (Ve =18 L/minute) was lower than that of individuals with asthma, which
effectively lowers the SO2 dose delivered to the lungs (Section 4.2.2. Neither the previous
nor recent epidemiologic study observed S02-associated decrements in lung function in
adults with COPD (Peacock et al.. 2011; Harre et al.. 1997). Both studies estimated SO2
exposure from a fixed-site monitor(s), and examined 24-h avg concentrations lagged
1 day. Whereas previous results were based on a multipollutant model (with PM10, NO2,
O3), results of the more recent study were based on a single-pollutant model.
Associations were imprecise with wide 95% CIs [e.g., 0.31 L/minute (95% CI: -0.10,
0.72) change in PEF per 10-ppb increase in SO2 and OR 1.01 (95% CI: 0.89, 1.15) for
5-89
-------
PEF decrement greater than 20%] (Peacock et al.. 2011). Mean and 75th percentile SO2
concentrations were 7.5 and 9.3 ppb, respectively. SO2 mostly was not associated with
dyspnea, sputum changes, wheeze/tight chest, or other respiratory symptoms (Wu et al..
2016; Peacock et al.. 2011). Wu et al. (2016) examined a period of higher SO2
concentration (median 17 ppb and 75th percentile 27 ppb) and observed dyspnea to
increase with an increase in 3- to 6-day avg SO2 (OR: 1.88 [95% CI: 1.06, 3.34] per
10-ppb increase in 3-day avg SO2). However, there was a wide range of distance from
subjects to the monitor (1.6-8.8 km), and associations also were observed with
moderately correlated (r = 0.51-0.68) PM25, PM10, and NO2.
Hospital Admissions and Emergency Department Visits
Of the studies evaluated in the 2008 SOx ISA, only one U.S. or Canadian-based study
examined the association between short-term SO2 exposure and COPD hospital
admissions or ED visits (Figure 5-7). Recent studies add to the initial evidence, which
generally indicates no association between short-term SO2 exposures and COPD hospital
admissions and ED visits. Additionally, most studies averaged SO2 concentrations over
multiple monitors and examined 24-h avg exposure metrics, which, may not adequately
capture the spatial and temporal variability in SO2 concentrations (Section 3.4.2). For
each of the studies evaluated in this section, Table 5-12 presents the air quality
characteristics of each city or across all cities, the exposure assignment approach used,
and information on copollutants examined in each COPD hospital admission and ED visit
study. Other recent studies of COPD hospital admissions and ED visits are not the focus
of this evaluation because of various study design issues, as initially detailed in
Section 5.2.1.2. but the full list of these studies, as well as study-specific details, can be
found in Supplemental Table 5S-6 (U.S. EPA. 2017c).
5-90
-------
Study
Location
Age Lag
fQiu et al. (2013) Hong Kong All 0-3 !•¦
I
I
I
f Wong et al. (2009) Hong Kong AH 0-1 ±
I
I
65+ 0-1
I
I
I
I
I
I
I
Peeletal. (2005)a Atlanta, GA AH 0-2 -+•
I
I
f Stieb et al. (2009) 7 Canadian cities All 1 —
I
I
I
f Arbex et al. (2009) Sao Paulo, Brazil 40+ 0-3 DL !
Hos pital Admissions
ED Visits
-10.0 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0
% Increase (95% Confidence Interval)
ED = emergency department; ISA = Integrated Science Assessment,
a = studies that used a 1-h max exposure metric.
Note: f and red text/symbols = recent studies published since the 2008 ISA for Sulfur Oxides; black text/symbols = U.S. and
Canadian studies evaluated in the 2008 ISA for Sulfur Oxides. Corresponding quantitative results are reported in Supplemental
Table 5S-7 (U.S. EPA. 2017c).
Figure 5-7 Percent increase in chronic obstructive pulmonary disease
hospital admissions and emergency department visits from U.S.
and Canadian studies evaluated in the 2008 SOx ISA and recent
studies in all-year analyses for a 10-ppb increase in 24-h avg or
40-ppb increase in 1-h max sulfur dioxide concentrations.
5-91
-------
Table 5-12 Study-specific details and mean and upper percentile concentrations
from chronic obstructive pulmonary disease hospital admission and
emergency department visit studies conducted in the U.S. and
Canada and evaluated in the 2008 SOx ISA and studies published
since the 2008 SOx ISA.
Upper
Mean Percentile
Location Exposure Concentration Concentrations Copollutants
Study Years Assignment Metric ppb ppb Examined
Hospital admissions
tQiu et al. (2013b), Hong Kong, Average of SO2 24-h avg 7.4 NR Correlations (r):
Ko et al. (2007a) China concentrations n,- n 17^
(1998-2007) from ' ...
10 monitoring CoP° utaJ*
stations models: PM1°
tWonq et al. (2009)
Hong Kong,
China
(1996-2002)
Average of SO2
concentrations
from eight
monitoring
stations
24-h avg
6.8
75th: 8.4
Max: 41.8
Correlations (r):
NR
Copollutant
models: none
ED visits
Peel et al. (2005)
Atlanta, GA
(1993-2000)
Average of SO2 1-h max
concentrations
across monitors
for several
monitoring
networks
16.5
90th: 39.0
Correlations (r):
PM2.5: 0.17
PM10: 0.20
PM10-2.5: 0.21
UFP: 0.24
PM2.5 water
soluble metals:
0.00
PM2.5 sulfate:
0.08
PM2.5 acidity:
-0.03
PM2.5 OC: 0.18
PM2.5 EC: 0.20
Oxygenated
HCs: 0.14
Os: 0.19
CO: 0.26
NO2: 0.34
Copollutant
models: none
5-92
-------
Table 5-12 (Continued): Study specific details and mean and upper percentile
concentrations from chronic obstructive pulmonary
disease hospital admission and emergency department
visit studies conducted in the U.S. and Canada and
evaluated in the 2008 SOx ISA and studies published
since the 2008 SOx ISA.
Study
Location
Years
Exposure
Assignment
Metric
Mean
Concentration
PPb
Upper
Percentile
Concentrations
PPb
Copollutants
Examined
tStieb et al. (2009)
Seven
Average SO2
24-h avg
2.6-10.0
75th: 3.3-13.4
Correlations (r)
Canadian
concentrations
only reported
cities
across all
by city and
(1992-2003)
monitors in
season.
each city.
Copollutant
Number of SO2
models: none
monitors in
each city
ranged from
1-11.
tArbex et al. (2009)
Sao Paulo,
Brazil
(2001-2003)
Average of SO2
concentrations
across
13 monitoring
stations
24-h avg
5.3
75th: 6.6
Max: 16.4
Correlations (r):
PM10: 0.77
NO2: 0.63
CO: 0.52
Copollutant
models: none
Avg = average; CO = carbon monoxide; EC = elemental carbon; HC = hydrocarbon; ISA = Integrated Science Assessment;
max = maximum; N02 = nitrogen dioxide; NR = not reported; 03 = ozone; OC = organic carbon; PM10 = particulate matter with
nominal aerodynamic diameter less than or equal to 10 |jm; PM25 = particulate matter with nominal aerodynamic diameter less
than or equal to 2.5 |jm; PM10-2 5 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm and greater
than a nominal 2.5 |jm; r= correlation coefficient; S02 = sulfur dioxide; UFP = ultrafine particle,
f = Studies published since the 2008 ISA for Sulfur Oxides.
Hospital Admissions
Of the studies evaluated in the 2008 SOx ISA, relatively few examined the association
between short-term SO2 exposure and COPD hospital admissions, and evidence of an
association was inconsistent across studies. Although several recent studies assessed the
relationship between short-term SO2 exposures and COPD hospital admissions, the
overall body of evidence remains limited.
Wong et al. (2009) in a study that examined the potential modification of the relationship
between air pollution and respiratory-related hospital admissions by influenza, also
focused on cause-specific respiratory hospital admissions, including COPD. When
focusing on the baseline effect of short-term SO2 exposures on COPD hospital
admissions, the authors found limited evidence of an association at lag 0-1 days for a
10-ppb increase in 24-h avg SO2 concentrations in analyses of both all ages [0.8% (95%
CI: -1.5, 3.1)] and individuals overthe age of 65 [0.5% (95% CI: -2.0, 3.0)].
5-93
-------
In an additional study conducted in Hong Kong, Qiu et al. (2013b) focused on whether
there is evidence of modification of the air pollution-COPD hospital admissions
relationship by season and humidity. Compared to Wong et al. (2009). Qiu et al. (2013b)
included 5 additional years of recent data through the year 2007. In single-pollutant
models focusing on the association between short-term SO2 exposures and COPD
hospital admissions, for amultiday lag of 0-3 days, the authors reported a 1.6% increase
(95% CI: 0.1, 3.1) for a 10-ppb increase in 24-h avg SO2 concentrations. The magnitude
of the SO2 association was found to differ between Qiu et al. (2013b) and Wong et al.
(2009). but the reason for the difference remains unclear, considering that similar data
sources were used in each study. It is important to note that neither study conducted
copollutant analyses for the entire study duration nor provided detailed information on
the correlation between the air pollutants examined to help in assessing whether SO2 has
an independent effect on COPD hospital admissions.
Emergency Department Visits
The 2008 SOx ISA identified relatively few studies that examined the association
between short-term SO2 exposure and COPD ED visits, and across studies there was
inconsistent evidence of an association. Although recent studies continued to assess the
relationship between short-term SO2 exposures and COPD ED visits, the overall body of
evidence remains limited.
In the seven Canadian cities study discussed previously, and consistent with the asthma
ED visits results, Stieb et al. (2009) did not find any evidence of associations between
24-h avg SO2 and COPD ED visits at single-day lags of 0 to 2 days. Additionally, there
was no evidence of consistent associations between any pollutant and COPD ED visits at
subdaily time scales (i.e., 3-h avg of ED visits vs. 3-h avg pollutant concentrations).
Arbex et al. (2009) also examined the association between COPD and several ambient air
pollutants, including SO2, in a single-city study conducted in Sao Paulo, Brazil for
individuals over the age of 40 years. The authors examined associations between
short-term SO2 exposures and COPD ED visits at single-day lags (0 to 6 days) and in a
polynomial distributed lag model (0-6 days). The authors found evidence that the
magnitude of the association was larger at multiday lags compared to single-day lags,
with the lag of 0-3 days from the distributed lag model [49.4% (95% CI: 4.1, 113.7) for a
10-ppb increase in 24-h avg SO2 concentrations] most representative of the pattern of
associations across single-day lags. Although the 0-6-day distributed lag model had the
largest risk estimate, it was not supported by the single-day lag results that showed the
strongest associations at lags of 0 and 1 day. It is important to note that Arbex et al.
(2009) did not conduct copollutant analyses, but unlike correlations with SO2 observed in
other locations, SO2 was highly correlated with PM10 (r = 0.77) and moderately
5-94
-------
correlated with NO2 (r = 0.63) and CO (r = 0.52) in this study. The results of Arbex et al.
(2009) provide evidence of a potentially prolonged SO2 effect on COPD ED visits;
however, the results should be viewed with caution because effect estimates are not
precise (i.e., wide confidence intervals), time series is short, and copollutant confounding
is possible.
Seasonal Analyses
Traditionally, epidemiologic studies have examined potential seasonal differences in
associations by stratifying by season. In the study of air pollution and COPD hospital
admissions in Hong Kong, Oiu et al. (2013b) examined potential seasonal differences in
associations by this traditional approach but also examined whether the combination of
season and humidity modifies the air pollution-health effect association. In seasonal
analyses, the authors found a stronger association at lag 0-3 for a 10-ppb increase in 24-h
avg SO2 concentrations during the cool season (November-April) [2.7% (95% CI: 0.5,
4.9)] compared to the warm season (May-October) [0.6% (95% CI: -1.1, 2.3)]. Qiu et al.
(2013b) then examined whether the seasonal differences in associations observed were
due to low humidity days (i.e., relative humidity <80%) or high humidity days
(i.e., relative humidity >80%) by examining the interaction between the various
combinations of season and humidity. When focusing on the combined effect of season
and humidity, SO2 concentrations were found to be highest on days with low humidity in
both seasons. In the warm season, there was no evidence of an association regardless of
whether the interaction between season and low or high humidity days were examined. In
the cold season, at lag 0-3 for a 10-ppb increase in 24-h avg SO2 concentrations, Qiu et
al. (2013b) reported the strongest association during days with low humidity [5.3% (95%
CI: 2.4, 8.3)] compared to high humidity [0.5% (95% CI: -2.6, 3.7)], suggesting that the
combination of season and humidity plays a role in the relationship between air pollution
and health effects. However, when examining copollutant models with PM10, associations
in all season and humidity combinations were attenuated, with only the association in the
cool season and low humidity combination remaining positive, albeit with large
uncertainty estimates [0.8% (95% CI: -2.1, 3.9); lag 0-3 for a 10-ppb increase in 24-h
avg SO2 concentrations]. The results from Oiu et al. (2013b) are consistent with evidence
from controlled human exposure studies demonstrating that SO2 responses are
exacerbated in colder and dryer conditions (Section 5.2.1.2). However, these studies
focused on lung function changes in people with asthma, and it is unclear how these
results correspond to results from an epidemiologic study of COPD hospital admissions.
Additionally, it is important to note the potential influence of geographic location on the
results from studies that examine the seasonal patterns of associations.
5-95
-------
Lag Structure of Associations
Only a limited number of studies examined the lag structure of associations for
S02-related COPD hospital admissions and ED visits. In the examination of air pollution
and COPD hospital admissions in Hong Kong, Qiu et al. (2013b) conducted analyses to
evaluate associations with SO2 at both single-day and multiday lags of 0-3 days. The
authors found the strongest evidence for an SO2-COPD hospital admission association at
a multiday lag of 0-3 days, with additional evidence of positive associations at single-day
lags of 1 day and 3 days.
Arbex et al. (2009). when examining associations between SO2 exposure and COPD ED
visits in Sao Paulo, Brazil, focused on both single-day lags (0 to 6 days) and a polynomial
distributed lag (0-6 day) model. The authors found evidence that the magnitude of the
association was larger at multiday lags compared to single-day lags, and the magnitude of
the association increased as the number of lag days examined increased, specifically
across lags of 0-1, 0-2, and 0-5 days. However, the 0-5-day distributed lag model
results were not supported by the single-day lag results, which indicated that the effect of
SO2 on COPD ED visits was rather immediate, occurring in the range of lag 0 and 1 days.
Collectively, the results of Qiu et al. (2013b) and Arbex et al. (2009) provide initial
evidence suggesting a potential prolonged effect of SO2 on COPD hospital admissions
and ED visits. However, the collective evidence indicating a potential association
between short-term SO2 exposures and COPD hospital admissions and ED visits remains
relatively small.
Summary of Chronic Obstructive Pulmonary Disease Exacerbation
Across disciplines and outcomes, evidence from previous and recent studies does not
clearly support a relationship between short-term SO2 exposure and COPD exacerbation.
The evidence base is much smaller than that for asthma exacerbation and mostly
comprises epidemiologic studies. Neither the single controlled human exposure study nor
the few epidemiologic studies indicate S02-related lung function changes in adults with
COPD, and recent epidemiologic studies mostly reported no association with an array of
respiratory symptoms, including sputum changes and dyspnea, which are characteristic of
COPD exacerbation. Evidence is similarly inconsistent for association between
short-term increases in ambient SO2 concentration and hospital admissions and ED visits
for COPD (Figure 5-7). Hospital admissions, ED visits, lung function, and symptoms
were examined in relation to 24-h avg SO2 concentrations, but an association was not
observed with 1-h max SO2 either. The supporting evidence is limited largely to an
association of COPD hospital admissions and ED visits with same-day and 4-day avg
SO2 concentrations. All epidemiologic studies estimated SO2 exposure from fixed-site
5-96
-------
monitors. SO2 generally has low to moderate spatial correlations across urban
geographical scales, and the potential error in the exposure estimates in adequately
representing the spatiotemporal variability is uncharacterized in the evidence
(Section 3.4.2.2V The uncertainty in exposure estimates especially applies to 1-h max
SO2. COPD hospital admissions were associated with PM10, NO2, and O3. PM10 was
highly correlated with SO2 (r = 0.77) (Arbcx et al.. 2009). or when analyzed in a
copollutant model, attenuated the SO2 association and produced wide 95% CIs (Qiu et al..
2013b). The copollutant model results have unclear implication as a result of uncertainty
in the exposure estimates and the unreported SO2-PM10 correlation. Overall, the studies
do not consistently report positive associations between short-term SO2 exposure and
COPD exacerbation. The consideration of copollutant confounding is limited so the
independent effect of SO2 is unclear.
5.2.1.5 Respiratory Infection
The respiratory tract is protected from exogenous pathogens and particles through various
lung host defense mechanisms that include mucociliary clearance, phagocytosis by
alveolar macrophages, and innate and adaptive immunity. There is a paucity of evidence
related to host defense from animal toxicological experiments using ambient-relevant
concentrations of SO2. Several studies of short-term exposure to SO2 were reported in the
1982 AQCD (U.S. EPA. 1982a) and discussed in the 2008 SOx ISA (U.S. EPA. 2008dV
Findings of short-term studies included some effects of 0.1-1 ppm SO2 on the clearance
of labeled particles. No new animal studies of the effects of SO2 exposure on lung host
defense have been conducted since the previous review. A small number of previous
epidemiologic studies reported S02-associated increases in respiratory infections as
self-reported or indicated by hospital admissions and ED visits. However, many results
were noted as being unreliable because they were based on statistical methods prone to
bias.
Recent contributions to the evidence are limited to epidemiologic studies, and the
evaluation of this evidence focuses on hospital admissions and ED visits. There are recent
studies of self-reported infections, and they inconsistently show associations with
ambient SO2 concentrations, [Supplemental Figure 5S-2 and Table 5S-8 (U.S. EPA.
2017c)]. Results based on school or home SO2 exposure estimates are limited by their
cross-sectional design or examination of nonspecific symptoms such as fever. Other
studies do not provide insight over studies of hospital admissions and ED visits on issues
such as exposure measurement error, copollutant confounding, or potentially relevant
exposure durations and concentrations. Recent studies of respiratory infection hospital
5-97
-------
admissions and ED visits provide some evidence for association with ambient SO2
concentrations. However, copollutant confounding remains an uncertainty.
Hospital Admissions and Emergency Department Visits
The 2008 SOx ISA contained limited evidence of an association between short-term SO2
concentrations and respiratory conditions other than asthma or COPD including
respiratory tract infections and pneumonia. Of these studies, only one study was
conducted in the U.S. or Canada [i.e., (Peel et al.. 2005)1. Recent studies have examined a
variety of outcomes indicative of respiratory infection; however, none have examined the
same respiratory infection outcome. Additionally, most studies averaged SO2
concentrations over multiple monitors and examined 24-h avg exposure metrics, which
may not adequately capture the spatial and temporal variability in SO2 concentrations
(Section 3.4.2V For each of the studies evaluated in this section, Table 5-13 presents the
air quality characteristics of each city, or across all cities, the exposure assignment
approach used, and information on copollutants examined in each respiratory infection
hospital admission and ED visit study. Other recent studies of respiratory infection
hospital admissions and ED visits are not the focus of this evaluation because of various
study design issues, as initially detailed in Section 5.2.1.2. but the full list of these
studies, as well as study specific details, can be found in Supplemental Table 5S-6 (U.S.
EPA. 2017c).
5-98
-------
Table 5-13 Study-specific details and mean and upper percentile concentrations from respiratory infection
hospital admission and emergency department visit studies conducted in the U.S. and Canada and
evaluated in the 2008 SOx ISA and studies published since the 2008 SOx ISA.
Study
Location (Years)
Type of Visit
(ICD 9/10)
Exposure
Assignment
Metric
Mean
Concentration
PPb
Upper Percentile
of
Concentrations
PPb
Copollutants
Examined
Hospital admissions
+HEI (2012)
Mehta et al. (2013)
Ho Chi Minh City,
Vietnam
(2003-2005)
Acute lower
respiratory
infection
(J 13—16, 18, 21)
Average of SO2
concentrations
across nine
monitors
24-h avg
8.2
Max: 30.5
Correlations (r):
Dry season:
PM10: 0.32
Os: 0.19
NO2: 0.29
Rainy season:
PM10: 0.36
Os: 0.65
NO2: 0.01
Copollutant
models: NO2,
PM10, O3
tSeqala et al.
(2008)
Paris, France
(1997-2001)
Bronchiolitis
Average SO2
concentrations
across
30 monitors
24-h avg
4.0
Max: 27.4
Correlations (r):
BS: 0.76
PM10: 0.73
NO2: 0.78
Copollutant
models: none
5-99
-------
Table 5-13 (Continued): Study specific details and mean and upper percentile concentrations from respiratory
infection hospital admission and emergency department visit studies conducted in the
U.S. and Canada and evaluated in the 2008 SOx ISA and studies published since the 2008
SOx ISA.
Study
Location (Years)
Type of Visit
(ICD 9/10)
Exposure
Assignment
Metric
Mean
Concentration
PPb
Upper Percentile
of
Concentrations
PPb
Copollutants
Examined
ED visits
Peel et al. (2005)
Atlanta, GA
(1993-2000)
Pneumonia
(480-486)
Average of SO2
concentrations
from monitors for
several monitoring
networks
1-h max
16.5
90th: 39.0
Correlations (r):
PM2.5: 0.17
PM10: 0.20
PM10-2.5: 0.21
UFP: 0.24
PM2.5 water
soluble metals:
0.00
PM2.5 sulfate: 0.08
PM2.5 acidity:
-0.03
PM2.5 OC: 0.18
PM2.5 EC: 0.20
Oxygenated HCs:
0.14
Os: 0.19
CO: 0.26
NO2: 0.34
Copollutant
models: none
tStieb et al. (2009) Seven Canadian Respiratory Average SO2 24-h avg 2.6-10.0 75th: 3.3-13.4 Correlations (r)
cities (1992-2003) infection concentrations only reported by
(464,466, across all city and season.
480-487) monitors in each Copollutant
city. Number of models: none
SO2 monitors in
each city ranged
from 1-11.
5-100
-------
Table 5-13 (Continued): Study specific details and mean and upper percentile concentrations from respiratory
infection hospital admission and emergency department visit studies conducted in the
U.S. and Canada and evaluated in the 2008 SOx ISA and studies published since the 2008
SOx ISA.
Study
Location (Years)
Type of Visit
(ICD 9/10)
Exposure
Assignment
Metric
Mean
Concentration
PPb
Upper Percentile
of
Concentrations
PPb
Copollutants
Examined
tSeqala et al.
(2008)
Paris, France
(1997-2001)
Bronchiolitis
Average SO2
concentrations
across
30 monitors
24-h avg
4.0
Max: 27.4
Correlations (r):
BS: 0.76
PM10: 0.73
NO2: 0.78
Copollutant
models: none
tZemek et al.
(2010)
Edmonton, AB
(1992-2002)
Otitis media
(382.9)
Average of SO2
concentrations
across three
monitors
24-h avg
All-year: 2.6
Warm (Apr-Sep):
2.1
Cold (Oct-Mar):
3.1
All-year
75th: 3.5
Correlations (r):
NR
Copollutant
models: none
Outpatient and physician visits
tSinclair et al.
(2010)
Atlanta, GA
(1998-2002)
Upper respiratory
infection
Lower respiratory
infection
SO2
concentrations
collected as part
of AIRES at
SEARCH
Jefferson Street
site
1-h max
1998-2000: 19.3
2000-2002: 17.6
1998-2002: 18.3
NR
Correlations (r):
NR
Copollutant
models: none
Avg = average; AIRES = Aerosol Research Inhalation Epidemiology Study; BS = black smoke; CO = carbon monoxide; EC = elemental carbon; HC = hydrocarbon;
ICD = International Classification of Diseases; ISA = Integrated Science Assessment; max = maximum; N02 = nitrogen dioxide; NR = not reported; 03 = ozone; OC = organic carbon;
PM2.5 = in general terms, particulate matter with a nominal aerodynamic diameter less than or equal to 2.5 |jm; PM10 = in general terms, particulate matter with a nominal
aerodynamic diameter less than or equal to 10 |jm; PM10-2 5 = in general terms, particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm and greater than
a nominal 2.5 |jm; r= correlation coefficient; SEARCH = Southeast Aerosol Research Characterization; S02 = sulfur dioxide; UFP = ultrafine particle.
f = studies published since the 2008 ISA for Sulfur Oxides.
5-101
-------
Hospital Admissions
Although recent studies have continued to examine the association between short-term
SO2 exposures and respiratory infection hospital admissions, the overall evidence
remains limited, primarily due to the variety of respiratory infection outcomes examined.
In a study conducted in Ho Chi Minh City, Vietnam Mehtaetal. (2013) and HEI (2012)
examined the association between short-term air pollution exposures and pediatric (ages
28 days-5 years) hospital admissions for acute lower respiratory infections (ALRI,
including bronchiolitis and pneumonia). In a time-stratified, case-crossover analysis
focusing only on the average of a 1-6 day lag, the study authors reported a positive
association, with large uncertainty estimates, between SO2 and ALRI hospital admissions
in the all-year analysis [7.0% (95% CI: -3.0, 19.1) for a 10-ppb increase in 24-h avg SO2
concentrations]. A larger association was observed in the time-series analysis (HEI.
2012) (Figure 5-8). When examining copollutant models with PM10 and O3, SO2
associations increased slightly, with the percent increase ranging from 7.5-8.0%,
respectively. However, in models with NO2, the SO2 association was attenuated, but
remained positive [4.9% (95% CI: -6.0, 17.0) for a 10-ppb increase in 24-h avg SO2
concentrations].
In another study that also examined respiratory infections (i.e., bronchiolitis) in children,
Segala et al. (2008) focused on associations with winter (October-January) air pollution
because that is the season when respiratory syncytial virus (RSV) activity peaks. Segala
et al. (2008) hypothesized that air pollution exposures may increase the risk of respiratory
infections, including bronchiolitis due to RSV. Focusing on children <3 years of age in
Paris, France, the study authors conducted a bidirectional case-crossover analysis along
with a time-series analysis to examine air pollution associations with bronchiolitis
hospital admissions and ED visits (see ED visits section below). Although the authors
specified that the bidirectional case-crossover approach was used to "avoid time-trend
bias," it must be noted that the bidirectional approach has been shown to bias results
(Segala et al.. 2008; Lew et al.. 2001). In the case-crossover analysis, SO2 was associated
with bronchiolitis hospital admissions at lag 0-4 days for a 10-ppb increase in 24-h avg
SO2 concentrations [34.8% (95% CI: 19.5, 47.8)] with a similar risk estimate observed
forthe time-series analysis [31.6% (95% CI: 13.7, 51.2)]. Although a positive association
was observed, the authors did not conduct copollutant analyses. This omission
complicates the interpretation of the results because SO2 was highly correlated with the
other pollutants examined, with correlations ranging from r = 0.73-0.87.
5-102
-------
Study Location Age
|HEI (2012); M ehta et al. (2013) Ho Chi Minh, Vietnam 28 days - 5 years
28 days - 5 years
jSegala et al. (2008) Paris, France < 3
Peel et al. (2005)a
jStieb et al. (2009)
jSegala et al. (2008)
Peel et al. (2005)a
jZemek et al. (2010)
Atlanta, GA
7 Canadian cities
Paris, France
Atlanta, GA
Edmonton, Canada
All
All
<3
All
1-3
Lag
l-6b
l-6c
0-4b
0-4c
0-2
2
0-4b
0-4c
0-2
4
Hospital Admissions
Respiratory Infection
Bronchiolitis
~
ED Visits
Respiratory Infection
Pneumonia
Otitis media
-20.0 -10.0 0.0 10.0 20.0 30.0 40.0
% Increase (95% Confidence Jhterrol)
ED = emergency department; ISA = Integrated Science Assessment,
a = studies that used a 1-h max exposure metric.
Note: f and red text/symbols = recent studies published since the 2008 ISA for Sulfur Oxides; Black text/symbols = U.S. and
Canadian studies evaluated in the 2008 ISA for Sulfur Oxides; circles = all-year results, diamonds = warm season results,
squares = cold season results. Corresponding quantitative results are found in Supplemental Table 5S-9 (U.S. EPA. 2017c).
Figure 5-8 Percent increase in respiratory infection hospital admissions and
emergency department visits from U.S. and Canadian studies
evaluated in the 2008 SOx ISA and recent studies in all-year and
seasonal analyses for a 10-ppb increase in 24-h avg or 40-ppb
increase in 1-h max sulfur dioxide concentrations.
Emergency Department Visits
Similar to respiratory infection hospital admissions, recent studies have examined
respiratory infection ED visits; however, these studies overall have not consistently
examined the same respiratory infection outcomes (Figure 5-8). In their study of seven
Canadian cities, Stieb et al. (2009) examined the association between short-term SO2
exposure and respiratory infection ED visits. The authors reported a positive association
at a 2-day lag [1.2% (95% CI: -2.5, 5.2) for a 10-ppb increase in 24-h avg SO2
concentrations], but this result was uncertain, with no evidence of an association at
single-day lags of 0 and 1 days. However, Segalaet al. (2008). in addition to examining
bronchiolitis hospital admissions, also examined bronchiolitis ED visits. The authors
reported evidence of an association between short-term SO2 exposures and bronchiolitis
ED visits [34.7% (95% CI: 25.5, 44.5); lag 0-4 for a 10-ppb increase in 24-h avg SO2
concentrations]. However, as mentioned previously, the interpretation of these results is
complicated by the lack of copollutant analyses and the high correlation between the
5-103
-------
pollutants examined (r = 0.73 to 0.87), along with the use of a bidirectional
case-crossover approach.
In an additional study conducted in Edmonton, AB, Zemek et al. (2010) examined a new
outcome for SO2, otitis media (i.e., ear infections) ED visits, for ages 1-3 years.
Associations were examined for single-day lags of 0 to 4 days in all-year as well as
seasonal analyses. The authors found no evidence of an association between short-term
SO2 exposures and increases in ED visits for otitis media at any single-day lag in the
all-year analysis.
Physician/Outpatient Visits
In a study conducted in Atlanta, GA as discussed in Section 5.2.1.2. Sinclair et al. (2010)
examined the association between air pollution and respiratory infection (e.g., upper
respiratory infections, lower respiratory infections) outpatient visits from a managed care
organization. As detailed previously, the authors separated the analysis into two time
periods—the first 25 months of the study period (i.e., August 1998-August 2000) and the
second 28 months of the study period (i.e., September 2000-December 2002). A
comparison of the two time periods indicated that risk estimates across outcomes tended
to be larger in the earlier 25-month period compared to the later 28-month period. An
examination of the respiratory infection outcomes found no evidence of an association for
upper respiratory infections at any lag and a positive association for lower respiratory
infections for only lag 0-2.
Multiday Lags
In the case of respiratory infection hospital admission and ED visit studies, none of the
studies evaluated conducted an extensive analysis of the lag structure of associations.
However, Segala et al. (2008) in a study of acute bronchiolitis examined multiday lags of
0-1 and 0-4 days, which does provide some indication of the lag structure of
associations. The authors found relatively similar associations for both multiday lags, but
the association was slightly larger for lag 0-4 days (i.e., 31.6 vs. 34.8%). These initial
results indicate a potential prolonged effect of SO2 that could lead to a respiratory
infection hospital admission or ED visit.
Seasonal Analyses
A few of the recent studies that examined respiratory infection-related hospital
admissions and ED visits also examined whether there was evidence of seasonal
differences in associations. Note that interpreting the results from these studies is
complicated by the different geographic locations as well as the respiratory infection
outcome examined in each study. Mehta et al. (2013) in the study of ALRI hospital
5-104
-------
admissions in Vietnam examined potential seasonal differences in associations by
dividing the year into the dry (November-April) and rainy seasons (May-October).
Within these seasons, SO2 concentrations differed drastically, with mean 24-h avg SO2
concentrations being 10.1 ppb in the dry season and 5.7 ppb in the rainy season. In
seasonal analyses, Mehta et al. (2013) reported that SO2 was consistently associated with
ALRI hospital admissions in the dry season [16.1% (95% CI: 1.2, 33.3) for a 10-ppb
increase in 24-h avg SO2 concentrations, lag 1-6 day avg], with no evidence of an
association in the rainy season. Of the other pollutants that were found to be positively
associated with ALRI hospital admissions during the dry season (i.e., PM10 andN02),
none were associated during the rainy season. In copollutant analyses for the dry season,
SO2 was robust to the inclusion of PM10 and O3 in the model, with the magnitude of the
effect remaining similar, 15.0 and 15.8%, respectively. However, in models with NO2,
the SO2-ALRI hospital admission association was attenuated, but remained positive with
large uncertainty estimates [10.0% (95% CI: -4.6, 26.9) for a 10-ppb increase in 24-h avg
SO2concentrations, lag 1-6 day avg].
Additionally, Zemek et al. (2010) in the study of otitis media ED visits in Alberta,
reported that the magnitude of the association was larger, albeit with wide confidence
intervals, in the warm months (April-September), 9.0% (95% CI: -8.4, 34.2), compared
to the cold months, (October-March), -4.3% (95% CI: -16.30, 9.0) at lag 4 for a 10-ppb
increase in 24-h avg SO2 concentrations.
Summary of Respiratory Infection
Recent evidence, which comes from epidemiologic studies, expands on that presented in
the 2008 ISA for Sulfur Oxides and provides some, but not entirely consistent, support
for an association between ambient SO2 concentrations and respiratory infection.
Whereas cross-sectional studies do not consistently link SO2 exposures estimated for
school or home to respiratory infections self-reported by children [Supplemental
Figure 5S-2 and Table 5S-8 (U.S. EPA. 2017c)!. some evidence points to an association
with hospital admission and ED visits (Figure 5-8). Associations are observed for all
respiratory infections combined and bronchiolitis but not pneumonia or otitis media. The
lack of multiple studies examining the same respiratory infection outcome complicates
the interpretation of the collective body of evidence, specifically because the etiologies of
upper and lower respiratory infections are vastly different.
Most supporting evidence points to associations with 24-h avg SO2 concentrations
averaged over 3 to 7 days, but an association was observed with temporally resolved 1-h
max as well. The relatively small number of studies does not provide a strong basis for
drawing inferences about the lag structure of associations with respiratory infection or
5-105
-------
potential seasonal differences in associations. An examination of potential factors that
could modify the S02-respiratory infection hospital admission or ED visit association
finds differences by SES but inconsistent differences by sex (Chapter 6). Recent studies
continued to rely on fixed-site monitors. SO2 generally has low to moderate spatial
correlations across urban geographical scales, and the uncharacterized spatial and
temporal variability may contribute to some degree of exposure measurement error
(Section 3.4.2.2). Another uncertainty that persists in the recent evidence is copollutant
confounding. Respiratory infection hospital admissions and ED visits were associated
with PM2 5, PM10, BS, and NO2. High SCh-copollutant correlations were observed
(r = 0.73-0.78). Correlations were low in some locations (r = 0.17-0.34) (Table 5-13).
but differences in exposure measurement error among different pollutants may influence
effect estimates, particularly for copollutants with different averaging times. New
information from copollutant models shows an SO2 association that is attenuated and
often imprecise (i.e., wide confidence intervals) with adjustment for NO2, but uncertainty
in the exposure estimates weakens inference about independent associations. Information
to assess the biological plausibility of epidemiologic findings is limited. There is some
evidence in rodents that SO2 exposures of 0.1-1 ppm alter clearance of particles, but
responses to infectious agents have not been examined in relation to ambient-relevant
exposures.
5.2.1.6 Aggregated Respiratory Conditions
In addition to individual respiratory conditions, epidemiologic studies examined
respiratory effects as an aggregate of multiple respiratory conditions (e.g., asthma,
COPD, respiratory infections). Epidemiologic studies examining the association between
short-term SO2 exposures and respiratory-related hospital admissions or ED visits,
including those discussed earlier in this chapter, were not available until after the
completion of the 1986 Supplement to the Second Addendum of the 1982 SOx AQCD
(U.S. EPA. 1994). Therefore, the 2008 SOx ISA (U.S. EPA. 2008d) included the first
thorough evaluation of respiratory morbidity in the form of respiratory-related hospital
admissions and ED visits. Of the studies evaluated, the majority consisted of single-city,
time-series studies that primarily examined all respiratory disease or asthma hospital
admissions or ED visits, with a more limited number of studies examining other
respiratory outcomes, as discussed in previous sections. Additionally, most studies
averaged SO2 concentrations over multiple monitors and examined 24-h avg exposure
metrics, which may not adequately capture the spatial and temporal variability in SO2
concentrations (Section 3.4.2). The studies that examined all respiratory disease hospital
admissions and ED visits generally reported positive associations (Figure 5-9). These
5-106
-------
associations were found to remain generally positive with some evidence of an
attenuation of the association in models with gaseous pollutants (i.e., NO2 and O3) and
particulate matter (U.S. EPA. 2008d').
Since the completion of the 2008 SOx ISA, recent studies have examined the association
between short-term exposure to ambient SO2 and all respiratory disease hospital
admissions and ED visits. For each of the studies evaluated in this section, Table 5-14
presents the air quality characteristics of each city or across all cities, the exposure
assignment approach used, and information on copollutants examined in each hospital
admission and ED visit study that examined all respiratory diseases. Other recent studies
that have examined all respiratory disease hospital admissions and ED visits are not the
focus of this evaluation because of various study design issues, as initially detailed in
Section 5.2.1.2. but the full list of these studies, as well as study specific details, can be
found in Supplemental Table 5S-6 (U.S. EPA. 2017c).
5-107
-------
Study
Location
Age
Lag
Cakmak et al. (2006)
10 Canadian cities
All
2.6
f Son et al. (2013)
8 South Korean cities
All
0-3
"fAtkinsonet al. (2012)
Meta-analysis (Asia)
All
NR
Burnett et al. (1997)a
Toronto, CAN
All
0-3
f Wong et al. (2009)
Hong Kong
All
0-1
Dales et al. (2006)
11 Canadian cities
0-27 days
2
Yang et al. (2003)
Vancouver, CAN
<3
2
f Son et al. (2013)
8 South Korean cities
0-14
0-3
Schwartz (1995)
Tacoma, WA
65+
0
Yang et al. (2003)
Vancouver, CAN
65+
0
Schwartz (1995)
New Haven, CT
65+
2
Schwartz et al. (1996)
Cleveland, OH
65+
0-1
f Wong et al. (2009)
Hong Kong
65+
0-1
Fung et al. (2006)
Vancouver, CAN
65+
0-6
f Son et al. (2013)
8 South Korean cities
75+
0-3
f Wong et al. (2009)b
Hong Kong
All
0-1
0-14
0-1
Wilson et al. (2005)
Peel et al. (2005)a
Tolbertet al. (2007)a
Wilson et al. (2005)
Portland, ME
Manchester, NH
Atlanta, GA
Atlanta, GA
Portland, ME
Manchester, NH
Portland, ME
Manchester, NH
Portland, ME
Manchester, NH
All
All
All
All
0-14
0-14
15-64
15-64
65+
65+
0
0
0-2
0-2
0
0
0
0
0
0
Hospital Admissions
ED Visits
i 1 1 1 1 1 1
-10 0 10 20 30 40 50
% Increase (95% Confidence Interval)
ED = emergency department; ISA = Integrated Science Assessment.
a = studies that used a 1-h max exposure metric.
b = Wong et al. (2009) also presented results for acute respiratory disease hospital admissions, which is a subset of total respiratory
hospital admissions.
Note: f and red text/symbols = recent studies published since the 2008 ISA for Sulfur Oxides; Black text/symbols = U.S. and
Canadian studies evaluated in the 2008 ISA for Sulfur Oxides. Corresponding quantitative results are found in Supplemental
Table 5S-10(U.S. EPA. 2017c).
Figure 5-9 Percent increase in respiratory disease hospital admissions and
emergency department visits from U.S. and Canadian studies
evaluated in the 2008 SOx ISA and recent studies in all-year and
seasonal analyses for a 10-ppb increase in 24-h avg or 40-ppb
increase in 1-h max sulfur dioxide concentrations.
5-108
-------
Table 5-14 Study-specific details and mean and upper percentile concentrations
from respiratory disease hospital admission and emergency
department visit studies conducted in the U.S. and Canada and
evaluated in the 2008 SOx ISA and studies published since the 2008
SOx ISA.
Upper Percentile
Mean of
Location Exposure Concentration Concentrations Copollutants
Study Years Assignment Metric ppb ppb Examined
Hospital admissions
Cakmak et al. 10 Canadian Average of 24-h avg 4.6 Max: 14-75 Correlations (r):
(2006) cities SO2 NR
(1993-2000) concentrations Copollutant
across all models: none
monitors in
each city
Dales et al. (2006) 11 Canadian Average of 24-h avg 4.3a 95th: 3.5-23.5 Correlations (r):
cities SO2 PM10: -0.09 to
(1986-2000) concentrations 0.61
across all O3: -0.41 to
monitors in
each city
0.13
NO2: 0.20 to
0.67
CO: 0.19 to
0.66
Copollutant
models: none
Burnett et al. (1997) Toronto, ON
(1992-1994)
Average of 1-h max 7.9
SO2
concentrations
from 4-6
monitors
during the
course of the
study
75th: 11 Correlations (r):
95th: 18 H+: 0.45
Max: 26 s°4: 0 42
PM10: 0.55
PM2.5: 0.49
PM10-2.5: 0.44
COH: 0.50
Os: 0.18
NO2: 0.46
CO: 0.37
Copollutant
models: COH,
PM10, PM10-2.5,
PM2.5
5-109
-------
Table 5-14 (Continued): Study specific details and mean and upper percentile
concentrations from respiratory disease hospital
admission and emergency department visit studies
conducted in the U.S. and Canada and evaluated in the
2008 SOx ISA and studies published since the 2008 SOx
ISA.
Study
Location
Years
Exposure
Assignment
Metric
Mean
Concentration
PPb
Upper Percentile
of
Concentrations
PPb
Copollutants
Examined
Fung et al. (2006)
Vancouver,
BC
(1995-1999)
Average of
SO2
concentrations
across all
monitors
within
Vancouver
24-h avg
3.46
Max: 12.5
Correlations (r):
CO: 0.61
COH: 0.65
Os: -0.35
NO2: 0.57
PM10: 0.61
PM2.5: 0.42
PM10-2.5: 0.57
Copollutant
models: none
Schwartz (1995)
New Haven,
Average of
24-h avg
New Haven:
New Haven:
Correlations (r):
CT
SO2
29.8
75th: 38.2
NR
Tacoma,
concentrations
Tacoma: 11.5
90th: 60.7
Copollutant
WA
across all
Tacoma:
models: PM10,
(1988-1990)
monitors in
75th: 21.4
O3
each city
90th: 28.2
Schwartz et al.
Cleveland,
Average of
24-h avg
35.0
75th: 45.0
Correlations (r):
(1996)
OH
SO2
90th: 61.0
NR
(1988-1990)
concentrations
Copollutant
across all
models: none
monitors
Yanq et al. (2003)
Vancouver,
Average of
24-h avg
4.8
75th: 6.3
Correlation (r):
BC
SO2
Max: 24.0
Os: -0.37
(1986-1998)
concentrations
Copollutant
across four
models: O3
monitors
tSon et al. (2013)
Eight South
Average of
24-h avg
3.2-7.3
NR
Correlation (r):
Korean
hourly
PM10: 0.5
cities
ambient SO2
Os: -0.1
(2003-2008)
concentrations
NO2: 0.6
from monitors
Copollutant
in each city
models: none
tAtkinson et al.
Meta-
NR
24-h avg
NR
NR
Correlation (r):
(2012)
analysis
NR
(Asia)
Copollutant
(1980-2007)
models: none
5-110
-------
Table 5-14 (Continued): Study specific details and mean and upper percentile
concentrations from respiratory disease hospital
admission and emergency department visit studies
conducted in the U.S. and Canada and evaluated in the
2008 SOx ISA and studies published since the 2008 SOx
ISA.
Upper Percentile
Mean of
Location Exposure Concentration Concentrations Copollutants
Study Years Assignment Metric ppb ppb Examined
tWona et al. (2009)
Hong Kong,
China
(1996-2002)
Average of
SO2
concentrations
from eight
monitoring
stations
24-h avg
6.8
75th: 8.4
Max: 41.8
Correlation (r):
NR
Copollutant
models: none
ED visits
Peel et al. (2005)
Atlanta, GA
(1993-2000)
Average of 1-h max
16.5
90th: 39.0
Correlations (r):
SO2
PM2.5: 0.17
concentrations
PM10: 0.20
from monitors
for several
PM10-2.5: 0.21
monitoring
UFP: 0.24
networks
PM2.5 water
soluble metals:
0.00
PM2.5 sulfate:
0.08
PM2.5 acidity:
-0.03
PM2.5 OC: 0.18
PM2.5 EC: 0.20
Oxygenated
HCs: 0.14
Os: 0.19
CO: 0.26
NO2: 0.34
Copollutant
models: none
5-111
-------
Table 5-14 (Continued): Study specific details and mean and upper percentile
concentrations from respiratory disease hospital
admission and emergency department visit studies
conducted in the U.S. and Canada and evaluated in the
2008 SOx ISA and studies published since the 2008 SOx
ISA.
Upper Percentile
Mean of
Location Exposure Concentration Concentrations Copollutants
Study Years Assignment Metric ppb ppb Examined
75th: 20.0 Correlations (r):
90th: 35.0 PMlo: °-21
03: 0.21
N02: 0.36
CO: 0.28
PM10-2.5: 0.16
PM2.5: 0.17
PM2.5 SO4: 0.09
PM2.5 EC: 0.22
PM2.5 OC: 0.17
PM2.5 TC: 0.19
PM2.5 water
soluble metals:
0.06
Organic
hydrocarbon:
0.05
Copollutant
models: none
NR Correlation (r):
Portland
Os: 0.05
Manchester
Os: 0.01
Copollutant
models: none
Avg = average; CO = carbon monoxide; COH = coefficient of haze; EC = elemental carbon; H+ = hydrogen ion; HC = hydrocarbon;
ISA = Integrated Science Assessment; max = maximum; N02 = nitrogen dioxide; NR = not reported; 03 = ozone; OC = organic
carbon; PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm; PM2.5 = particulate matter with
a nominal aerodynamic diameter less than or equal to 2.5 |jm; PM10-2 5 = particulate matter with a nominal aerodynamic diameter
less than or equal to 10 |jm and greater than a nominal 2.5 |jm; r= correlation coefficient; S02 = sulfur dioxide; S04 = sulfate;
TC = total hydrocarbon; UFP = ultrafine particle.
f = studies published since the 2008 SOx ISA.
Tolbert et al. (2007) Atlanta, GA
(1993-2004)
Average of 1-h max 14.9
SO2
concentrations
from monitors
for several
monitoring
networks
Wilson et al. (2005) Portland, SO2 24-h avg Portland:
ME concentrations 11.1
Manchester, from one Manchester:
NH monitor in 16 5
(1996-2000) each city
5-112
-------
Hospital Admissions
A recent multicity study conducted in Korea (Son et al.. 2013) and a single-city study
conducted in Hong Kong (Wong et al.. 2009) provide additional insight into the
relationship between short-term SO2 exposures and hospital admissions for all respiratory
diseases.
Son etal. (2013) examined the association between short-term exposures to air pollution
and respiratory-related hospital admissions in eight South Korean cities. The city-specific
means ranged from 3.2 to 7.3 ppb in this study. It is important to note that South Korea
has unique demographic characteristics with some indicators more in line with other
developed countries (e.g., life expectancy, percent of population living in urban areas),
but because it represents a rapidly developing Asian country, it is likely to have different
air pollution, social, and health patterns than less industrialized Asian nations or Western
nations that developed earlier (Son et al.. 2013). In a time-series analysis using a
two-stage Bayesian hierarchical model, Son et al. (2013) examined both single-day lags
and multiday lags up to 3 days (i.e., lag 0-3). For a lag of 0-3 days the authors reported a
5.6% increase (95% CI: 1.4, 10.0) in respiratory disease hospital admissions for a 10-ppb
increase in 24-h avg SO2 concentrations. The authors did not conduct copollutant
analyses; however, SO2 was found to be moderately correlated with PM10 (r = 0.5), NO2
(r = 0.6), and CO (r = 0.6). The results of Son et al. (2013) add additional support to the
results from the multicity studies evaluated in the 2008 SOx ISA [i.e., (Cakmak et al.
(2006); Dales et al. (2006))! in terms of the lag in which the strongest associations were
observed and the magnitude of the association (Figure 5-9).
A greater degree of variability in the magnitude of the association between short-term
SO2 exposures and all respiratory hospital admissions was observed when evaluating
single-city studies in the 2008 SOx ISA (Figure 5-9). Wong et al. (2009) in a study
conducted in Hong Kong reported results consistent with these earlier single-city studies
for individuals over the age of 65 [1.0% (95% CI: -0.8, 2.8) for a 10-ppb increase in 24-h
avg SO2 concentrations at lag 0-1], However, compared to studies that examined all
ages, the magnitude of the association was much smaller [0.8% (95% CI: -0.6, 2.3) for a
10-ppb increase in 24-h avg SO2 concentrations at lag 0-11. Wong et al. (2009) also
examined acute respiratory disease, which represents a smaller subset of outcomes within
all respiratory diseases. When focusing on only acute respiratory disease, Wong et al.
(2009) reported no evidence of an association at a 0-1-day lag for all ages [-2.0% (95%
CI: -4.4, 0.4) for a 10-ppb increase in 24-h avg SO2 concentrations].
The all-respiratory-disease hospital admissions results of Son et al. (2013) and Wong et
al. (2009) are supported by the results of a meta-analysis conducted by Atkinson et al.
5-113
-------
(2012) that focused on studies conducted in Asian cities since 1980. The six estimates
from studies that examined the association between SO2 and all respiratory hospital
admissions were included in a random effects model, which yielded a 1.3% increase in
respiratory hospital admissions (95% CI: -0.4, 3.2) for a 10-ppb increase in 24-h avg SO2
concentrations. However, Atkinson et al. (2012) found some evidence of publication bias
for associations between SO2 and respiratory hospital admissions.
Emergency Department Visits
The 2008 SOx ISA evaluated a few studies that examined the association between
short-term SO2 exposures and all respiratory ED visits [Figure 5-9. Supplemental
Table 5S-10 (U.S. EPA. 2017c)l. These studies reported evidence of a positive
association, but the magnitude of the association varied across study locations. However,
these studies were limited in that they did not examine copollutant confounding. Recent
studies that examined the association between air pollution and all respiratory ED visits
have not examined associations with SO2.
Model Specification—Sensitivity Analyses
A question that often arises when evaluating studies that examine the association between
air pollution and a health effect is whether the statistical model employed adequately
controls for the potential confounding effects of temporal trends and meteorological
conditions. Son et al. (2013). in the study of eight South Korean cities, conducted
sensitivity analyses to identify whether risk estimates changed depending on the df used
to control for temporal trends and meteorological covariates (i.e., temperature, humidity,
and barometric pressure). The authors reported that the association between short-term
SO2 exposures and all of the respiratory hospital admission outcomes examined (i.e., all
respiratory diseases, allergic disease, and asthma) was sensitive to using less than 7 df per
year, indicating inadequate control for temporal trends, but was stable when using
7-10 df per year. These results suggest that at least 7 df per year are needed to adequately
account for temporal trends when examining the relationship between short-term SO2
exposures and respiratory disease hospital admissions. However, additional studies have
not systematically examined this issue.
In an additional sensitivity analysis focusing on meteorological covariates
(i.e., temperature, relative humidity, and barometric pressure), Son et al. (2013) examined
whether risk estimates were sensitive to the degree of smoothing used and to the lag
structure. The authors found that when varying the number of df for each covariate from
3 to 6 df and varying the lag structure (i.e., lag 0 and lag 0-3 days), the SO2 association
remained robust for all respiratory hospital admission outcomes.
5-114
-------
Lag Structure of Associations
As stated previously, when examining associations between air pollution and a specific
health outcome, it is informative to assess whether there is a specific exposure window
for SO2 that results in the strongest association with the health outcome of interest. In the
examination of all respiratory disease hospital admissions, Son et al. (2013) focused on
both single-day and multiday lags to address whether there is evidence of an immediate
or persistent effect of SO2. Across single-day lags of 0 to 3 days, positive associations
were observed across each lag with the magnitude of the association being relatively
similar across each lag (i.e., 2.4% for lag 0 and 2.1% for lags 1 to 3 days for a 10-ppb
increase in 24-h avg SO2 concentrations). When examining multiday lags of 0-1, 0-2,
and 0-3 days, the authors reported an increase in the magnitude of the association as the
length of the multiday lag increased with a 3.5% increase reported at lag 0-1 and a 5.6%
increase reported for lag 0-3 days. Therefore, the limited evidence suggests that SO2
effects occur within the first few days after exposure, but also that SO2 effects on
respiratory disease hospital admissions may persist over several days.
Examination of Seasonal Differences
Of the studies that examined all respiratory disease hospital admissions or ED visits, only
Son etal. (2013) in the analysis of eight South Korean cities examined potential seasonal
differences in SO2 associations. However, it is important to note the potential influence of
geographic location on the results from studies that examine potential seasonal
differences in associations. For all outcomes examined, including respiratory diseases,
the association with SO2 was largest in magnitude during the summer, although
confidence intervals were quite large [respiratory diseases: 21.5% (95% CI: -0.7, 48.3),
lag 0-3, for a 10-ppb increase in 24-h avg SO2 concentrations] with additional evidence
of a positive association in the fall [8.9% (95% CI: -1.4, 20.7), lag 0-3, for a 10-ppb
increase in 24-h avg SO2 concentrations]. There was no evidence of an association
between short-term SO2 exposures and respiratory disease hospital admissions in either
the spring or winter seasons. Across the eight cities, mean 24-h avg SO2 concentrations
were lowest during the summer season (4.4 ppb compared to a range of 4.8 to 7.0 in the
other seasons) as was also the case for NO2 and CO.
Summary of Aggregate Respiratory Conditions
Recent studies add to the evidence detailed in the 2008 SOx ISA that indicated a
generally positive association between short-term SO2 exposures and respiratory disease
hospital admissions and ED visits (Figure 5-9). These recent studies provide some insight
5-115
-------
into previously identified limitations (i.e., model specification, lag structure of
associations, and potential seasonal differences) in the S02-respiratory disease hospital
admission and ED visits relationship. Initial evidence from a limited number of studies
suggests that SO2 associations are robust to alternative model specifications for weather
covariates and that SO2 associations are relatively stable in the range of df per year
indicative of reasonable control for temporal trends (i.e., 7-10 df per year); however,
more studies are needed to confirm these findings. Additionally, an examination of the
lag structure of associations is in line with the results reported in studies that focused on
a priori lags [i.e., associations tend to be strongest within the first few days after
exposure, primarily within the range of 0 to 3 days (Figure 5-9)1. The potential seasonal
patterns in SO2 associations remain unclear due to the variability in SO2 associations
observed across different geographic locations, as reflected in studies of other respiratory
hospital admission and ED visit outcomes. Some studies have also examined whether
there is evidence that specific factors modify the S02-respiratory disease hospital
admission or ED visit relationship and have found some evidence for potential
differences by lifestage and influenza intensity (see Chapter 6. Studies of all respiratory
hospital admissions and ED visits have not conducted extensive analyses to examine
potential copollutant confounding. However, studies that reported SO2 correlations with
other pollutants found low (r < 0.4) to moderate (r = 0.4-0.7) correlations. Overall, the
results of recent studies are limited in that they do not further inform the understanding of
potential confounding by copollutants on the relationship between short-term SO2
concentrations and respiratory disease hospital admissions and ED visits.
5.2.1.7 Respiratory Effects in General Populations and Healthy Individuals
The 2008 SOx ISA (U.S. EPA. 2008d) reported respiratory effects of SO2 in general
populations and healthy individuals but did not make specific conclusions about the
relationship. Respiratory effects were demonstrated in healthy individuals following SO2
exposures >1.0 ppm in controlled human exposure studies. Animal toxicological studies
demonstrated bronchoconstriction after a single SO2 exposure and provided evidence for
increased airway responsiveness and inflammation after repeated SO2 exposures.
Epidemiologic evidence was weak. The few recent toxicological studies corroborate
previous results, but recent epidemiologic and controlled human exposure studies provide
inconsistent results, including new results for pulmonary inflammation.
5-116
-------
Lung Function Changes in General Populations and Healthy Individuals
Compared with evidence for lung function changes in individuals with asthma, evidence
for SC>2-induced lung function effects in healthy individuals is weak. Most of the
controlled human exposure studies evaluating these effects in healthy individuals were
discussed in the 1982 SOx AQCD (U.S. EPA. 1982a). While some studies showed that
transient decreases in lung function can occur at concentrations of 1.0 ppm SO2 under
exercising or forced oral breathing conditions, the evidence was more consistent for
exposures >1.0 ppm (U.S. EPA. 2008d). Epidemiologic associations between ambient
SO2 concentrations and lung function continue to be inconsistent in children. While
recent results indicate associations in adults, inferences about SO2 exposure still are weak
because of uncertainty in the exposure estimates and copollutant confounding.
Controlled Human Exposure Studies
Evidence from controlled human exposure studies evaluating SC>2-induced lung function
changes in healthy adults was extensively discussed in the 1982 AQCD (U.S. EPA.
1982a). In general, these studies demonstrated respiratory effects such as increased
airway resistance and decreased FEVi following exposures to concentrations
>1.0-5.0 ppm, while some studies demonstrated respiratory effects at 1.0 ppm.
Lung function changes in response to SO2 exposure in controlled human exposure studies
have been investigated since the early 1950s. Respiratory effects including increased
respiration rates, decrements in peak flow, bronchoconstriction, and increased airway
resistance have been observed in healthy human volunteers at concentrations >1.0 ppm
(Lawther et al.. 1975; Andersen et al.. 1974; Snell and Luchsinger. 1969; Abe. 1967;
Frank etal.. 1962; Sim and Pattle. 1957; Lawther. 1955; Amduretal.. 1953). Although
bronchoconstriction was observed in healthy subjects exposed to concentrations
>5.0 ppm, shallow rapid respiration and increased pulse rate, decreased maximum
expiratory flow from one-half vital capacity, and increased sRaw were observed
following exposures as low as 1.0 ppm (Lawther et al.. 1975; Snell and Luchsinger. 1969;
Amdur et al.. 1953). Overall, only these few studies have reported SCh-induced
respiratory effects in healthy individuals for 5-10-minute exposures at concentrations
>1.0 ppm SO2.
A limited number of studies examined lung function changes in healthy populations in
response to >1 hour exposures to SO2. Controlled human exposure studies examining
lung function changes in healthy individuals exposed to SO2 are summarized in
Table 5-15. Andersen et al. (1974) reported that exposures of up to 6 hours to 1.0 ppm
SO2 in resting healthy adults induced decreases in FEF25-75 and to a lesser extent FEVi.
Another human exposure study (van Thriel et al.. 2010) reported that healthy subjects
5-117
-------
exposed to SO2 concentrations of 0.5, 1.0, or 2.0 ppm for 4 hours with two 15-minute
exercise periods separated by about 140 minutes did not show changes in FEVi following
the 4-hour SO2 exposure. However, lung function measurements in this study were not
performed until at least 35-85 minutes after exercise. Healthy individuals at rest or
exercising exhibited no changes in several measures of lung function following a 1-hour
exposure to 0.2-0.6 ppm SO2 (Tunnicliffc et al.. 2003; Linnetal.. 1987).
The interaction of SO2 exposure with O3 was reported in two studies. Hazucha and Bates
(1975) demonstrated that a combined 2-hour exposure to low concentrations of O3
(0.37 ppm) and SO2 (0.37 ppm) had a greater effect on lung function than exposure to
either agent alone in exercising adults. However, Bedi et al. (1979). using a similar study
design, did not observe a greater effect of the combined exposures compared with
exposure to only O3; exposure to SO2 alone had no effect.
5-118
-------
Table 5-15 Study-specific details from controlled human exposure studies of
lung function and respiratory symptoms in healthy adults.
Reference
Disease Status; n;
Sex; Age
(mean ± SD)
Exposure Details
(Concentration; Duration)
Endpoints Examined
Andersen et Healthy; n = 15; 15 M; 0, 1, 5, or 25 ppm SO2 for 6 h at
al. (1974) 20-28 yr rest
Nasal mucociliary flow
Area of the nasal airway
Airway resistance (FEV1, FEF25-75%)
Nasal removal of SO2
Discomfort level symptoms
Linn et al. Healthy; n = 24; 15 M,
(1987) 9 F; 18-37 yr
0, 0.2, 0.4, or 0.6 ppm SO2
1 h exposures
3 x 10-min exercise (bicycle)
periods -40 L/min
Exposures were repeated for a
total of eight
Lung function measure pre-exposure,
-15 min, and -55 min into exposure
sRaw, FVC, FEV1, peak expiratory flow
rate, maximal midexpiratory flow rate
Continuously EKG
Midway-HR
Before, during, 1-day after, and 1-wk after
symptom score, self-rated activity
Immediately after exposure-bronchial
reactivity percent change in FEV induced
by 3 min normocapnic hyperpnea with
cold, dry air
Raulf-
Heimsoth et
al. (2010)
Healthy; n = 16; 8 M,
8 F; 19-36 yr
0, 0.5, 1.0, or 2.0 SO2 for 4 h
with exercise for 15 min (bicycle,
75 Watts) two times during each
session
Exhaled NO, biomarkers of airway
inflammation in EBC and NALF
Tunnicliffe et
al. (2003)
Asthma; n = 12
adults, 35.7 yr
Healthy; n = 12
adults, 34.5 yr
0 or 0.2 ppm SO2 for 1 h at rest
Symptoms, FEV1, FVC, MMEF, exhaled
NO, ascorbic and uric acid in nasal
lavage fluid
van Thriel et
al. (2010)
Healthy; n = 16; 8 M,
8 F; M: 28.4 ± 3.9 yr,
F: 24.3 ± 5.2 yr
0, 0.5, 1.0, or 2.0 ppm SO2 for
4 h with exercise for 15 min
(bicycle, 75 Watts) two times
during each session
Symptoms, FEV1
EBC = exhaled breath condensate; EKG = electrocardiogram; F = female; FEF25-75% = forced expiratory flow at 25-75% of exhaled
volume; FEV = forced expiratory volume; FEVt = forced expiratory volume in 1 sec; FVC = forced vital capacity; HR = heart rate;
M = male; MMEF = maximum midexpiratory flow; n = sample size; NALF = nasal lavage fluid; NO = nitric oxide; SD = standard
deviation; S02 = sulfur dioxide; sRaw = specific airway resistance.
5-119
-------
Epidemiologic Studies
Previous epidemiologic evidence was inconsistent for an association between ambient
SO2 concentrations and lung function in healthy adults or children and people recruited
from the general population (U.S. EPA. 2008d). Studies mostly estimated SO2 exposure
from fixed-site monitors and did not report whether the measurements well captured the
spatiotemporal variability in the study areas. Some recent studies measured SO2 at
subjects' locations and observed associations with lung function decrements in adults but
not consistently in children. Most studies examined 24-h avg SO2 concentrations, which
are much longer than the 5-10 minute exposures inducing lung function decrements in
experimental studies. Inconsistency also is observed among recent results for temporally
resolved metrics such as 1-h max and 1- to 10-h avg SO2 concentrations, which is similar
to controlled human exposure findings for 1- to 6-hour exposures to SO2.
Adults. An S02-associated decrease in lung function was observed in adults in Beijing,
China where coal was used for domestic heating (Xu et al.. 1991). Recent results are
based on much lower SO2 concentrations [means 7.3-8.6 ppb vs. 6.8-49 ppb in Xu et al.
(1991)1. Associations are observed with lung function decrements in adults without
respiratory disease (Table 5-16). with some based on relatively good exposure
characterization (Dales et al.. 2013).
The exposure characterization of Dales et al. (2013) is judged to be good because SO2
was measured on site of adults' scripted exposures near (0.87 km) and away from
(4.5 km at a college campus) a steel plant in Ontario. Another strength was the
well-defined 8-hour exposure duration and lag between exposure and lung function
testing. Higher SO2 concentrations averaged over 10 hours (8 a.m.-6 p.m.) were
associated with decreases in several lung function parameters measured just after
exposure (Table 5-16). For example, a 10-ppb increase in SO2 was linked to a -0.50%
FEVi change (95% CI: -1.0, 0.05). Son et al. (2010) also examined air pollution from
industry, in this case a petrochemical complex in Ulsan, South Korea. Ambient SO2
concentrations across the study area were highly variable. Between-monitor correlation
varied widely (0-0.8), even for those 5 km apart, and the mean decreased from about 0.4
to 0.2 with increasing distance up to 20 km. Investigators aimed to capture this
spatiotemporal variability by combining SO2 measurements across monitors with inverse
distance weighting or kriging. These metrics and that for the nearest monitor to the
subjects' home, all 24-h avg SO2, were associated with forced vital capacity (FVC) but
not FEVi (Table 5-16). The implications overall are unclear because many subjects lived
far from a monitor, and potential confounding by meteorological factors and season were
not considered. Both studies observed associations with copollutants among PM2 5, PM10,
ultrafine particle (UFP), CO, NO2, and O3. Correlations between copollutants and
5-120
-------
analyses of confounding or interactions were not reported for personal exposures near the
steel plant (Dales et al.. 2013). For the study near the petrochemical complex, the
decrease in FEVi for kriged SO2 was larger after CO adjustment (Son et al.. 2010)
(Table 5-16). The effect estimate for CO became null, but the range of between-monitor
correlations was 0-0.8. The effect estimate for SO2 was attenuated with adjustment for
O3, and the attenuation could be influenced by differences in exposure measurement error
between the two pollutants. Between-monitor correlations were 0.4 to 0.8 for O3.
Other studies reported S02-associated lung function decrements, but inference about SO2
is weaker (Steinvil et al.. 2009; Min et al.. 2008a). Associations were observed for SO2
after adjustment for NO2 or CO, but correlations with SO2 were 0.62-0.70, and
single-pollutant associations for SO2 were in opposing directions across lags and limited
to lags of 3 or more days (Steinvil et al.. 2009). Associations were observed with 1-h avg
SO2 concentrations lagged 5-30 hours, but confounding by meteorological factors was
not considered (Min et al.. 2008a). Also, both studies had cross-sectional design and
estimated SO2 exposure from monitors up to 11 km or unspecified distance from homes.
Children. Similar to previous studies, many recent studies of children examined
populations with high prevalence (8-35%) of respiratory disease, such as asthma, and
populations outside the U.S. and Canada. As examined in several recent studies, SO2 at
schools was inconsistently associated with lung function (Table 5-17). Previously, 1-h
max SO2 concentrations at school were not associated with lung function. Additional
results for temporally resolved SO2 metrics, both school and fixed-site, are inconsistent.
5-121
-------
Table 5-16 Epidemiologic studies of lung function in healthy adults and adults in the general population
published since the 2008 ISA for Sulfur Oxides.
Study Population and
Methodological Details
SO2 Exposure
Estimates (ppb)
SO2 Averaging
Time and Lag
Day
Effect Estimate (95% CI)
Single-Pollutant Modela
SO2 Copollutant Model Results
and Correlations
tDales etal. (2013)
Sault Ste. Marie, ON, May-Aug 2010
N = 61, mean age 24 yr. 100% healthy.
Cross-over, with scripted outdoor exposures
near and away from steel plant. Five
consecutive 8-h days at each site, with 9-day
washout period in between. Supervised
spirometry. Recruited from university.
Required not to live in neighborhood bordering
steel plant.
Monitor on site of
outdoor exposures
Mean (SD)
Near steel plant
7.8 (13)
College campus
1.6 (4.2)
10-h avg
(8 a.m.-6 p.m.
Lag 0 h
Percent change
FEV1: -0.50 (-1.0, 0.05)
FVC: -0.45 (-1.1, 0.19)
FEV1/FVC: -0.15 (-0.31, 0.01)
FEF25-75%: -0.44 (-0.74, -0.14)
Total lung capacity
-0.42 (-0.70, -0.13)
Residual volume
-2.1 (-4.1, -0.18)
No copollutant model
Copollutant correlations NR.
5-122
-------
Table 5-16 (Continued): Epidemiologic studies of lung function in healthy adults and adults in the general
population published since the 2008 ISA for Sulfur Oxides.
Study Population and
SO2 Exposure
SO2 Averaging
Time and Lag
Effect Estimate (95% CI)
SO2 Copollutant Model Results
Methodological Details
Estimates (ppb)
Day
Single-Pollutant Modela
and Correlations
tSon etal. (2010)
13 monitors in city
24-h avg
Change in percent predicted
FVC, lag 0-2 avg, kriged SO2
Ulsan, South Korea, 2003-2007
Mean (SD), 75th
0-2 avg
FVC
Persists with: CO or O3
N = 2,102, ages 7-97 yr. Mean age 45 yr.
percentile, max
Kriging
PM2.5 not examined.
Mean percent predicted FEVi 83%.
Kriging
-6.2 (-8.2, -4.2)
Copollutant correlations NR.
Cross-sectional. Supervised spirometry.
8.3 (4.4), 9.6, 25
IDW
Recruited from a meeting of residents near a
Nearest monitor
-5.3 (-7.1, -3.5)
petrochemical complex. Did not examine
7.3 (5.9), 9.5, 34
Nearest monitor
confounding by meteorological factors or
IDW
-5.6 (-7.4, -3.9)
season.
8.4 (5.3), 11, 29
Average of 13 monitors
8.6 (4.1), 10, 24
Average of 13 monitors
-7.0 (-9.0, -4.8)
FEV1
Kriging
-0.08 (-0.76, 0.60)
IDW
0.31 (-0.32, 0.95)
Nearest monitor
0.35 (-0.21, 0.92)
Average of 13 monitors
-0.15 (-0.89, 0.58)
tSteinvil et al. (2009)
Three monitors within
24-h avg
Change in FEV1 (mL)
FEV1 (mL), lag 5
Tel Aviv, Israel, 2002-2007
11 km of home
0
93 (-90, 277)
Persists with: O3, NO2, or CO
N = 2,380, mean age 43 yr. 100% healthy.
Mean (SD): 2.8 (1.2)
5
-300 (-487, -113)
r= 0.70 NO2, 0.62 CO, -0.24 O3.
Cross-sectional. Supervised spirometry.
75th percentile: 3.4
0-6 avg
-447 (-750, -143)
Recruited from ongoing survey of individuals
Max: 9.4
Change in FVC (mL)
attending health center.
0
5
0-6 avg
0
5
0-6 avg
53 (-167, 273)
-373 (-600, -147)
-560 (-927, -193)
Percent change in FEV1/FVC
716 (-6.5, 4,233)
237 (-79, 2,195)
220 (-217, 657)
5-123
-------
Table 5-16 (Continued): Epidemiologic studies of lung function in healthy adults and adults in the general
population published since the 2008 ISA for Sulfur Oxides.
Study Population and
Methodological Details
SO2 Exposure
Estimates (ppb)
SO2 Averaging
Time and Lag
Day
Effect Estimate (95% CI)
Single-Pollutant Modela
SO2 Copollutant Model Results
and Correlations
100% no serious
tMin et al. (2008a)
South Korea, 2006
N = 867, ages 20-86 yr.
medical conditions.
Cross-sectional. Supervised spirometry.
Recruitment not described. Did not examine
confounding by meteorological factors.
Monitors in city
Number and distance
NR
Mean: 6
1-h avg
Lag 1 h
Results presented only in figure.
Associations observed only in
smokers. FEV1 and FVC
decrease after lag of 5-6 h. No
association after 30 h.
No copollutants examined.
Avg = average; CI = confidence interval; CO = carbon monoxide; FEF25-75% = forced expiratory flow at 25-75% of forced vital capacity; FE\A| = forced expiratory volume in 1 sec;
FVC = forced vital capacity; IDW = inverse distance weighting; max = maximum; N = population number; n=sample size; N02 = nitrogen dioxide; NR = not reported; 03 = ozone;
r = correlation coefficient; PM2 5 = particulate matter with nominal aerodynamic diameter less than or equal to 2.5 |jm; PM10 = particulate matter with nominal aerodynamic diameter
less than or equal to 10 |jm; SD = standard deviation; S02 = sulfur dioxide.
aEffect estimates are standardized to a 10-ppb increase in 1-h to 24-h avg S02.
f = Studies published since the 2008 ISA for Sulfur Oxides.
5-124
-------
Table 5-17 Epidemiologic studies of lung function in healthy children and children in the general population
published since the 2008 ISA for Sulfur Oxides.
Study Population and
Methodological Details
SO2 Exposure
Estimates (ppb)
SO2 Averaging
Time and Lag
Day
Effect Estimate (95% CI)
Single-Pollutant Modela
SO2 Copollutant Model Results
and Correlations
tCorreia-Deur et al. (2012)
Sao Paolo, Brazil, Apr-Jul 2004
N = 31, ages 9-11 yr. 100% no allergic
sensitization.
Daily measures for 15 days. Supervised
spirometry. Recruited from schools.
Monitor at school
Mean (SD): 8.8 (3.3)
75th percentile: 11
90th percentile: 13
2-h avg Percent change in PEF
0 -0.24 (-0.96, 0.49)
24-havg -0.20 (-1.4, 0.96)
0 No association for 3-, 5-, 7-, or
10-day avg
Remains null with: PM10, NO2, or
CO (analysis includes 65 children
. with atopy).
Pearson r= 0.75 PM10, 0.60 NO2,
0.60 CO.
tAltua et al. (2014)
Monitor at school
24-h avg
Relative ratio for change
No copollutant model
Eskisehir, Turkey, Feb-Mar2007
Mean and max
0-6 avg
Subjects without URS
PM2.5 and PM10 not examined.
N = 535, ages 9-13 yr
Suburban: 21, 29
FVC: 1.00 (0.97, 1.03)
r= 0.49 NO2, -0.40 O3.
Cross-sectional. Supervised spirometry.
Recruited from schools from participants of a
larger study.
Urban: 29, 44
Urban-traffic: 22, 27
FEV1: 1.00 (0.97, 1.03)
PEF: 1.00 (0.97, 1.03)
MMEF: 1.00 (0.92, 1.08)
Subjects with URS
FVC: 1.00 (0.97, 1.03)
FEV1: 1.00 (0.97, 1.03)
PEF: 1.00 (0.97, 1.03)
MMEF: 1.03 (0.95, 1.11)
5-125
-------
Table 5-17 (Continued): Epidemiologic studies of lung function in healthy children and children in the general
population published since the 2008 ISA for Sulfur Oxides.
SO2 Averaging
Study Population and
SO2 Exposure
Time and Lag
Effect Estimate (95% CI)
SO2 Copollutant Model Results
Methodological Details
Estimates (ppb)
Day
Single-Pollutant Modela
and Correlations
tAltuq etal. (2013)
Monitor at school
24-h avg
OR for impaired lung function
Remains null with: O3 or NO2
Eskisehir, Turkey, Jan 2008-Mar 2009
Mean and max
0-6 avg
(predicted values <85% for
PM2.5 and PM10 not examined.
N = 1,880, 9-13 yr. 7% asthma. 11% hay fever
Summer
FEV1 or FVC or <75% for PEF
r= 0.49 NO2, -0.40 O3. (winter).
Two measures: summer and winter.
Suburban: 8.5, 16
or MMEF)
Summer correlations NR.
Supervised spirometry. Recruited from
Urban: 10, 16
Summer
schools. Did not examine confounding by
meteorological factors.
Urban-traffic: 6.3, 8.9
Winter
Girls: 1.22 (0.72, 2.09)
Boys: 0.83 (0.47, 1.45)
Winter
Suburban: 21, 29
Urban: 29, 44
Girls: 1.00 (0.76, 1.32)
Urban-traffic: 22, 33
Boys: 0.83 (0.61, 1.11)
tCastro et al. (2009)
Monitor at school
24-h avg
Change in PEF (L/min)
No copollutant model
Rio de Janeiro, Brazil, 2004
Mean (SD): 7.1 (6.8)
1
-0.73 (-2.5, 0.99)
PM2.5 not examined.
N = 118, ages 6-15 yr. 18% asthma.
90th percentile: 16
2
-0.99 (-2.6, 0.61)
Copollutant correlations NR.
Daily measures for 6 wk. Supervised PEF.
Max: 37
3
0.34 (-1.1, 1.8)
Recruited from schools.
0-1 avg
0-2 avg
-1.8 (-3.8, 0.17)
-1.5 (-3.4, 0.46)
tChanq etal. (2012b)
Five monitors averaged
4-h avg
Change in FEV1 (mL)
No copollutant model
Taipei, Taiwan, 1996-1997
within 2 km of schools
0
0.4 (-32, 33)
PM2.5 not examined.
N = 2,919, ages 12-16 yr.
Means across districts
Copollutant correlations NR.
Cross-sectional. Supervised spirometry.
4-h avg (8 a.m.-12 p.m.):
10-h avg
Recruited from schools.
4.6-10
1
-117 (-193, -42)
10-h avg (8 a.m.-6 p.m.):
1.8-5.4
1-h max
1-h max:
0
3.6 (-21, 28)
5.9-35
1
-85 (-129, -41)
5-126
-------
Table 5-17 (Continued): Epidemiologic studies of lung function in healthy children and children in the general
population published since the 2008 ISA for Sulfur Oxides.
Study Population and
Methodological Details
SO2 Exposure
Estimates (ppb)
SO2 Averaging
Time and Lag
Day
Effect Estimate (95% CI)
Single-Pollutant Modela
SO2 Copollutant Model Results
and Correlations
tLinares et al. (2010)
Salamanca, Mexico, Mar 2004-Feb 2005
N = 464, ages 6-14 yr. 0.6% asthma.
Daily measures for 20 days in each season.
Supervised spirometry. Recruited from
schools.
Monitors within 2 km of
school
Means spring-winter
School 1: 12, 12, 10, 9.8
School 2: 9.1, 8.7, 10, 13
24-h avg Units not reported
0 FVC:-0.06 (-0.13, 0)
FEV1: -0.01 (-0.01, -0.00)
PEF: -0.03 (-0.05, 0)
FEV1/FVC: -0.07 (-0.18, 0.03)
No copollutant model
PM2.5 not examined.
Copollutant correlations NR.
tReddv et al. (2012)
Durban, South Africa, 2004-2005
N = 129, ages 9-11 yr. 37% asthma.
Daily measures for 3 wk each season.
Supervised spirometry. Recruited from
schools. Did not examine confounding by
meteorological factors except season.
Monitor at school
Mean (SD): 5.8 (0.2)
Max: 41
24-h avg Percent change FEV1 diurnal
variability (increase = poorer
function)
0-4 avg By GSTM1 gene variant
Null: -1.2 (-3.0, 0.54)
Positive: 1.1 (0.45, 2.7)
No copollutant model
PM2.5 not examined.
Copollutant correlations NR.
By GSTP1 gene variant
AG/GG: 3.1 (1.6, 4.7)
AA: -0.73 (-2.2, 0.70)
tMakamure et al. (2016a)
Durban, South Africa, 2004-2005
N = 71, ages 9-11 yr. 35% asthma.
Part of the same cohort as Reddv et al. (2012)
above.
Daily measures for 3 wk each season.
Supervised spirometry. Recruited from
schools. Did not examine confounding by
meteorological factors except season.
Monitor at school
Mean (SD): 5.8 (0.2)
Max: 41
24-h avg Percent change FEV1 diurnal
1 variability (increase = poorer
function)
All subjects: 1.6 (-0.03, 3.3)
By CD14 gene variant
CC: -1.5 (-3.4, -0.37)
CT/TT: -3.6 (-7.1, -0.17)
No copollutant model
PM2.5 not examined.
Copollutant correlations NR.
5-127
-------
Table 5-17 (Continued): Epidemiologic studies of lung function in healthy children and children in the general
population published since the 2008 ISA for Sulfur Oxides.
Study Population and
Methodological Details
SO2 Exposure
Estimates (ppb)
SO2 Averaging
Time and Lag
Day
Effect Estimate (95% CI)
Single-Pollutant Modela
SO2 Copollutant Model Results
and Correlations
tMakamure et al. (2016b)
Durban, South Africa, 2004-2005
N = 104, ages 9-11 yr. 39% asthma.
Part of the same cohort as Reddv et al. (2012)
above.
Daily measures for 3 wk each season.
Supervised spirometry. Recruited from
schools. Did not examine confounding by
meteorological factors except season.
Monitor at school
Mean (SD): 5.8 (0.2)
Max: 41
24-h avg Percent change FEV1 diurnal
variability (increase = poorer
function)
By TNF-a gene variant
AA/GA: 2.3 (-0.29, 5.0)
GG: 0.83 (-1.32, 3.0)
2 AA/GA: 2.7 (0.52, 4.8)
GG: 0.24 (-1.19, 1.68)
No copollutant model
PM2.5 not examined.
Copollutant correlations NR.
tAmadeo et al. (2015)
Pointe-a-Pitre, Guadeloupe, 2008-2009
N = 354, ages 8-13 yr. 17% asthma.
Cross-sectional. Supervised spirometry.
Recruited from schools.
Monitors in city
Number and distance NR
Mean (SD): 1.8 (1.4)
Max: 4.9
1-h max All subjects No copollutant model
0 Percent change post 6-min run PM2.5 not examined.
43 (-3,787, 3,873) Copollutant correlations NR.
24-h avg Children without asthma
0-13 avg Change in pre-run PEF (L/min)
18 (-84, 119)
Percent change post 6-min run
4.5 (-24, 33)
Avg = average' CI = confidence interval; CO = carbon monoxide; FE\A| = forced expiratory volume in 1 sec; FVC = forced vital capacity; ISA = Integrated Science Assessment;
max = maximum; MMEF = maximum midexpiratory flow; N = population number; n = sample size; N02 = nitrogen dioxide; NR = not reported; 03 = ozone; OR = odds ratio;
PEF = peak expiratory flow; PM2.5 = particulate matter with nominal aerodynamic diameter less than or equal to 2.5 |jm; PM10 = particulate matter with nominal aerodynamic diameter
less than or equal to 10 |jm; r= correlation coefficient; SD = standard deviation; S02 = sulfur dioxide; TNF-a = tumor necrosis factor-alpha; URS = upper respiratory symptoms.
+Effect estimates are standardized to a 10-ppb increase in 1-h to 24-h avg S02 or a 40-ppb increase in 1-h max S02.
f = Studies published since the 2008 ISA for Sulfur Oxides.
5-128
-------
For SO2 measured at schools, there is no evidence for association with lung function in
groups of children without respiratory disease or symptoms in Turkey or Brazil (Altug et
al.. 2014; Correia-Deur et al.. 2012). Altug et al. (2014) examined only 1-wk avg SO2,
but Correia-Deur et al. (2012) was noteworthy for examining multiple averaging times
and lags (i.e., 3-to 10-day avg). PEF, which also was measured at school, was not
associated with SO2 concentrations averaged over the preceding 2 hours [-0.24% change
(95% CI: -1.4, 0.96) in PEF per 10-ppb increase in SO2]. Another strength of this study
over similar ones is its repeated-measures design and clinical assessment of children's
respiratory health status. Among the studies of school SO2, an association with lung
function was observed in another cohort of children from Brazil (Castro et al.. 2009). The
impact of the 18% of children with asthma on these results is unknown. The effect
estimate was largest for 2-day avg SO2 concentrations and imprecise for lag 1 and 2
(Table 5-17). Missing SO2 concentration data for 52% of days could be one reason for the
imprecision.
Some results for SO2 measured at children's schools have more ambiguous implication
(Makamure et al.. 2016a. b; Altug et al.. 2013; Reddv et al.. 2012) (Table 5-17). For
children in Turkey, lung function was analyzed dichotomously based on a cutpoint of 85
or 75% of the predicted value (Altug et al.. 2013). Healthy children may not experience
such decrements, and the 7% of the cohort with asthma may influence results. In a South
African cohort, results were in opposing directions across the many comparisons made
among lung function parameters, pollutants, exposure lags, and gene variants (Makamure
et al.. 2016a. b; Reddv et al.. 2012). For example, an association of SO2 and FEVi was
found in children with the GSTP1 variant but not in children with the GSTM1 variant
(Table 5-17 and Section 6.4). Confounding by meteorology was not considered in either
cohort.
For exposures estimated from fixed-site monitors, lung function associations were
inconsistent for 1-h max SO2 (Amadeo et al.. 2015; Chang et al.. 2012b). which may be
more variable within a community and if uncharacterized, is subject to greater exposure
error. For children in Taiwan, a 40-ppb increase in 1-h max SO2 lagged 1 day was
associated with a -85 mL (95% CI: -129, -41) change in FEVi (Chang et al.. 2012b).
SO2 concentrations were averaged from five monitors within 2 km of children's schools.
For children in Guadeloupe, West Indies, the distance to monitors was not reported. Daily
1-h max SO2 concentrations were not associated with PEF (Amadeo et al.. 2015).
Although PEF was measured before and after a 6-minute exercise period, similar to
procedures in controlled human exposure studies, the SO2 metric was not likely matched
temporally with PEF measurements. Lung function in populations of children with low or
no prevalence of asthma was inconsistently associated with 24-h avg SO2 measured at
fixed-site monitors (Amadeo et al.. 2015; Linares et al.. 2010). although the null findings
5-129
-------
are for 13-day avg SO2 (Amadeo et al.. 2015). Airway responsiveness increased with
increases in 24-h avg SO2 in a population of children in which 8% had asthma and 18%
had atopy (Sovseth et al.. 1995). SO2 exposures were estimated from monitors within
2 km of homes, which is similar to studies observing associations with 24-h avg and 1-h
max SO2 (Chang et al.. 2012b; Linares et al.. 2010).
For the few associations observed for SO2 with lung function or airway responsiveness,
the potential for copollutant confounding or interactions was not addressed, including the
study conducted near an aluminum smelter that also emitted PM (Sovseth et al.. 1995).
Associations were observed for PM10, CO, NO2, and O3 measured at schools and
fixed-site monitors, but neither correlations with SO2 nor copollutant model results were
reported (Chang et al.. 2012b; Linares et al.. 2010; Castro et al.. 2009). Altug et al. (2014)
reported a moderate correlation with NO2 of 0.49 and observed no association for either
NO2 or SO2. Copollutant models were analyzed for long-term SO2, which was not
associated with lung function decrements in single-pollutant models (Linares et al..
2010). Importantly, none of the studies examined PM2 5
Animal Toxicological Studies
Lung function was examined in numerous studies reported in the 1982 SOx AQCD (U.S.
EPA. 1982a) and the 2008 SOx ISA (U.S. EPA. 2008d). The majority of these were
conducted in naive animals rather than in animal models of allergic airway disease.
Bronchoconstriction, indicated by increased pulmonary resistance, was identified as the
most sensitive indicator of lung function effects of acute SO2 exposure, based on the
observation of increased pulmonary resistance in guinea pigs acutely exposed to
0.16 ppm S02 (U.S. EPA. 2008d. 1982a). The 2008 SOx ISA (U.S. EPA. 2008d) reported
a few additional studies conducted at concentrations below 2 ppm. Animal toxicological
studies examining lung function changes in naive animals exposed to SO2 are
summarized in Table 5-18. Increased pulmonary resistance and decreased dynamic
compliance were observed in conscious guinea pigs exposed to 1 ppm SO2 for 1 hour
(Amdur et al.. 1983). Effects were seen immediately after exposure and were not present
1 hour post-exposure. No changes in tidal volume, minute volume, or breathing
frequency were found. These same investigators also exposed guinea pigs to 1 ppm SO2
for 3 hours/day for 6 days (Conner et al.. 1985). No changes were observed in lung
function or respiratory parameters (i.e., diffusing capacity for CO, functional reserve
capacity, vital capacity, total lung capacity, respiratory frequency, tidal volume,
pulmonary resistance, or pulmonary compliance). In another study, Barthelemv et al.
(1988) demonstrated a 16% increase in airway resistance following a 45-minute exposure
of anesthetized rabbits to 0.5 ppm SO2 via an endotracheal tube. This latter exposure is
more relevant to oronasal than to nasal breathing.
5-130
-------
Table 5-18 Study-specific details from animal toxicological studies of lung
function.
Species (Strain); n; Sex; Exposure Details
Study Lifestage/Age or Weight (Concentration; Duration) Endpoints Examined
Amdur et al. (1983) Hartley guinea pig; ~1 ppm (2.62 mg/m3); head Endpoints examined during
n = 8-23/group; M; age only for 1 h exposure and up to 1 h
NR; 200-300 g post-exposure.
Lung function—pulmonary
resistance, dynamic
compliance, breathing
frequency, tidal volume, and
minute volume
Endpoints examined 1, 24,
and 48 h after the sixth
exposure.
Lung function—residual
volume, functional residual
capacity, vital capacity, total
lung capacity, respiratory
frequency, tidal volume,
pulmonary resistance,
pulmonary compliance,
diffusing capacity for CO, and
alveolar volume
Barthelemv et al. (1988)
Rabbit; n = 5-9/group;
sex NR; adult; mean
2.0 kg; rabbits were
mechanically ventilated
0.5 ppm (1.3 mg/m3) for
45 min; intratracheal
Endpoints examined 5 min
before and up to 1 h
post-exposure.
Lung function—pulmonary
resistance
Amdur et al. (1988)
Guinea pig; n = 8
1 ppm for 1 h
Endpoints examined 2 h
following exposure
Airway responsiveness to
acetylcholine
Bronchial obstruction
determined by examination of
the respiratory loop
measured by whole-body
plethysmography in
spontaneously breathing
animals after each bronchial
provocation
5-131
Conner et al. (1985) Hartley guinea pig; 1 ppm (2.62 mg/m3); nose
n < 18/group/time point; only for 3 h/day for 6 days
M;age NR; 250-320 g
Riedel et al. (1988) Guinea pigs (Perlbright- 0.1, 4.3, and 16.6 ppm whole
White); n = 5-14; M; age body; 8 h/day for 5 days
NR; 300-350 g Animals were sensitized to
ovalbumin (ovalbumin
aerosol) on the last 3 days of
exposure
Bronchial provocation every
other day with aerosolized
0.1% ovalbumin began at
1 wk after the last exposure
to SO2 and continued for
14 days
4 groups:
Control
0.1 ppm SO2
4.3 ppm SO2
16.6 ppm SO2
-------
Table 5-18 (Continued): Study specific details from animal toxicological studies of
lung function.
Study
Species (Strain); n; Sex;
Lifestage/Age or Weight
Exposure Details
(Concentration; Duration)
Endpoints Examined
Park etal. (2001)
Guinea pigs
(Dunkin-Hartley);
n = 7-12/group; M; age
NR; 250-350 g
0.1 ppm whole body; 5 h/day
for 5 days
Animals were sensitized to
ovalbumin (0.1% ovalbumin
aerosol) on the last 3 days of
exposure
Bronchial challenge with
1% ovalbumin aerosol
occurred at 1 wk after the last
exposure to SO2
4 groups:
Control
Ovalbumin
Bronchial
obstruction—measurement of
Penh by whole-body
plethysmography
CO = carbon monoxide; M = males; n = sample size; NR = not reported; Penh = enhanced pause; S02 = sulfur dioxide.
The 2008 SOx ISA (U.S. EPA. 2008cD also described studies that examined airway
responsiveness following SO2 exposure. In several different animal species, a single
exposure to SO2 at a concentration up to 10 ppm failed to increase airway responsiveness
to a challenge agent. These studies were mainly conducted in naive animals rather than in
models of allergic airways disease. Only one was conducted at a SO2 concentration of
less than 2 ppm. This study found no change in airway responsiveness to acetylcholine
measured 2 hours following a 1-hour exposure in guinea pigs to 1 ppm SO2 (Amdur ct
al.. 1988). However, two toxicological studies (Park et al.. 2001; Riedel et al.. 1988)
described in the 2008 SOx ISA (U.S. EPA. 2008d'). provided evidence that repeated SO2
exposure of guinea pigs to concentrations as low as 0.1 ppm resulted in increased airway
responsiveness following subsequent sensitization and challenge with ovalbumin.
Summary of Lung Function Changes in General Populations and
Healthy Individuals
Across disciplines, there is limited evidence that short-term SO2 exposure induces lung
function changes in healthy people. Evidence from controlled human exposure studies of
healthy individuals shows that transient decreases in lung function can occur at
concentrations of 1.0 ppm SO2 under exercising or forced oral breathing conditions, but
the evidence is more consistent for exposures >1.0 ppm. Animal toxicological studies
demonstrated that acute exposure of guinea pigs to 0.16-1.0 ppm SO2 results in increased
airway resistance, and provided evidence that repeated exposure of guinea pigs to
concentrations of SO2 as low as 0.1 ppm led to increased airway responsiveness
following sensitization and challenge with an allergen. Epidemiologic studies do not
clearly indicate S02-associated decreases in lung function in healthy adults or children or
5-132
-------
groups from the general population with varying prevalence of respiratory disease.
Results are mixed for SO2 measured at subjects' locations and at fixed-site monitors.
Similar to experimental studies in healthy humans and animals without allergen challenge
plus 1-to 6-hour SO2 exposures, epidemiologic findings are mixed for temporally
resolved metrics such as 1-h max or 1- to 4-h avg SO2. Associations were observed for
populations living in locations with steel, aluminum, or petrochemical industry or coal
heating, but SO2 was one of many pollutants implicated.
Respiratory Symptoms in General Populations and Healthy Individuals
Respiratory symptoms in relation to short-term SO2 exposure have been investigated in a
limited number of studies of general populations or healthy individuals. The 2008 SOx
ISA (U.S. EPA. 2008d) described some controlled human exposure and epidemiologic
studies of respiratory symptoms among children or adults without asthma. Most
controlled human exposure studies reported no respiratory symptoms at concentrations up
to 2.0 ppm. Evidence from both previous and recent epidemiologic studies is
inconsistent.
Controlled Human Exposure Studies
Controlled human exposure studies examining respiratory symptoms in healthy
individuals exposed to SO2 are summarized in Table 5-15. Briefly, Tunnicliffe et al.
(2003) found no association between respiratory symptoms (i.e., throat irritation, cough,
and wheeze) and 1-hour exposures at rest to 0.2 ppm SO2 in either healthy adults or those
with asthma. Similarly, Andersen et al. (1974) reported no change in respiratory
symptoms in resting adults exposed to 1.0 ppm SO2 for 6 hours. A more recent study in
which exercising healthy adults were exposed to SO2 concentrations as high as 2.0 ppm
for 4 hours confirms these null findings (van Thriel et al.. 2010).
Epidemiologic Studies
Associations for ambient SO2 with respiratory symptoms in populations of healthy adults
and children are inconsistent. Most results are from Europe and Asia. There are more
studies of children than adults, but studies of adults focus on healthy individuals. Many
previous studies of children examined populations in which 5-81% of the children had
chronic wheeze, asthma, or atopy (Boezen et al.. 1999; Neas et al.. 1995). Some recent
studies examine populations of children with low (0.6-4%) prevalence of respiratory
disease, but like previous studies do not consistently associate increases in SO2
concentrations with respiratory symptoms (Table 5-19). Previous results were largely
based on 24-h avg SO2 concentrations measured at fixed-site monitors. Many recent
studies have improved exposure assessment, examining temporally resolved 1-hour SO2
5-133
-------
concentrations for adults or SO2 concentrations at children's schools. These associations
with respiratory symptoms also are inconsistent. Other uncertainties include confounding
by meteorological factors and copollutants.
For adults, a study on Miyakejima Island, Japan 5 years after a volcano eruption provided
information on effects related to SO2 concentrations and durations comparable to those
examined in experimental studies (Ishigami et al.. 2008). Compared to the reference
group (those living near monitors with 1-h avg S02 concentrations below 10 ppb),
incidence of many symptoms increased among people living near monitors with
concentrations above 100 ppb (Table 5-19). Although temporally resolved metrics were
analyzed, inference about an SO2 effect is weak. SO2 concentrations were measured
within 2 km of volunteer workers' homes and work sites, no other air pollutants or other
potential confounders were examined, and 80% of concentrations were in the reference
category. Results linking long-term air pollution from volcanoes to respiratory symptoms
also are uncertain because they are based on ecological comparisons of areas with low
and high air pollution mixtures in which SO2 is one constituent (Section 5.2.2.1).
5-134
-------
Table 5-19 Epidemiologic studies of respiratory symptoms in healthy adults and children and groups in the
general population published since the 2008 ISA for Sulfur Oxides.
Study Population
and Methodological Details
SO2 Exposure
Estimates (ppb)
SO2 Averaging
Time and
Lag Day
Effect Estimate (95% CI)
Single-Pollutant Modela
SO2 Copollutant Model
Results and Correlations
Adults
tlshiaami et al. (2008)
Monitors within 2 km of
1-h avg
Cough crude incidence rate, males
No copollutant model
Miyakejima Island, Japan, 2005
residence/work area
<10 ppb: 4.8, 10-20 ppb: 1.4,
No copollutants examined
N =611, ages >15 yr, 100% healthy
Daily diaries for 1-15 days. Recruited
Means across monitors
0-3,550
20-30 ppb: 2.9, 30-100 ppb: 6.6,
>100 ppb: 19.3. p for trend <0.01
from volunteers working on an active
volcanic island 5 yr after eruption. Did not
examine potential confounding factors
Max across monitors
3,790-10,320
1-h max
<10 ppb: 4.7, 10-20 ppb: 4.3,
20-60 ppb: 8.1, 60-2,000 ppb: 16.4,
>2,000 ppb: 58.3. p for trend <0.01
Children
tZhao et al. (2008)
Monitor at school
24-h avg
Outdoor SO2
No copollutant model
Taiyuan, China, Dec 2004
Mean (SD) and max
0-6 avg
Wheeze
PM2 5 not examined.
N = 1,993, ages 11—15 yr. 2% asthma. 4%
Outdoor: 271 (72), 386
OR: 1.01 (0.98, 1.04)
r= 0.74 NO2.
with furry pet or pollen allergy.
Indoor: 101 (53), 244
Daytime attacks of breathlessness
Cross-sectional. Recruited from schools.
OR: 0.99 (0.97, 1.01)
Likely temporal mismatch between current
Nocturnal attacks of breathlessness
SO2 concentrations and symptoms
OR: 1.01 (0.96, 1.06)
assessed as any occurrence in preceding
Indoor SO2
12 mo.
Wheeze
OR: 1.04 (1.01, 1.08)
Daytime attacks of breathlessness
OR: 1.02 (0.99, 1.04)
Nocturnal attacks of breathlessness
OR: 1.07 (1.01, 1.13)
5-135
-------
Table 5-19 (Continued): Recent epidemiologic studies of respiratory symptoms in healthy adults and children and
groups in the general population published since the 2008 ISA for Sulfur Oxides.
Study Population
SO2 Exposure
SO2 Averaging
Time and
Effect Estimate (95% CI)
SO2 Copollutant Model
and Methodological Details
Estimates (ppb)
Lag Day
Single-Pollutant Modela
Results and Correlations
tAltuq etal. (2014)
Monitor at school
24-h avg
Complaints of the throat in last 7 days
No copollutant model
Eskisehir, Turkey, Feb-Mar 2007
Mean and max
0-6 avg
RR: 0.83 (0.59, 1.15)
PM2.5 not examined.
N = 605, ages 9-13 yr. 7% asthma,
Suburban: 21, 29
Complaints of the throat at the moment
r= 0.40 Os, 0.49 NO2.
44% eczema.
Urban: 29, 44
RR: 1.03 (0.72, 1.47)
Cross-sectional. Recruited from schools
Urban-traffic: 22, 27
Runny nose in last 7 days
from participants of a larger study.
RR: 0.95 (0.74, 1.22)
Runny nose at the moment
RR: 0.92 (0.69, 1.23)
Shortness of breath/wheeze in last
7 days RR: 1.72 (1.05, 2.81)
Medication for shortness of
breath/wheeze in last 7 days RR: 1.44
(0.69, 2.99)
Shortness of breath/wheeze today
RR: 1.79 (0.90, 3.58)
Medication for shortness of
breath/wheeze today RR: 0.74 (0.16,
3.33)
tLinares et al. (2010)
Monitors within 2 km of
24-h avg
Wheezing OR: 1.06 (1.00, 1.11)
No copollutant model
Salamanca, Mexico, Mar 2004-Feb 2005
school
0
Rhinorrhea OR: 0.98 (0.92, 1.05)
PM2 5 not examined.
N = 464, ages 6-14 yr. 0.6% asthma.
Means spring-winter
Dyspnea OR: 1.02 (0.97, 1.07)
Copollutant correlations NR.
Cross-sectional. Recruited from schools.
School 1: 12, 12, 10, 9.8
School 2: 9.1, 8.7, 10, 13
5-136
-------
Table 5-19 (Continued): Recent epidemiologic studies of respiratory symptoms in healthy adults and children and
groups in the general population published since the 2008 ISA for Sulfur Oxides.
Study Population
and Methodological Details
SO2 Averaging
SO2 Exposure Time and
Estimates (ppb) Lag Day
Effect Estimate (95% CI)
Single-Pollutant Modela
SO2 Copollutant Model
Results and Correlations
tMoon et al. (2009)
Monitors in city
24-h avg LRS OR: 1.00 (0.93, 1.08)
0 URS OR: 1.11 (1.03, 1.20)
No copollutant model
PM2.5 not examined.
Copollutant correlations NR.
Seoul, Incheon, Busan, Jeju, South Korea, Number and distance NR
2003
N = 696, ages <13 yr
Means NR
Max: 38
Daily diaries for 2 mo. Recruited from
schools.
Avg = average; CI = confidence interval; ISA = Integrated Science Assessment; LRS = lower respiratory symptoms; max = maximum; n = sample size; N = population number;
N02 = nitrogen dioxide; NR = not reported; 03 = ozone; OR = odds ratio; PM2.5 = particulate matter with a nominal aerodynamic diameter less than or equal to 2.5 |jm; r= correlation
coefficient; RR = relative risk or ratio; SD = standard deviation; S02 = sulfur dioxide; URS = upper respiratory symptoms.
aEffect estimates are standardized to a 10-ppb increase in 1-h avg and 24-h avg S02 and a 40-ppb increase in 1-h max S02.
fStudies published since the 2008 ISA for Sulfur Oxides.
5-137
-------
For children, associations with SO2 concentrations were inconsistent within studies
among the array of symptoms examined (Table 5-19). Results across studies were
consistent for wheeze, an asthma symptom that is less likely to be experienced by healthy
children. The study in South Korea (Moon et al.. 2009) has many limitations, including
estimating SO2 exposure from fixed-site monitors at an unspecified distance from
children and observing only a few isolated associations among the numerous pollutants,
symptoms, exposure lags, and cities examined. Other studies had cross-sectional design
and measured SO2 at school or within 2 km from school (Altug et al.. 2014; Linares et al..
2010; Zhao et al.. 2008). The study in China examined high SO2 concentrations similar to
those in the Japanese volcano study. Mean school SO2 concentrations were 101 ppb
indoors and 271 ppb outdoors. Indoor, but not outdoor, 1-wk avg SO2 concentrations
were associated with symptoms (Zhao et al.. 2008) (Table 5-19). However, temporal
mismatch is likely because measured SO2 concentrations were compared with symptoms
that may have appeared at any time in the 12 months preceding the measurement. The
other study with 1-wk avg school SO2 measures, conducted in Turkey, observed an
association with any shortness of breath or wheeze in the previous 7 days but not throat
symptoms, runny nose, or medication use concurrently or in the previous 7 days (Altug et
al.. 2014). It is not clear whether the single positive association applied to the entire
population, the 7% with asthma, or 27% with hay fever. Among mostly healthy children
(0.6% asthma) in Mexico, lag 0 SO2 concentration was associated with wheeze, but SO2
was measured up to 2 km from children's schools (Linares et al.. 2010). SO2
concentrations were not associated with runny nose or difficulty breathing.
For the few observations of S02-associated increases in respiratory symptoms in healthy
adults and children, the potential for copollutant confounding was not examined. PM10,
CO, and formaldehyde were also associated with symptoms, but PM2 5 was not examined
(Table 5-19). Most studies did not report copollutant correlations, and none examined
copollutant models. Symptoms were not associated with outdoor NO2 (Altug et al.. 2014;
Linares et al.. 2010; Zhao et al.. 2008). but an association was observed with indoor NO2
(Zhao et al.. 2008). Indoor school SO2 and NO2 were highly correlated (r = 0.74), and it
is unclear to what extent the association with breathlessness can be attributed
independently to SO2, NO2, or to a combined effect of those and other copollutants.
Summary of Respiratory Symptoms in General Populations and Healthy
Individuals
There is little evidence for an effect of short-term SO2 exposure on respiratory symptoms
in healthy individuals. Controlled human exposure studies of healthy adults did not
demonstrate effects for 1- to 6-hour SO2 exposures up to 2 ppm, and epidemiologic
findings are inconsistent for healthy adults and children. For epidemiologic studies, there
is uncertain representativeness of SO2 exposures estimated from fixed-site monitors.
5-138
-------
However, as shown in recent studies, respiratory symptoms are also inconsistently
associated with SO2 measured at children's schools. A biological explanation for
associations observed with 1-wk avg SO2 concentrations is unclear. For associations
observed with 1-h avg or max concentrations and the evidence overall, potential for
confounding by PM2 5, PM10, NO2, CO, and formaldehyde is not addressed.
Subclinical Respiratory Effects in Healthy Individuals
Pulmonary inflammation is a key subclinical effect in the pathogenesis of asthma and
other respiratory diseases. It consists of both acute and chronic responses and involves
the orchestrated interplay of the respiratory epithelium and both the innate and adaptive
immune system. The immunohistopathologic features of chronic inflammation involve
the infiltration of airway epithelium by inflammatory cells, such as eosinophils,
lymphocytes, mast cells, and macrophages, and the release of inflammatory mediators
like cytokines and leukotrienes. The 2008 ISA for Sulfur Oxides described limited
evidence from animal toxicological studies for S02-induced pulmonary inflammation and
allergic sensitization in rodents exposed to an allergen. A single new animal toxicological
study also provides evidence of S02-induced pulmonary inflammation. Recent controlled
human exposure and epidemiologic studies add to the evidence base but do not clearly
support S02-related pulmonary inflammation in healthy populations.
Controlled Human Exposure Studies
A recent controlled human exposure study examined eNO and other biomarkers of
pulmonary inflammation in the nasal lavage fluid (NALF) and EBC after exposures to 0,
0.5, 1, and 2 ppm SO2 for 4 hours in exercising healthy adults (Raulf-Hcimsoth ct al..
2010). Data demonstrated no statistically significant changes in eNO; leukotriene B4,
prostaglandin E2, and 8-iso-prostaglandin F2 alpha in EBC; or substance P, interleukin-8,
and brain derived neurotrophic factor in NALF after SO2 exposures, compared to air.
Epidemiologic Studies
Unlike the study reviewed in the 2008 ISA for Sulfur Oxides (Adamkiewicz et al.. 2004).
recent studies measured SO2 near subjects' homes, schools, or work. SO2 concentrations
at a site within 1 km of most homes were not associated with pulmonary inflammation in
a population of children with high prevalence (33%) of asthma or atopy (Chen et al..
2012a). Previous results were similar for a population of older adults that included people
with respiratory disease. Recent examination of healthy adults and children in Beijing,
China indicates S02-associated increases in pulmonary inflammation or oxidative stress.
These recent studies were conducted before, during, and after the 2008 Olympics (Roy et
al.. 2014; Lin et al.. 2011b). Concentrations of SO2 and other pollutants were lower
5-139
-------
during the Olympics than before or after (e.g., mean 24-h avg 3.0 vs. 7.5 and 6.8 ppb).
During one period in winter 2007, mean 24-h avg SO2 concentrations were 45 ppb (Lin et
al.. 201 lb). Pollutants were measured 0.65 km from the school that the study children
attended and the hospital where most of the study adults worked. A 10-ppb increase in
lag 0 24-h avg SO2 was associated with a 7.6% (95% CI: 5.9, 9.3) increase in eNO of
children (Lin et al.. 2011b) and, in adults, a 0.67 standard deviation (95% CI: 0.48, 0.86)
increase in an index of pulmonary inflammation and oxidative stress combining eNO and
EBC markers (Rov etal.. 2014). Associations were also observed with PM2 5, sulfate,
EC/BC, CO, NO2, and OC. Copollutant models were analyzed for children, and SO2
effect estimates remained positive but decreased substantially with adjustment for PM2 5
or BC (Lin etal.. 2011b). Conversely, the effect estimate for BC was robust to
adjustment for SO2. Correlations with SO2 concentrations were not reported, and
confounding by other copollutants was not examined.
Animal Toxicological Studies
The 2008 SOx ISA (U.S. EPA. 2008d') described several animal toxicological studies that
examined the effects of repeated exposure to SO2 on inflammation. These and other
animal toxicological studies examining inflammation in naive animals exposed to SO2
are summarized in Table 5-20. In two of these studies, repeated exposure to SO2 was
found to promote allergic sensitization and to enhance allergen-induced bronchial
obstruction in guinea pigs. Riedel et al. (1988) examined the effect of SO2 exposure on
local bronchial sensitization to inhaled antigen. Guinea pigs were exposed by inhalation
to 0.1, 4.3, and 16.6 ppm SO2 for 8 hours/day for 5 days. During the last 3 days, SO2
exposure was followed by exposure to nebulized ovalbumin for 45 minutes. Following
bronchial provocation with inhaled ovalbumin (0.1%) 1 week later, bronchial obstruction
was measured by examining the respiratory loop obtained by whole-body
plethysmography. In addition, specific antibodies against ovalbumin were measured in
serum and BALF. The results showed significantly higher bronchial obstruction in
animals exposed to both SO2, at all concentration levels, and ovalbumin compared with
animals exposed only to ovalbumin. In addition, significant increases in antiovalbumin
immunoglobulin G (IgG) antibodies were detected in BALF of animals exposed to 0.1,
4.3, and 16.6 ppm SO2 and in serum from animals exposed to 4.3 and 16.6 ppm SO2 and
ovalbumin compared with controls exposed only to ovalbumin. These results
demonstrated that repeated exposure to SO2 enhanced allergic sensitization and bronchial
obstruction in the guinea pig at a concentration as low as 0.1 ppm.
5-140
-------
Table 5-20 Study-specific details from animal toxicological studies of
subclinical effects.
Species (strain);
n; Sex;
Lifestage/Age or
Exposure Details
Study
Weight
(Concentration; Duration)
Endpoints Examined
Conner et al. (1989)
Guinea pigs
1 ppm nose only; 3 h/day for
BAL performed each day.
(Hartley); n = 4; M;
1-5 days
BALF—total and differential cell
age NR;
counts
250-300 g;
Riedeletal. (1988)
Guinea pigs
0.1, 4.3, and 16.6 ppm whole body;
Endpoints examined 48 h after
(Perlbright-White);
8 h/day for 5 days
the last provocation.
n = 5-14/group; M;
Animals were sensitized to ovalbumin
Serum—anti IgG levels
age NR;
(ovalbumin aerosol) on the last 3 days
BALF—anti IgG levels
300-350 g;
of exposure
Bronchial provocation every other day
with 0.1% ovalbumin aerosol began at
1 wk after the last exposure to SO2
and continued for 14 days
Four groups:
Control
0.1 ppm SO2
4.3 ppm SO2
16.6 ppm SO2
Park et al. (2001)
Guinea pigs
0.1 ppm whole body; 5 h/day for
Endpoints examined 24 h after
(Dunkin-Hartley);
5 days
the bronchial challenge.
n = 7-12/group; M;
Animals were sensitized to ovalbumin
BALF—differential cell counts
age NR;
(0.1% ovalbumin aerosol) on the last
cells
250-350 g;
3 days of exposure
Lung and bronchial
Bronchial challenge with 1%
tissue—histopathology
ovalbumin aerosol occurred at 1 wk
after the last exposure to SO2
Four groups:
Control
Ovalbumin
SO2
Ovalbumin/SC>2
Li et al. (2007)
Rats (Wistar);
2 ppm SO2 for 1 h/day for 7 days
Endpoints examined 24 h
n = 6/group; M; age
following the last exposure
NR
BALF—inflammatory cell counts
Lung—histopathology and
immunohistochemistry
Lung and tracheal
tissue—mRNA and protein
levels of MUC5AC and ICAM-1
5-141
-------
Table 5-20 (Continued): Study specific details from animal toxicological studies of
subclinical effects
Study
Species (strain);
n; Sex;
Lifestage/Age or
Weight
Exposure Details
(Concentration; Duration)
Endpoints Examined
Li et al. (2014) Rats (Wistar); 2 ppm SO2 for 1 h/day for 7 days
n = 6/group; M; age
NR; 180-200 g
Endpoints examined
BALF—inflammatory cell counts
and cytokines IL-4, IFN-y,
TNF-a, IL-6
Serum—IgE
Lung—histopathology,
Lung and tracheal
tissue—mRNA and protein
levels NFkB, IkBq, IKK(3, IL-6,
IL-4, TNF-a, FOXp3,
EMSA NFkB binding activity
BAL = bronchoalveolar lavage; BALF = bronchoalveolar lavage fluid; EMSA = electrophoretic mobility shift assay;
FOXp3 = forkhead box p3; ICAM-1 = intercellular adhesion molecule 1; IFN-y = interferon gamma; IgE = immunoglobulin E;
IgG = immunoglobulin G; IKK(B = inhibitor of nuclear factor kappa-B kinase subunit beta; IL-4 = interleukin-4; IL-6 = interleukin-6;
IkBo = nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, alpha; M = male; MUC5AC = mucin 5AC
glycoprotein; n = sample size; NFkB = nuclear factor kappa-light-chain-enhancer of activated B cells; NR = not reported;
S02 = sulfur dioxide; TNF-a = tumor necrosis factor alpha.
Similarly, guinea pigs were exposed to 0.1 ppm SO2 for 5 hours/day for 5 days and
sensitized with 0.1% ovalbumin aerosols for 45 minutes on Days 4 and 5 (Park et al..
2001). One week later, the animals were subjected to bronchial challenge with 0.1%
ovalbumin, and lung function was evaluated 24 hours later by whole-body
plethysmography. The results demonstrated a significant increase in enhanced pause
(Penh), a measure of airway obstruction, in animals exposed to both SO2 and ovalbumin
but not in animals treated with ovalbumin or SO2 alone. In animals treated with both SO2
and ovalbumin, increased numbers of eosinophils were found in the lavage fluid. In
addition, infiltration of inflammatory cells, bronchiolar epithelial cell damage, and
plugging of the airway lumen with mucus and cells were observed in bronchial tissues.
These cellular changes were not observed in animals treated with ovalbumin or SO2
alone. Results indicate that repeated exposure to near-ambient levels of SO2 enhanced the
allergic inflammation due to subsequent sensitization and challenge with ovalbumin in
the guinea pig. Furthermore, increases in bronchial obstruction suggest that SO2 exposure
increased airway responsiveness in the animals subsequently made allergic to ovalbumin.
Park et al. (2001) demonstrated that repeated exposure of guinea pigs to 0.1 ppm SO2
alone did not lead to allergic inflammation or morphologic changes in the lung. Conner et
al. (1989) found no changes in total cells and neutrophils in BALF from guinea pigs
exposed repeatedly to 1 ppm SO2. In contrast, repeated exposure of rats to 2 ppm SO2
resulted in mild pathologic changes in the lung, including inflammatory cell influx and
5-142
-------
smooth muscle hyperplasia (Li et al.. 2014; Li et al.. 2007). Several other indicators of
inflammation and immune response were not changed by exposure to SO2 alone.
Summary of Subclinical Respiratory Effects in Healthy Individuals
There is limited evidence for inflammatory and other subclinical respiratory effects in
healthy populations following short-term exposure to SO2. As newly informed by recent
studies, SO2 is not clearly related to pulmonary inflammation in healthy populations in
controlled human exposure or epidemiologic studies. Associations were observed in
some epidemiologic studies, but confounding by PM2 5, sulfate, BC, or NO2 is not well
addressed. Studies in animals demonstrated that repeated exposure of guinea pigs to 0.1
or 1 ppm SO2 had no effect on inflammation. However, when followed by sensitization
with an allergen, exposure of guinea pigs to 0.1 ppm SO2 enhanced allergic inflammation
and bronchial obstruction to an allergen challenge. These results point to the potential for
SO2 exposure to increase airway responsiveness to an allergen. In addition, repeated
exposure of rats to 2 ppm SO2 resulted in inflammation and smooth muscle hyperplasia,
early indicators of airway remodeling.
Studies of Mixtures of Particles and Sulfur Oxides in Healthy Individuals
The 1982 AQCD (U.S. EPA. 1982a) addressed the question of possible effects of PM on
the response to SO2. It was noted that sorption of SO2 onto liquid or solid particles, which
may act as carriers, tended to increase its potency in animal toxicological experiments.
However, the mechanism for the effect was not known. Between the 1982 AQCD and the
2008 ISA for Sulfur Oxides (U.S. EPA. 2008d'). additional animal studies demonstrated
respiratory responses following inhalation of SO2 that was adsorbed onto metal oxide or
carbon particles. These studies are summarized in Annex Table E-4 of the 2008 ISA for
Sulfur Oxides and confirm and extend earlier findings. In all of the more recent studies,
the resulting particles were submicron in size and would be expected to deposit in the
lower respiratory tract. Acute and subacute exposures to SO2 and PM resulted in additive
or more-than-additive effects on pulmonary resistance, diffusing capacity for CO, airway
responsiveness following an acetylcholine challenge, and host defense responses. Many
of these studies reported transformation of SO2 to sulfite, sulfate, sulfur trioxide and
sulfuric acid (H2SO4), depending on temperature and relative humidity. Respiratory
responses observed in these experiments were in some cases attributed to the formation
of particulate sulfur-containing species. For example, repeated exposure to 20 |_ig/m3
carbon black-associated sulfate resulted in impaired host defense. Some studies of
laboratory-generated complex mixtures did not include a S02-only or a metal-only
exposure group, making it difficult to determine the relative contribution of these species.
In addition, the relevance of these animal toxicological studies has been called into
5-143
-------
question because the concentrations of both PM (1 mg/m3 and higher) and SO2 (1 ppm
and higher) used in these studies are much higher than ambient levels. Furthermore, the
SCh-adsorbed PM in some of these studies is not representative of ambient PM. For
example, some of the laboratory-generated aerosols contained sulfite but atmospheric
chemistry studies do not indicate significant amounts of sulfite ion in ambient PM. In
summary, animal toxicological studies suggest that SO2 effects may be potentiated by
coexposure to PM, but the relevance of these results to ambient exposures is not clear.
Summary of Respiratory Effects in General Populations and Healthy
Individuals
Overall, epidemiologic evidence for SC>2-induced lung function effects, symptoms, or
inflammation in healthy individuals is weak. Epidemiologic studies do not clearly support
associations of SO2 with lung function in healthy children or adults. Associations of SO2
with respiratory symptoms in children and pulmonary inflammation in healthy
populations are not consistently observed. Decrements in lung function, but not increases
in respiratory symptoms or inflammation were demonstrated in healthy individuals
following SO2 exposures >1.0 ppm in controlled human exposure studies. Animal
toxicological studies demonstrated bronchoconstriction after a single SO2 exposure.
Repeated exposures to SO2 were found to enhance allergic sensitization, allergic
inflammation, and bronchial obstruction in response to a subsequent allergen challenge.
In the absence of an allergen challenge, repeated exposure to SO2 resulted in
inflammation and smooth muscle hyperplasia, which is an early indicator of airway
remodeling. Furthermore, studies of mixtures of particles and sulfur oxides indicate some
enhanced effects on lung function parameters, airway responsiveness, and host defense.
However, some of these studies lack appropriate controls and others involve species of
sulfur oxides that may not be representative of ambient exposures.
5.2.1.8 Respiratory Mortality
Studies evaluated in the 2008 SOx ISA that examined the association between short-term
SO2 exposure and cause-specific mortality found consistent positive associations with
respiratory mortality using a 24-h avg exposure metric with some evidence indicating that
the magnitude of the association was larger compared to all-cause and cardiovascular
mortality. Recent multicity studies conducted in Asia (Chen et al.. 2012b; Kan et al..
2010) and Italy (Bellini et al.. 2007). a meta-analysis of studies conducted in Asia
(Atkinson et al.. 2012). and a four-city study conducted in China that focused specifically
on COPD mortality (Meng et al.. 2013) add to the initial body of evidence indicating
5-144
-------
increased respiratory mortality with short-term SO2 exposure (Section 5.5.1.3.
Figure 5-18).
Studies evaluated in the 2008 SOx ISA or in earlier documents [e.g., 1982 AQCD (U.S.
EPA. 1982a) I that examined the association between short-term SO2 exposures and
respiratory mortality focused exclusively on single-pollutant analyses. Therefore,
questions arose regarding the independent effect of SO2 on respiratory mortality, and
whether associations remained robust in copollutant models. A few recent multicity
studies conducted in China (Meng et al.. 2013; Chen et al.. 2012b) and multiple Asian
cities (Kan et al.. 2010) examined both of these questions. Chen et al. (2012b) found that
the SCh-respiratory mortality association was attenuated, but remained positive in
copollutant models with PM10 [2.03% (95% CI: 0.89, 3.17) for a 10-ppb increase in 24-h
avg SO2 concentrations at lag 0-1 days] and NO2 [1.16% (95% CI: -0.03, 2.37) for a
10-ppb increase in 24-h avg SO2 concentrations at lag 0-1 days]. These results are
similar to what the authors reported when examining the SCh-total mortality association
in models with PM10 (i.e., -40% reduction), but more attenuation was observed in models
with NO2 (i.e., -80% reduction for total mortality and 65% reduction for respiratory
mortality) (Section 5.5.1.4). Kan et al. (2010). as part of the Public Health and Air
Pollution in Asia (PAPA) study, also examined the effect of copollutants (i.e., NO2,
PM10, and O3), but only in each city individually. The study authors found that although
the S02-respiratory mortality association remained positive in copollutant models, there
was evidence of an attenuation of the association in models with PM10 and more so in
models with NO2 (Figure 5-10). Meng et al. (2013) in a four-city analysis of COPD
mortality in China reported evidence consistent with Chen et al. (2012b) and Kan et al.
(2010). The authors observed a 3.7% (95% CI: 2.4, 4.9) increase in COPD mortality for a
10-ppb increase in 24-h avg SO2 concentrations at lag 0-1 days. However, compared to
the results for respiratory mortality from copollutant models reported in Chen et al.
(2012b). Meng et al. (2013) found attenuation in models with PM10, -50% reduction
[1.9% (95% CI: 0.3, 3.5)] and N02, -99% reduction [0.0% (95% CI: -1.8, 1.9)]
compared to the SO2 results from the single pollutant model. The larger degree of
attenuation of the SO2-COPD mortality association in Meng et al. (2013). compared to
respiratory mortality in Chen et al. (2012b) could be a reflection of the smaller sample
size and smaller number of cities included in Meng et al. (2013). Overall, the studies that
examined the potential confounding effects of copollutants on the S02-respiratory
mortality relationship show results consistent with what has been observed for total
mortality. However, the overall assessment of potential copollutant confounding remains
limited, and it is unclear how the results observed in Asia are generalizable to other
locations, specifically due to the unique air pollution mixture and higher concentrations
observed in Asian cities.
5-145
-------
cc
t
o
E
Q
Q_
O
U
0)
CL
%
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
-0.5
-1.0
0 | 1 2 3 ! 4 5 6 7 101 04
S02
07
Lag
COPD = chronic obstructive pulmonary disease; S02 = sulfur dioxide.
Source: Adapted from Mena et al. (20131. Reprinted with persmission of Elevier.
Figure 5-10 Percent increase in chronic obstructive pulmonary disease
mortality associated with a 10 |jg/m3 (3.62 ppb) increase in 24-h
avg sulfur dioxide concentrations at various single and multiday
lags in four Chinese cities.
Of the studies evaluated, only Bellini et al. (2007) (in a multicity study conducted in
Italy) examined potential seasonal differences in the SC^-cause-specific mortality
relationship. Bellini et al. (2007) reported that risk estimates for respiratory mortality
were dramatically increased in the summer compared to the winter from 4.1 to 12.0% for
a 10-ppb increase in 24-h avg SO2 concentrations at lag 0-1 day (the all-year and winter
were similar). These results are consistent with the seasonal pattern of SO2 associations
observed in Bellini et al. (2007) for total and cardiovascular mortality. However, it
remains unclear whether this seasonal pattern of SC^-respiratory mortality associations is
observed in other locations.
An uncertainty that often arises when examining the relationship between short-term air
pollution exposures and cause-specific mortality is whether the lag structure of
associations and the C-R relationship is consistent with what is observed for total
mortality. Meng et al. (2013) addressed both the lag structure of associations and the C-R
5-146
-------
relationship in a study of short-term air pollution exposures and COPD mortality in four
Chinese cities. Although not explicitly part of the China Air Pollution and Health Effects
Study (CAPES) study, Meng et al. (2013) focused on four CAPES cities over the same
time period as Chen et al. (2012b). In comparison to Chen et al. (2012b). who found a
steady decline in risk estimates at single-day lags of 0 to 7 days with the largest effect at
lag 0-1, Meng et al. (2013) observed a steady decline over single lag days, but some
indication of larger associations, although highly uncertain, at longer multiday lags
(i.e., 0-4 and 0-7 days) (Figure 5-10). Note that Chen et al. (2012b) did not examine
multiday lags longer than 0-1 days, but the magnitude of the association for all
respiratory mortality [3.3% (95% CI: 2.1, 4.6) for a 10-ppb increase in 24-h avg SO2
concentrations] is similar to that reported in Meng et al. (2013) for COPD [3.7% (95%
CI: 2.4,4.9)].
Meng et al. (2013) also examined the shape of the SO2-COPD mortality C-R relationship.
To examine the assumption of linearity, the authors modeled the relationship between air
pollution exposures and COPD mortality using a natural spline with 3 df. Meng et al.
(2013) then computed the difference between the deviance of the linear and spline
models to assess whether there was evidence of nonlinearity in the SO2-COPD
relationship. As depicted in Figure 5-11. the authors found no evidence that the spline
model resulted in a better fit of the S02-mortality relationship compared to the linear
model. However, the authors did not present confidence intervals for each of the C-R
curves, which complicates the interpretation of the results.
5-147
-------
pg
o o
E
o
£L
o
o _
s-
(/>
sm
>i>
ty.
o
Beijing
Shanghai
Guangzhou
Hongkong
!¦¦¦¦
50 100
S02 concentration at lag 01 day
III il m
150
COPD = chronic obstructive pulmonary disease; S02 = sulfur dioxide.
Source: Adapted from Meng et al. (2013).
Figure 5-11 City-specific concentration-response curves for short-term sulfur
dioxide exposures and daily chronic obstructive pulmonary
disease mortality in four Chinese cities.
Overall, recent multicity studies report evidence of consistent positive associations
between short-term SO2 concentrations and respiratory mortality, which is consistent
with those studies evaluated in the 2008 SOx ISA. Unlike studies evaluated in the 2008
SOx ISA, recent studies examined whether copollutants confound the relationship
between short-term SO2 concentrations and respiratory mortality. Overall, these studies
reported evidence that the S02-respiratory mortality association was attenuated in models
with NO2 and PM10, but the analyses are limited to Asian cities where the air pollution
mixture and concentrations are different from those reported in other areas of the world.
Additional analyses focusing on seasonal patterns of associations, lag structure of
associations, and the C-R relationship are limited in number, but suggest evidence of:
larger associations in the summer/warm season, larger and more precise associations at
shorter lag periods (in the range of 0 and 1 days), and a log-linear, no-threshold C-R
relationship, respectively. However, for both total and cause-specific mortality, the
studies have conducted a rather limited exploration of potential alternatives to linearity
when examining the shape of the C-R relationship, which in combination with the
5-148
-------
potential measurement error due to uncharacterized spatial and temporal variability in
SO2 concentrations, complicates the interpretation of the SCh-mortality C-R relationship
(Sections 3.4.2.2 and 3.4.2.3).
5.2.1.9 Summary and Causal Determination
Strong evidence indicates that there is a causal relationship between short-term SO2
exposure and respiratory effects, particularly for respiratory effects in the at-risk
population of individuals with asthma (Section 6.3.1). This determination is based on
the consistency of SC>2-induced bronchoconstriction in exercising individuals with
asthma in controlled human studies, coherence of asthma-related effects among multiple
lines of evidence, and biological plausibility for effects specifically related to asthma
exacerbation. There is limited support for a relationship between short-term SO2 exposure
and other respiratory effects, including exacerbation of COPD, allergy exacerbation,
respiratory infection, respiratory effects in healthy populations, and respiratory mortality.
The limited and inconsistent evidence for these nonasthma-related respiratory effects
does not substantially contribute to the causal determination.
The determination of a causal relationship is the same as the conclusion of the 2008 SOx
ISA (U.S. EPA. 2008d). This conclusion is substantially based on controlled human
exposure studies that showed lung function decrements and respiratory symptoms in
adult individuals with asthma exposed to SO2 for 5-10 minutes under increased
ventilation conditions. These findings are consistent with the current understanding of
biological plausibility described in the mode of action section (Section 4.3.6). Previous
epidemiologic studies provided supporting evidence indicating associations between
short-term increases in ambient SO2 concentration and respiratory-related ED visits and
hospital admissions as well as respiratory symptoms. The evidence for a causal
relationship is detailed below using the framework described in the Preamble to the ISAs
(U.S. EPA. 2015b). While new evidence adds to the existing body of evidence, the
determination remains largely based on previous controlled human exposure studies. The
key evidence as it relates to the causal framework is presented in Table 5-21.
5-149
-------
Table 5-21 Summary of evidence for a causal relationship between short-term
sulfur dioxide exposure and respiratory effects.
so2
Concentrations
Rationale for Causal Associated with
Determination3 Key Evidence13 Key References'3 Effects0
Asthma exacerbation
Consistent evidence
from multiple,
high-quality controlled
human exposure
studies rules out
chance, confounding,
and other biases
Decreased lung function following exposures Section 5.2.1.2 400-600 ppb
of 5-10 min in exercising individuals with Table 5-2
asthma
A group of responders (defined as having Section 5.2.1.2 300 ppb
>15% decrease in FEVi after exposure to Table 5-3
0.6 or 1.0 ppm SO2) showed statistically
significant decrements in FEV1 following
5-10 min of exposure to 0.3 ppm SO2
Decreased lung function following exposures Section 5.2.1.2 200-300 ppb
of 5-10 min in 5-30% of exercising Table 5-2
individuals with asthma
Increased respiratory symptoms following Section 5.2.1.2 400-1,000 ppb
exposure of 5-10 min in exercising Table 5-2
individuals with asthma
Generally supporting
evidence from
multiple
epidemiologic studies
at relevant SO2
concentrations
Increase in asthma hospital admissions and
ED visits in single- and multicity studies,
among all ages, children and older adults
Section 5.2.1.2
1-h max:
9.6-10.8 ppb
24-h avg:
1.03-36.9 ppb
Limited evidence for respiratory symptoms in
children with asthma with school and/or
home SO2 measurements judged to well
represent subjects' exposure
tSpira-Cohen et al.
(2011). tVelicka et al.
(2015)
Sections 5.2.1.2. 3.5
24-h avg:
median 4.0 ppb
Uncertainty regarding
exposure
measurement error
SO2 exposures estimated from fixed-site
monitors may not capture spatiotemporal
variability of SO2 across a community
Section 3.4.2
Uncertainty regarding
potential copollutant
confounding
Some SO2 associations were relatively
unchanged in magnitude in copollutant
models with NO2, PM2.5, or PM10. Others
were attenuated. Uncertainty in extent to
which exposure measurement error is
comparable for SO2 and copollutants. SO2
showed a wide range of correlations with
copollutants across studies (r= 0.4-0.9).
Attenuated: tSpira-
Cohen et al. (2011)
Sections 5.2.1.2. 3.4.3
5-150
-------
Table 5-21(Continued): Summary of evidence for a causal relationship between
short term sulfur dioxide exposure and respiratory effects
Rationale for Causal
Determination3
Key Evidence13
Key References'3
SO2
Concentrations
Associated with
Effects0
Evidence for key
events in proposed
mode of action
Neural reflexes and/or inflammation lead to
bronchoconstriction
Section 4.3.6
Increased airway eosinophils in adults with
asthma exposed to SO2
Enhanced allergic inflammation in rats
previously sensitized with an allergen and
then repeatedly exposed to SO2
Gona etal. (2001). Li et
al. (2007). +Li et al.
(2014)
750-2,000 ppb
Enhancement of allergic sensitization,
allergic inflammation and airway obstruction
in guinea pigs exposed to SO2 repeatedly
over several days and subsequently
sensitized and challenged with an allergen
Park etal. (2001). Riedel
etal. (1988)
100 ppb
Allergic inflammation leads to increased
airway responsiveness. Association with
increased AHR in a population of adults with
asthma and a high prevalence of atopy
Taaaart et al. (1996)
24-h avg: max
39 ppb
Other respiratory effects
Limited and
inconsistent evidence
across disciplines
and outcomes
Inconsistent evidence for allergy
exacerbation, COPD exacerbation,
respiratory infection, respiratory diseases,
hospital admissions and ED visits, and
respiratory effects in healthy individuals
Section 5.2.1.3.
Section 5.2.1.4,
Section 5.2.1.5,
Section 5.2.1.6, and
Section 5.2.1.7
Respiratory mortality
Consistent
epidemiologic
evidence from
multiple studies at
relevant SO2
concentrations
Increases in respiratory mortality in multicity
studies conducted in the U.S., Canada,
Europe, and Asia
Section 5.2.1.8 and
Section 5.2.1.3
Fiqures 5-8 and 5-16
Mean 24-h avg:
U.S., Canada,
Europe:
0.4-28.2d ppb
Asia:
0.7->200 ppb
Table 5-39
Uncertainty regarding
potential confounding
by copollutants
No copollutant models with PM2.5. SO2
associations remained positive but
decreased in magnitude with adjustment for
PM10 or NO2, suggesting confounding.
Studies limited to areas with high SO2
concentrations, which complicates the
interpretation of independent association for
SO2.
Sections 5.2.1.8. 3.4.3
5-151
-------
Table 5-21(Continued): Summary of evidence for a causal relationship between
short term sulfur dioxide exposure and respiratory effects
Rationale for Causal
Determination3
Key Evidence13
Key References'3
SO2
Concentrations
Associated with
Effects0
Uncertainty regarding
SO2 exposures estimated from fixed-site
Section 3.4.2
exposure
monitors may not capture spatiotemporal
measurement error
variability of SO2 across a community.
Avg = average' COPD = chronic obstructive pulmonary disease; ED = emergency department; FE\A| = forced expiratory volume in
1 second; max = maximum; N02 = nitrogen dioxide; PM10 = particulate matter with a nominal aerodynamic diameter less than or
equal to 10 |jm; PM25 = particulate matter with a nominal aerodynamic diameter less than or equal to 2.5 |jm; r= correlation
coefficient; S02 = sulfur dioxide.
aBased on aspects considered in judgments of causality and weight of evidence in causal framework in Tables I and II of the
Preamble to the ISAs (U.S. EPA. 2015b).
bDescribes the key evidence and references, supporting or contradicting, contributing most heavily to causal determination and,
where applicable, to uncertainties or inconsistencies. References to earlier sections indicate where the full body of evidence is
described.
°Describes the S02 concentrations with which the evidence is substantiated (for experimental studies, below 2,000 ppb).
dThe value of 28.2 represents the median concentration from Katsouvanni et al. (1997).
fStudies published since the 2008 ISA for Sulfur Oxides.
5-152
-------
Evidence for Asthma Exacerbation
A causal relationship between short-term SO2 exposure and respiratory effects is
primarily supported by evidence from controlled human exposure studies of respiratory
effects in adults with asthma. These studies consistently demonstrated that the majority of
individuals with asthma experience a moderate or greater decrement in lung function, as
defined by a >100% increase in sRaw or >15% decrease in FEVi. This decrement is
frequently accompanied by respiratory symptoms following exposures of 5-10 minutes,
with elevated ventilation rates at concentrations of 0.4-0.6 ppm (Johns et al.. 2010; Linn
et al.. 1990; Linn et al.. 1988; Balmes et al.. 1987; Linn et al.. 1987; Horstman et al..
1986; Linn et al.. 1983b). A fraction of the population with asthma (-5-30%) has also
been observed to have decrements in lung function at lower SO2 concentrations
(0.2-0.3 ppm) (Linn et al.. 1990; l inn et al.. 1988; l inn et al.. 1987; Bethel et al.. 1985).
Although the degree of lung function decrements are considered moderate, they are less
likely to be accompanied by respiratory symptoms at these lower concentrations (Linn et
al.. 1990; Linn et al.. 1988; Linn et al.. 1987; Roger et al.. 1985; Linnetal.. 1983b). A
group of responders (defined as having >15% decrease in FEVi after exposure to 0.6 or
1.0 ppm SO2) showed statistically significant decrements in FEVi following 5-10 minute
exposure to 0.3 ppm SO2 (Johns et al.. 2010) (Table 5-3). While SC>2-induced respiratory
effects have been examined in individuals classified as having mild and moderate asthma,
these individuals are relatively healthy. Thus, extrapolating to individuals with severe
asthma is difficult because such individuals cannot be tested in an exposure chamber due
to the severity of their disease. Therefore, it is unknown whether people with severe
asthma are at increased risk of having respiratory effects due to short-term SO2 exposure.
The same may be said about children with asthma. There are no laboratory studies of
children exposed to SO2, but a number of studies have assessed airway responsiveness of
children and adults exposed to the bronchoconstrictive stimuli methacholine. Based
largely on those studies, school-aged children (-5-11 years of age), particularly boys and
perhaps obese children, might be expected to have greater responses (i.e., larger
decrements in lung function) following exposure to SO2 than adolescents and adults.
The coherence of the epidemiologic findings (Section 5.2.1.2) is supporting evidence for
a causal relationship. Epidemiologic evidence for lung function changes in adults and
children with asthma is inconsistent. However, short-term increases in ambient SO2
concentration are associated with increases in asthma hospital admissions and ED visits
among all ages, children (i.e., <18 years of age) and older adults (i.e., 65 years of age and
older) (Figure 5-3). as well as asthma symptoms in children (Velicka etal.. 2015; Spira-
Cohen etal.. 2011). Epidemiologic associations between short-term increases in ambient
5-153
-------
SO2 concentration and respiratory mortality provide support for a potential continuum of
effects between respiratory morbidity and respiratory mortality.
Most epidemiologic studies indicating associations between short-term SO2 exposures
and asthma exacerbation assigned exposure using SO2 concentrations measured at
fixed-site monitors. The use of fixed-site monitors to assign exposure, particularly to 1-h
max SO2, may introduce exposure measurement error if the spatiotemporal variability in
SO2 concentrations is not captured. The studies did not statistically correct for
measurement error. A few recent results reduce the uncertainty with SO2 measured or
modeled at or near children's school or home (Velicka et al.. 2015; Spira-Cohen et al..
2011). Additional uncertainty exists regarding potential copollutant confounding. In
many studies, SO2 was moderately to highly correlated with PM2 5, larger sized PM,
EC/BC, NO2, and VOCs (r = 0.4-0.9). The few available results show association with
sulfate. A small number of studies examined copollutant models. Some associations were
relatively unchanged in magnitude after adjustment for a copollutant; others did not
persist. However, inference from copollutant models is limited given potential differences
in exposure measurement error for SO2 compared to NO2, CO, PM, and O3 and in some
studies, high copollutant correlations. Copollutant interactions are not well studied. Some
controlled human exposure studies demonstrate increased asthma-related effects with
coexposure to SO2 and NO2 or O3. Limited epidemiologic evidence shows increased
asthma-related effects with joint increases in SO2 and copollutants but does not clearly
show a joint association that is greater than a single-pollutant association.
There is supportive evidence for a relationship between short-term SO2 exposure and
both airway responsiveness and pulmonary inflammation. Limited epidemiologic
evidence points to an association with increased AHR in a population of adults with
asthma and a high prevalence of atopy (Taggart et al.. 1996). Gong et al. (2001)
demonstrated an increase in airway eosinophils in adults with asthma 2 hours after a
10-minute exposure to 0.75 ppm SO2. This effect, along with bronchoconstriction, was
attenuated by pretreatment with a leukotriene receptor antagonist. Other pharmacologic
studies have demonstrated that inflammatory mediators play an important role in SO2
exposure-induced bronchoconstriction in people with asthma (Section 4.3.1). Further
support for an important role of airway inflammation, including allergic inflammation, is
provided by animal toxicological studies of repeated SO2 exposure in allergic animals
that are used to model the asthmatic phenotype (Li et al.. 2014; Li et al.. 2007). In
addition, repeated exposure of naive animals promoted allergic sensitization and
enhanced allergic inflammation (Park et al.. 2001; Riedel et al.. 1988). Increases in
bronchial obstruction also observed in these studies suggest that SO2 exposure increased
airway responsiveness in the animals subsequently made allergic to ovalbumin. These
5-154
-------
latter studies point to a possible increased sensitivity to allergens following SO2
exposure.
Evidence for Other Respiratory Effects
Epidemiologic studies demonstrate some associations of ambient SO2 concentrations
with hospital admissions and ED visits for all respiratory causes combined (Figure 5-9).
While these results suggest that the respiratory effects of short-term SO2 exposure could
extend beyond exacerbation of asthma, evidence across disciplines is inconsistent and/or
lacks biological plausibility for conditions such as allergy exacerbation (Section 5.2.1.3).
COPD exacerbation (Section 5.2.1.4). and respiratory infection (Section 5.2.1.5). Where
epidemiologic associations were found, potential copollutant confounding is uncertain.
For COPD exacerbation, a controlled human exposure study demonstrated no effect of
SO2 exposure, and epidemiologic associations are inconsistent for lung function,
respiratory symptoms, hospital admissions, and ED visits. Some evidence supports
SCh-associated increases in hospital admissions and ED visits due to respiratory
infections. However, the lack of multiple studies examining the same respiratory
infection outcome, inconsistent findings for self-reported infections in children, and the
lack of evidence from controlled human exposure and animal toxicological studies
produces uncertainty as to whether a relationship exists. Controlled human exposure
studies in healthy individuals provide evidence for transient decreases in lung function
with >1 ppm SO2 exposures for 5-10 minutes under exercising or a forced oral breathing
condition with no evidence for increased respiratory symptoms. Epidemiologic evidence
is inconsistent for SO2 associations with lung function, respiratory symptoms, and
pulmonary inflammation in healthy children and adults.
Conclusion
The evidence integrated across disciplines supports a causal relationship between
short-term SO2 exposure and respiratory effects, particularly asthma exacerbation.
This determination is primarily based on decreased lung function and increased
respiratory symptoms observed in controlled human exposure studies in adults with
asthma. Epidemiologic studies of asthma hospital admissions and ED visits and asthma
symptoms in children provide supporting evidence. There is also supportive evidence for
a relationship of short-term SO2 exposure with pulmonary inflammation and AHR.
Evidence for an effect of SO2 exposure on allergy exacerbation, COPD exacerbation,
respiratory infection, respiratory effects in healthy populations, and respiratory mortality
is inconsistent within and across disciplines and outcomes, and there is uncertainty
related to potential confounding by copollutants. The limited and inconsistent evidence
5-155
-------
for these nonasthma-related respiratory effects does not substantially contribute to the
causal determination.
5.2.2 Long-Term Exposure
The 2008 SOx ISA (U.S. EPA. 2008d) reviewed the epidemiologic and toxicological
evidence for long-term exposure to SO2 and respiratory effects and concluded that the
evidence was inadequate to infer a causal relationship. Although some positive
associations with asthma prevalence, bronchitis, symptoms, and lung function were
observed among children, uncertainties made it difficult at that time to assess the
evidence as a whole. Uncertainties related to assessing the consistency of findings across
a diverse set of respiratory outcomes, the potential for exposure measurement error to
influence results, and the lack of information available to assess the impact of copollutant
confounding were cited in the document. The studies of long-term exposure to SO2 and
respiratory morbidity that were considered in the last review are found in Supplemental
Table 5S-11 (U.S. EPA. 2017c). Animal toxicological studies of the effects of long-term
exposure to SO2, which were reviewed in the 2008 SOx ISA (U.S. EPA. 2008d).
examined lung function, morphology, and host defense. Most of these studies involved
SO2 concentrations well above 2 ppm. Recent toxicological studies add to this database.
Both older and more recent epidemiologic and toxicological studies that evaluate the
relationship between long-term SO2exposure and asthma (Section 5.2.2.1). allergy
(Section 5.2.2.2). lung function (Section 5.2.2.3). respiratory infection (Section 5.2.2.4).
other respiratory diseases (Section 5.2.2.5). and respiratory mortality (Section 5.2.2.6) are
discussed below. Overall, the collective evidence is strengthened by recent epidemiologic
studies reporting increases in asthma incidence among children and findings of animal
toxicological studies that provide a pathophysiologic basis for the development of
asthma.
Recent cohort studies of asthma incidence (Nishimura et al.. 2013; Clark et al.. 2010) use
a longitudinal design, a methodological enhancement over the cross-sectional studies of
asthma prevalence available in the 2008 SOx ISA (U.S. EPA. 2008d). In a recent study,
Ierodiakonou etal. (2015). using a longitudinal design, provided the first epidemiological
report relating SO2 exposure to increases in AHR (decreases in PC20) in children with
asthma. Uncertainties related to exposure estimates based on SO2 measurements from
monitors combined by inverse distance weighting (see Section 3.3.2.3) may limit the
inferences that can be made for these recent studies. The majority of other recent and
earlier epidemiologic studies used cross-sectional designs evaluating prevalence. Results
were generally positive, although the strength of the associations varied across studies.
5-156
-------
The designs used (i.e., ecological, cross-sectional) limit the contribution of these studies
to possible inferences about causality of relationships between long-term SO2 exposure
and respiratory effects. The caution expressed in the 2008 SOx ISA (U.S. EPA. 2008d)
related to the limitation of attributing an independent effect to SO2 (due to the
relationship of SO2 levels to PM levels) is still a concern. The evidence base does not
include studies evaluating concentration-responses, and few studies provide copollutant
model analyses. The 2008 SOx ISA (U.S. EPA. 2008d) found that animal toxicological
studies did not provide sufficient evidence to assess the effects of long-term SO2
exposure on lung function, morphology, or host defense. The one new subchronic animal
toxicological study that is discussed in this review found effects of SO2 exposure on
airway responsiveness, airway remodeling, and allergic inflammation. Short-term
toxicological studies also provide some evidence for these responses to SO2 exposure.
5.2.2.1 Development and Severity of Asthma
Development of Asthma
Asthma is described by the National Heart, Lung, and Blood Institute (NHLBINAEPP.
2007) as a chronic inflammatory disease of the airways that develops overtime.
Pulmonary inflammation can increase airway responsiveness, resulting in
bronchoconstriction (bronchial smooth muscle contraction), and in turn, episodes of
shortness of breath, coughing, wheezing, and chest tightness. When symptoms progress
to the stage that people seek medical treatment, a diagnosis of asthma can result.
Epidemiologic studies of SO2 used self- or parental report of a diagnosis to define
asthma. Epidemiologic studies reviewed in the 2008 SOx ISA (U.S. EPA. 2008d) were
limited to those with cross-sectional designs [Supplemental Table 5S-11 (U.S. EPA.
2017c)]. The majority of these studies reported positive associations of long-term SO2
exposure with asthma prevalence. A few recent longitudinal epidemiologic studies
support associations with asthma incidence and provide coherent evidence for
associations with respiratory symptoms in healthy populations. Uncertainty remains in
the adequacy of SO2 exposure estimates and copollutant confounding. However, some
support for an effect of SO2 exposure comes from a recent toxicological study providing
evidence for increased airway responsiveness.
A strength of recent epidemiologic studies of asthma development is their longitudinal
design (see Table 5-22). The follow-up of children over time to mark the first record of a
physician diagnosis with no prior record of diagnosis can better characterize the temporal
sequence between SO2 exposure and the incidence of asthma. In this regard, longitudinal
studies can better distinguish between onset of asthma and the exacerbation of asthma. In
5-157
-------
a large multicity study (N = 4,320 from Chicago, IL, Bronx, NY, Houston, TX, San
Francisco Bay Area, CA, and the territory of Puerto Rico), Nishimura et al. (2013)
observed that for SO2 exposures during the first year of life, the OR for asthma incidence
was 0.95 (95% CI: 0.59, 1.47) per 5 ppb change. SO2 exposure during the first 3 years of
life produced an OR for asthma incidence of 1.16 (95% CI: 0.73, 1.84) per 5 ppb SO2.
SO2 exposures were estimated using the inverse distance weighted average of the four
monitors within 50 km of the subject's residence. Selection bias due to differential loss to
follow-up is not an issue given the retrospective design.
Table 5-22 Selected epidemiologic studies of long-term exposure to SO2 and the
development of asthma and intervention studies/natural
experiments.
Location Mean SO2 Exposure Selected Effect
Study/Population Years ppb Assessment Estimates (95% Cl)a
Longitudinal studies of the development of asthma
tNishimura et al. Chicago, IL; Bronx, 4.0
(2013)
GALA II and
SAGE II cohorts
(Latinos and
African Americans
NY; Houston, TX;
San Francisco Bay
Area, CA; and the
territory of Puerto
Rico
8-21 yr) N = 4,320) (2006-2011)
IDW avg of
monitors within
50 km of
residence;
annual avg and
concentration
during first 3 yr
of life
Copollutant
correlations: NR
0.95 (0.59-1.47)—annual
avg
1.16 (0.74-1.84—early
life exposure
Covariate adjustment:
age, sex, ethnicity, and
composite SES.
tClark et al. (2010) Southwest British In utero
British Columbia Columbia Controls. 5.11
1999-2000 Cases: 5.22
1st yr of life:
Controls: 5.22
Cases: 5.37
Birth Cohort
(N = 37,401)
IDW avg of three
monitors within
50 km of postal
code centroid.
Concentrations
for in utero and
1st yr of life
estimated.
Copollutant
correlations: NR
1.47 (1.30-1.89) (both
in utero/1st yr of life)
Covariate adjustment:
native status,
breast-feeding, maternal
smoking, income quartile,
birth weight, and
gestational length
5-158
-------
Table 5-22 (Continued) Selected epidemiologic studies of long term exposure to
SO2 and the development of asthma and intervention
studies/natural experiments.
Study/Population
Location
Years
Mean SO2
PPb
Exposure
Assessment
Selected Effect
Estimates (95% Cl)a
tChianq et al.
(2016a). Chiang et
al. (2016b)
Recruited
587 children aged
between 11 and
14 yr from junior
high schools in
each of nine
townships.
N = 587
Incidence rates for
asthma (ICD-9;
493) were obtained
from the Taiwan
Health Insurance
Database.
Taiwan, near a
petrochemical
complex which
yields a diverse
pollution mix.
1999 to 2010
The 3-year average of the
99th percentile of SO2
levels in high and low
exposure areas after 2003
was 137.3 ppb and
32.0 ppb in the high and
low exposure areas,
respectively, between 2003
and 2006. From 2003 to
2010, there were 138 h
with hourly SO2
concentrations above
75 ppb each year in the
high exposure areas and
2 h in low exposure areas.
Two air quality
monitoring
stations, part of
the Taiwan
Environmental
Protection
Administration
(TEPA),
provided the
SO2 levels in the
HE and LE
areas. One is
located 8.1 km
south of the
complex, and
the other
16.2 km east
and south of the
complex. Three
exposure
periods were
reported since
opening of the
complex.
Copollutant
correlations NR.
The incidence rate of
asthma in the HE group
(18.5%) was significantly
higher than that in the LE
group (11.0%) in the first
4 yr after the complex
began its operations. A
difference in the
incidence of asthma
between the two groups
emerged after 12 mo, and
the maximum difference
appeared at 40 mo. The
hazard ratios of the
incidences of asthma,
during the different study
periods were adjusted for
group, age, gender, living
near roads, incense
burning and passive
smoking exposure. In
example for the third
study period
(1999-2010), HR (CI):
1.29 (0.91 to 1.83) for the
difference between high
and low exposure areas.
Intervention studies and natural experiments
Peters et al. (1996b)
Children
N = 3,521
Hong Kong,
China
(Kwai Tsing and
Southern
districts)
Period of study:
1989-1991
Annual avg (|jg/m3):
Southern
1989: 11
1990
1991
Kwai Tsing
1989
1990
1991
111
67
23
Pre- and
post-regulation
concentrations
compared in
natural
experiment; SO2
emissions were
reduced by 80%
post-regulation.
Associations between
respiratory symptoms and
living in polluted areas
observed and greater
decline in symptoms
post-regulation.
Covariate adjustment:
age, gender,
environmental tobacco
smoking in the family
home, housing and
father's education.
tWonq et al. (1998)
Children (9-
N =423
-12 yr)
Hong Kong, China
(Kwai Tsing and
Southern districts)
Period of study:
1989-1991
Annual avg (|jg/m3)
Southern
1989: 11
1990
1991
Kwai Tsing
1989
1990
1991
111
67
23
Pre- and
post-regulation
concentrations
compared in
natural
experiment; SO2
emissions were
reduced by 80%
post-regulation.
Decreased bronchial
responsiveness observed
post-intervention.
5-159
-------
Table 5-22 (Continued) Selected epidemiologic studies of long term exposure to
SO2 and the development of asthma and intervention
studies/natural experiments.
Study/Population
Location
Years
Mean SO2
PPb
Exposure
Assessment
Selected Effect
Estimates (95% Cl)a
tlwasawa et al.
(2009)
Miyake adults
(N = 823)
Miyakejima
Island, Japan,
near Mt. Oyama
volcano
2004-2006
31, post volcano (range:
19-45)
Inhabited areas were
classified into one lower
SO2 and three higher SO2
areas to gauge exposure.
Seven monitors
in residential
areas used to
estimate 2 yr
avg; natural
experiment
comparing
symptom
prevalence pre-
and
post-volcano
eruption.
Copollutant
correlations: NR
Minor health effects on
the respiratory system
observed.
Phlegm higher in higher
exposure areas.
Note: no consistent
differences in lung
function observed.
Logistic regression model
used. Covariate
adjustment: sex, age,
current smoking status,
residential area, and
hyper-susceptibility.
tlwasawa et al.
(2015)
120 Miyake school
children
Feb. 2005 to Nov.
2011
Average concentrations
(ppb) of SO2 decreased
year-by-year and ranged
from 11.3 to 2.47 in low
area, from 32.2 to 12.2 in
high area-1, and from 75.1
to 12.1 in high area-2.
Six monitors in
residential areas
used to estimate
post-volcano
eruption
concentrations
in different
residential
areas.
Other volcanic
gases were
measured and
considered to be
unlikely to cause
the health
effects seen in
the study.
Prevalence of respiratory
symptoms (cough,
phlegm, wheeze,
shortness of breath) was
increased in areas with
higher post-volcano SO2
concentrations compared
to areas with lower
concentrations.
Exposure-dependent
increases in symptoms
observed (no effects
observed at
concentrations lower than
30 ppb).
Logistic regression model
used. Covariate
adjustment: age, sex, and
hyper-susceptibility.
tLonao et al. (2008)
tLonqo (2009)
Adults (>20 yr)
N = 115 exposed
N = 110 unexposed
Kilauea volcano,
Hawaii
Apr. to Jun. 2004
24.5 (exposed)
0.7 (unexposed).
The volcanic plume is
carried over the exposed
area of the island by the
Pacific trade winds. The
unexposed area is located
at the extreme end of the
island from the volcano.
Ambient and
indoor SO2
concentrations
measured using
a network of
70 passive
samplers over a
3-wk sample
period.
Copollutant
correlations: NR
Cough on most days for
3 consecutive mo or more
(acute bronchitis) peryr
increased in areas with
higher levels.
Note: associations with
other symptoms also
reported.
Logistic regression model
used. Covariate
adjustment: age, sex,
race, smoking, dust, and
body mass index.
5-160
-------
Table 5-22 (Continued) Selected epidemiologic studies of long term exposure to
SO2 and the development of asthma and intervention
studies/natural experiments.
Study/Population
Location
Years
Mean SO2
PPb
Exposure
Assessment
Selected Effect
Estimates (95% Cl)a
tTametal. (2016)
1,836 4th/5th
graders mean age
10,1 yr
Kilauea volcano,
Hawaii
2002 to 2005
SO2, PM2.5, and particulate
acid in four exposure
zones. Mean (SD) SO2
across zones ranged from
0.3 to 10.1 ppb.
SO2 measured
by passive
diffusion for 1- to
4-wk intervals to
determine zone
levels at
representative
sites in each
zone.
Strongly acidic respirable
particulates associated
with cough. SO2 not
evaluated specifically but
included in the area mix
which was not related to
cough.
Cross-sectional study
with adjustments for age,
race, sex, sitting height,
BMI, premature birth,
maternal smoking during
pregnancy, current
smokers in the home, and
visible mold in the home.
Avg = average; BMI = body mass index; CI = confidence interval; HR = heart rate; ICD = International Classification of Diseases;
IDW = inverse distance weighting; n = sample size; N = population number; NR = not reported; PM25 = particulate matter with a
nominal aerodynamic diameter less than or equal to 2.5 |jm; SD = standard deviation; SES = socioeconomic status; S02 = sulfur
dioxide.
aEffect estimates are standardized per 5-ppb increase in S02 concentrations unless otherwise noted.
fStudies published since the 2008 ISA for Sulfur Oxides.
In a study of the British Columbia Birth Cohort (n =3,394 asthma cases), Clark et al.
(2010) used inverse distance weighted estimate-based concentrations from the three
closest monitors within 50 km of the participant's postal code to estimate SO2 exposure.
These authors observed an adjusted OR per 5 ppb of 1.48 (95% CI: 1.3, 1.9) due to
average exposures during both pregnancy and the first year of life. Conducted in
Southwest British Columbia, the study had 14 SO2 monitors available to provide data.
Clark et al. (2010) conducted a quartile analysis to explore the exposure-response
relationship and observed that the trend across quartiles was not linear [i.e., for the
first-year exposure model the second quartile was smaller (negative with confidence
intervals less than 1.0) than the positive first and last quartiles], lessening the strength of
the association. In this nested case-control study (n = 37,401), medical records of children
aged 3-4 years (born 1999-2000) were reviewed for asthma diagnosis (Clark et al..
2010). Selection bias due to differential loss to follow-up is not an issue because of the
records-based analysis used.
Asthma incidence for school children from the Taiwan Health Insurance Database was
evaluated contrasting high and low air pollution areas near a petrochemical complex for
three time periods after the opening of the complex. The areas were indexed by 3-year
annual average levels of the 99th percentile of SO2 levels and periods above 75 ppb
(Chiang et al.. 2016a. b). The hazard ratios (HRs) were positive with wide confidence
5-161
-------
intervals for the three periods. Caution is required in making inferences about an SO2
effect because (1) the air in the areas examined represent complicated mixes from
petrochemical complexes, (2) the uncertainty for exposure error may be too high to
include area comparisons rather than individual level comparisons, and (3) there is an
absence of evaluation for potential asthma risk factors.
The use of questionnaires in these studies to ascertain parents' report of
physician-diagnosed asthma, a strength of the study design (Burr. 1992; Ferris. 1978).
adds to the strength of inference about associations with SO2. A limitation of these
longitudinal studies include the potential for exposure measurement error related to the
use of inverse distance weighting for SO2 exposure estimates and comparison of high and
low concentration areas (see Section 3.3.2.3). Validation of SO2 exposures was not
discussed for these studies. The standard increment used in the current ISA, 5 ppb for an
annual average, is larger than the mean exposures in these studies, especially so for Clark
et al. (2010) where the mean exposure (SD) is 1.98 (0.97) ppb. Additionally, the strongest
associations observed in both studies were with NO2 concentration. Correlations between
pollutant concentrations were not reported by Chiang et al. (2016a); Nishimura et al.
(2013). while Clark et al. (2010) noted that correlations between pollutant concentrations
were generally high but did not provide quantitative data. These studies suggest a
potential relationship between long-term SO2 exposure and the development of asthma.
However, these results do little to reduce uncertainty related to potential copollutant
confounding.
These studies considered confounding by asthma risk factors, although the relation to
SO2 exposure and potential for confounding, is not well characterized for many of these
risk factors. Adjustment was made for SES with individual characteristics such as
parental education, income, or health insurance status (Nishimura et al.. 2013; Clark et
al.. 2010) (Table 5-22). Clark et al. (2010) additionally adjusted for neighborhood income
levels, and Nishimura et al. (2013) adjusted for race/ethnicity. These studies also adjusted
for maternal smoking, which has an unclear relationship to SO2 exposure. Obesity has
been identified as a potential risk factor for asthma in children (Gilliland et al.. 2003;
Gold et al.. 2003). In the cohorts examined by Nishimura et al. (2013). obesity was
associated with poorer asthma control (Borrell et al.. 2013). The role of obesity as an
effect modifier of SO2 was not examined in any of the studies of asthma. Nishimura et al.
(2013) did not find that atopy or family history of atopy modified the SO2 association.
Several recent studies presented in Supplemental Table 5S-12 (U.S. EPA. 2017c) also
examined the association of long-term exposure to SO2 with the prevalence of asthma in
cross-sectional designs with various SO2 exposure estimates as discussed in the table.
While these studies involve uncertainties, with the exception of Portnov et al. (2012).
5-162
-------
most reported positive associations (Liu et al.. 2016; Deng et al.. 2015a; Liu et al.. 2014a;
Dong et al.. 2013c; Dong et al.. 2013b; Kara et al.. 2013; Deger et al.. 2012; Portnov et
al.. 2012; Akinbami et al.. 2010; Sahsuvaroglu et al.. 2009). These studies are consistent
with similar studies in the 2008 SOx ISA (U.S. EPA. 2008d). Deng et al. (2015a')
examined copollutant models and reported that adjusting for PMio only slightly changes
the SO2 asthma risk. However, the SO2 association did not persist after adjusting for NO2.
Liu et al. (2016) only examined multipollutant models, which can be unreliable because
of multicollinearity among pollutants, and found that adjusting for both NO2 and PM10
attenuated the SO2 association. No longitudinal study of asthma incidence evaluated
copollutant models. Thus, within the recent epidemiologic evidence base, the studies
provided inconsistent new data and do not reduce the uncertainty related to whether the
effect was from SO2 or another pollutant. Studies of asthma incidence strengthen the
inference by addressing the temporality of exposure and response.
Supportive evidence for a relationship between long-term SO2 exposure and the
development of asthma is provided by cross-sectional studies of respiratory symptoms
related to asthma. In the 2008 SOx ISA (U.S. EPA. 2008d). studies examining an array of
respiratory symptoms related to SO2 exposure are presented in Supplemental Table F-6
(U.S. EPA. 2008d) and others are noted in the text of the 2008 SOx ISA (U.S. EPA.
2008d; Ware et al.. 1986; Chapman et al.. 1985; Dodge et al.. 1985). These
cross-sectional studies used monitors for the SO2 exposure estimate. While associations
were generally positive, some inverse or null associations were also observed. Recent
studies evaluating the relationship between long-term SO2 exposure and the prevalence of
asthma symptoms [Supplemental Table 5S-12 (U.S. EPA. 2017c)l also found positive
associations (Altug et al.. 2013; Pan et al.. 2010; Arnedo-Pena et al.. 2009).
Additional epidemiologic evidence for a link between long-term exposure to SO2 and the
development of asthma may come from intervention or natural experiment studies (see
Table 5-22). Physicians diagnose asthma, in part, based on the occurrence or
exacerbation of asthma symptoms, such as cough and wheeze, and the level of AHR in
the subjects. Decline in such symptoms and AHR in relation to a decline of a pollutant
level may support a relationship between asthma development and exposure to pollutants
such as SO2. Decreases in respiratory symptoms, including any wheeze or symptoms of
asthma, cough and sore throat, in 3,521 healthy children (mean age of 9.51 years) were
associated with decreases in SO2 concentrations in Hong Kong due to a government
restriction of sulfur content of fuels as discussed in the 2008 SOx ISA [see Peters et al.
(1996b). within U.S. EPA (2008d)l. During the same period, Wong et al. (1998)
examined the effect of the same decrease in SO2 concentrations on AHR in children aged
9-12 who were nonwheezing and did not have asthma at study entry. In the cohort
analysis, which compared measurements made before the intervention and 1 year
5-163
-------
afterwards, AHR declined. The subjective health measures seen in Peters et al. (1996b)
were corroborated by the objective data of the histamine challenge test in Wong et al.
(1998). These results should be interpreted with caution given the uncertainty of whether
changes in AHR and respiratory symptoms were independently related to SO2 in light of
the concomitant decline in sulfate respirable suspended particles (RSP) (<10 Over
the study period, SO2 declined about 80% (from about 111 to 23 |_ig/m3) while annual
mean sulfate concentrations in RSP fell from 12.5 to 7.7 (ig/m3. It is difficult to determine
whether one was more important than the other. However, these studies add to the
information base relating long-term SO2 exposure and asthma-related outcomes.
Recent cross-sectional studies that estimated long-term SO2 exposure from volcano
emissions in Japan and Hawaii were conducted (Table 5-22). Iwasawa et al. (2009)
observed increased frequencies of phlegm and minor effects on the respiratory system
among both adults and children residing near the Mt. Oyama volcano in Japan across four
inhabitant areas with varying SO2 levels. Iwasawa et al. (2015) further followed the
children yearly from 2006 to 2011, finding the prevalence of respiratory symptoms
(cough, phlegm, wheeze, shortness of breath) to be related to the higher SO2 exposure.
Studies conducted near the Kilauea volcano in Hawaii observed an adjusted increase in
cough on most days for 3 consecutive months or more per year in children and adults
(Longo. 2009; Longo and Yang. 2008; Longo et al.. 2008). Tam et al. (2016) related
cough to a mixture containing acidic respirable particulates, but not to SO2 exposure
directly, in children near the Kilauea volcano. As a whole, these studies are supportive of
a link between SO2 exposure and respiratory symptoms. However, such studies compare
areas of high volcano emissions to areas of lower emissions (indexed by SO2
concentration), and thus, results may be confounded by copollutant exposures.
Severity of Asthma
NHLBINAEPP (2007) identifies stages of asthma such as mild, moderate,
moderate-persistent, and severe. When going from mild to severe, the likelihood of acute
exacerbations increases. Stages of worsening of asthma are usually based on severity
scores as used in the following studies [Supplemental Table 5S-12, (U.S. EPA. 2017c)l.
Rage et al. (2009) examined severity of asthma in adults. Long-term SO2 exposure was
correlated with a higher asthma severity score. Ozone showed the strongest relationship
while NO2 was unrelated. In 17-year-old male military recruits, Greenberg et al. (2016)
related asthma severity to SO2 measured as low, intermediate, and high. The observed
associations between asthma severity and air pollution support the notion that air
pollutants may increase asthma severity. However, the uncertainty as to whether these
effects are influenced by short-term exposure needs to be examined. Degeretal. (2012)
observed an association with long-term SO2 exposure among children with active asthma
5-164
-------
and a more marked association among children with poor asthma control. No other
pollutants were examined. Adjusting for child's age and sex, parental atopy, and
environmental tobacco smoke exposure slightly decreased the association, and
stratification according to age (<6 years and >6 years) showed that associations with SO2
were mainly observed in the older age group. Adjusting for socioeconomic status
(i.e., household income and maternal educational level) had limited influence on the
results of the analyses (<5%).
AHR is a key component of asthma. In a recent study, long-term exposures to SO2,
estimated from monitors up to 50 km from subjects' ZIP code centroid, were associated
with increased methacholine responsiveness (i.e., decreased PC20) determined by FEVi
decreasing by 20% or more (Ierodiakonou et al.. 2015). A large number of comparisons
were made between pollutants, exposure lags, lung function parameters, cities, and
asthma medication groups possibly increasing probability that the few associations
observed may be due to chance. The PC20 percent change per interquartile range (2 ppb
4-month moving average) was -6% (95% CI, -11 to -1.5%) in 2,661 observations in the
Childhood Asthma Management Program (CAMP), a randomized clinical trial involving
eight cities in North America. The PC20 standardized to per 5 ppb is -15% (-27.5 to
-3.75%). The 4-month average SO2 was not associated with changes in lung function
measured before or after bronchodilator treatment. Health outcome results for 1-day and
1-week exposure periods are discussed earlier in Section 5.2.1.2; only the 4-month
moving average results are discussed here. The original health study, a longitudinal
prospective cohort study with repeated measures but without a pollution component, was
designed to examine the long-term safety and effectiveness of daily inhaled
anti-inflammatory medication in children diagnosed with mild to moderate asthma and
was sponsored by the NHLBI. The children were 5 to 12 years of age and
hyperresponsive to methacholine at study entry. Recruitment occurred from late
December 1993 to early September 1995 (CAMP Research Group. 1999; Cherniack et
al.. 1999) at two health maintenance organizations and six academic institutions.
Monitoring data on 24-h avg concentrations of pollutants ozone, CO, NO2, and SO2 were
obtained for each metropolitan area from the Aerometric Information Retrieval System
for the U.S. cities and from the Air Quality and Reporting Unit for Toronto were linked
to the ZIP code of the subject's address at study entry. There is uncertainty in the
measurement estimate and a potential for exposure measurement error. Distance or
proximity of sites to subjects' residence is not known. For long-term studies, bias can go
in either direction. Thus, the evidence base for a relationship between long-term SO2
exposure and AHR is limited.
5-165
-------
Animal Toxicological Studies
A single animal study of chronic SO2 exposure-related effects on lung morphology was
discussed in the 2008 SOx ISA (U.S. EPA. 2008d'). Study characteristics are summarized
in Table 5-23. Smith et al. (1989) found that rats exposed to 1 ppm of SO2 had an
increased incidence of bronchiolar epithelial hyperplasia and increased numbers of
nonciliated epithelial cells after 4 months of exposure. However, these effects were not
present at 8 months of exposure, suggesting that repair and/or adaptation may have taken
place.
5-166
-------
Table 5-23 Study-specific details from animal toxicological studies.
Study
Species (strain); n;
Sex; Lifestage/Age
Exposure Details
(Concentration; Duration)
Endpoints Examined
Smith et al. (1989)
Rats
(Sprague-Dawley);
n = 12-15 per data
point; M; young adult;
normal or
elastase-impaired
1 ppm (2.62 mg/m3) SO2
whole body; 5 h/day,
5 days/wk for 4 or 8 mo
8-mo exposure group
sacrificed immediately or
3 mo after exposure ended
Endpoints examined prior to
sacrifice
Lung function—residual volume,
functional residual capacity,
quasi-static compliance,
residual volume/total lung
capacity, N2 washout
Morphological effects
Song et al. (2012)
Rats
(Sprague-Dawley);
n = 10/group; M; 4 wk
old neonates
Sensitization by i.p. injection
of 10 mg ovalbumin followed
by booster injection of 10 mg
ovalbumin after 7 days
Challenge with 1% ovalbumin
aerosol for 30 min daily for
4 wk beginning at 15 days
Exposure to 2 ppm SO2 for
4 h/day for 4 wk beginning at
15 days
Exposure groups:
(1) Control
(2) SO2 alone
(3) Ovalbumin alone
(4) Ovalbumin + SO2
Endpoints examined 24 h after
challenge
Lung function—whole body
plethysmography (MCh
challenge)
BALF-IL-4, IFN-y
Serum-IL-4, IFN-y
Lung—histopathology
In vitro culture of airway smooth
muscle cells from
experimentally treated
animals—stiffness and
contractility
BALF = bronchoalveolar lavage fluid; IFN-y = interferon gamma; IL-4 = interleukin-4; i.p. = intraperitoneal; M = male;
MCh = methacholine; n = sample size; N2 = nitrogen; S02 = sulfur dioxide.
No studies on airway responsiveness or pulmonary inflammatory responses to long-term
exposure to SO2 concentrations of 2 ppm and lower were discussed in the 2008 SOx ISA
(U.S. EPA. 2008d). One new animal toxicological study of subchronic SO2 exposure has
become available since the last review. Key findings are discussed here, and study
characteristics are summarized in Table 5-23. Song et al. (2012) found that Penh, a
measure of bronchial obstruction, was enhanced following methacholine challenge in a
model of allergic airways disease. In this model, rats were first sensitized and challenged
with ovalbumin and then exposed to 2 ppm SO2 for 4 hours/day for 28 days. Penh was
not changed with exposure to SO2 alone in naive rats. However, Song et al. (2012)
observed hyperemia in the lung parenchyma and inflammation in the airways of naive
rats exposed only to SO2. SO2 exposure also increased the inflammatory responses in rats
made allergic to ovalbumin. Airway remodeling was found in ovalbumin-treated rats with
and without exposure to SO2. A more pronounced increase in the airway smooth muscle
layer was found in the ovalbumin/SCh group compared to the ovalbumin group. The
authors concluded that the effects of SO2 on bronchial obstruction and airway remodeling
5-167
-------
were dependent on ovalbumin sensitization and challenge. Song et al. (2012) also
measured concentrations of IL-4 and IFN-y in the BALF and serum of rats exposed to
SO2, with and without prior sensitization and challenge with ovalbumin. Concentrations
of IL-4 in the BALF were increased in the ovalbumin and the SO2 groups, with the
greatest increase occurring in the combined ovalbumin/SCh group. An increase in IL-4 in
serum occurred only in the ovalbumin/SCh group. Concentrations of IFN-y in the BALF
were decreased in the ovalbumin, SO2, and ovalbumin/S02 groups. A decrease in serum
IFN-y was observed in the ovalbumin and ovalbumin/SCh groups.
IL-4 is a Th2 cytokine associated with allergic responses, while IFN-y is a Thl cytokine.
An increase in the ratio of Th2 to Thl cytokines indicates Th2 polarization (or possibly a
Type 2 immune response mediated by Group 2 innate lymphoid cells), a key step in
allergic sensitization. These findings provide evidence that repeated SO2 exposure
enhances allergic responses and airway remodeling. Furthermore, increases in bronchial
obstruction suggest that SO2 exposure increased airway responsiveness in this model of
allergic airway disease. Furthermore, repeated SO2 exposure in naive rats increased levels
of the Th2 cytokine IL-4, decreased levels of the Thl cytokine IFN-y in the BALF, and
increased airway inflammation suggesting that SO2 exposure may induce allergic
sensitization on its own. Because allergic sensitization, airway remodeling, and increased
airway responsiveness are key events (or endpoints) in the proposed mode of action for
the development of asthma (Section 4.3.6). these results suggest that long-term exposure
to SO2 may lead to the development of an asthma-like phenotype in this animal model
involving newborn rats.
Summary of Asthma Development and Severity
Recent epidemiologic evidence from a limited number of longitudinal studies report
associations between asthma incidence among children and long-term SO2 exposures.
Additional supportive evidence for a link between long-term SO2 exposure and the
development of asthma is provided by cross-sectional studies of asthma prevalence. The
longitudinal studies help reduce the uncertainty associated with the temporality of
exposure and response that is inherent in cross-sectional study designs. This evidence is
coherent with animal toxicological evidence of inflammation, allergic sensitization and
other allergic responses, airway remodeling, and increased airway responsiveness, which
are key events (or endpoints) in the proposed mode of action for the development of
asthma (Section 4.3.6). The animal toxicological evidence provides support for an
independent effect of SO2 and strengthens the link between long-term exposure to SO2
and the development of asthma in children. Additional evidence supportive of this link
comes from cross-sectional studies of respiratory symptoms and respiratory allergies
among children and from natural experiments. Thus, multiple lines of evidence suggest
5-168
-------
that long-term SO2 exposure results in a coherent and biologically plausible
sequence of events that culminates in the development of asthma, especially allergic
asthma, in children.
The potential for a relationship between long-term SO2 exposure and severity of asthma
has been examined in a few studies. One study in adults correlated exposure with higher
asthma severity scores. A study in children found a more marked association in those
with poor asthma control. AHR, measured as PC20, worsened with long-term SO2
exposure in a multicity cohort of children. Thus, evidence of asthma control and
increased AHR provides suggestive but limited support for this relationship.
5.2.2.2 Development of Allergy
There is some evidence for a potential relationship between long-term SO2 exposure and
indicators of respiratory allergies and inflammation among children. Several recent
cross-sectional studies examined the prevalence of respiratory allergies using different
markers for respiratory allergies including IgE antibodies, rhinitis, eczema, sensitization
to pollen, and hay fever related to long-term SO2 exposure (Liu et al.. 2016; Chan et al..
2013; Bhattacharvva and Shapiro. 2010; Penard-Morand et al.. 2010; Parker etal.. 2009;
Nordling et al.. 2008) [see Supplemental Table 5S-11 (U.S. EPA. 2017c)l. Positive
results were observed for children using these various indicators of allergy. Further, a
weak relationship was found by Dales et al. (2008) between long-term SO2 exposure and
eNO, an indicator of inflammation [see Supplemental Table 5S-13 (U.S. EPA. 2017c)l.
Recent studies examine two-pollutant models for allergic rhinitis prevalence. Results for
allergic rhinitis prevalence based on responses from the ISAAC (International Study of
Asthma and Allergies in Childhood) questionnaire data in Changsha, China (Chan et al..
2013) did not find an association for SO2 for site-specific background SO2 and allergic
rhinitis in children 3-6 years old, but did find an association for age-related accumulative
exposure in a single pollutant model using the closest monitor to kindergartens. The
two-pollutant model with PM10 was attenuated. For SO2 exposures during the first year of
life in Shanghai, China, Liu et al. (2016) found an association with allergic rhinitis in
children at age 6 years, which was attenuated when adjusted for other pollutants using
district monitors. These findings suggest the possibility that chronic exposure to SO2 may
play a role in the development of allergic conditions based on results for various allergic
markers. The cross-sectional design of these studies makes these relationships uncertain
with regard to the temporal relationship between exposure and outcome. Further, the
exposure estimates from monitors may not adequately characterize the spatial and
temporal variation in SO2 concentrations potentially leading to exposure measurement
5-169
-------
error (Section 3.4.2). Although two pollutant models have begun to address the role of
SO2 exposure in the development of allergic rhinitis, the evidence base for a relationship
between long-term SO2 exposure and this response is limited.
5.2.2.3 Lung Function
Epidemiologic Studies
Longitudinal epidemiologic studies examine associations between long-term SO2
exposure and decrements in lung function. Lung function increases through early
adulthood with growth and development, then declines with aging (Stanoicvic et al..
2008; Zeman and Bennett. 2006; Thurlbeck. 1982V Thus, a relationship between
long-term SO2 exposure and decreased lung function over time in school-age children
into early adulthood would be an indicator of decreased lung development.
As discussed in the 2008 SOx ISA (U.S. EPA. 2008d). earlier cross-sectional studies
(Dockery et al.. 1989; Schwartz. 1989) found no association between long-term SO2
exposure and lung function in U.S. children. A longitudinal cohort study (Frischer et al..
1999) reported that long-term SO2 exposure was associated with decrements in lung
function in the summer but not in the winter. In Poland, a prospective cohort study of
children (Jedrvchowski et al.. 1999) found decrements in lung function growth related to
a polluted area where concentrations of both TSP and SO2 were high compared to a
cleaner area where concentrations of both TSP and SO2 were low, thus not providing
results specifically for SO2. In a cross-sectional study in adults in Switzerland,
Ackermann-Liebrich et al. (1997) observed an association between SO2 concentration
and lung function, but after controlling for PM10, this association was no longer evident.
In former East Germany from 1992 to 1999, Frye et al. (2003) reported improvements in
lung function associated with declines in SO2 concentrations in 2,493 children over three
cross-sectional surveys. These studies are presented in Supplemental Table 5S-11 (U.S.
EPA. 2017c).
Recent studies in children and adults add to this evidence base [see Supplemental
Table 5S-14 (U.S. EPA. 2017c)l. In the repeated measure prospective Taiwan Children's
Health Study, Hwang et al. (2015a) examined lung function over a 2-year period from
age 12 to 14 years. No association was found for SO2 exposure and FEVi or FVC for
boys and girls, but a deficit was observed for boys for FEF25 75. A single measure
longitudinal study in several U.S. cities observed an association between FEVi and SO2
during the first year of life [-1.01 (95%CI: -3.25, 1.27) per 1 ppb increase]. In this study,
Neophvtou et al. (2016) examined the same cohort that Nishimura et al. (2013) did, as
5-170
-------
discussed earlier in this section for asthma incidence in the same cities with the same SO2
exposure method evaluating the same confounding factors plus obesity.
In a longitudinal repeated-measures study of children, Linares et al. (2010) reported a
decline in FEVi related to long-term SO2 exposure in the entire study group. This study
included children from two schools in different locations relative to a petrochemical
zone. In an analysis of the children by sex, in single- and copollutant models with PM10
or O3, the association was attenuated. In a cross-sectional study of children in
14 communities in Taiwan, Lee etal. (2011c') found a reduction in FEVi related to
long-term SO2 exposure with larger reductions related to NO2 and CO exposure. Yogev-
Baggio et al. (2010) related the effect of the interaction, NOx x SO2 "event," to reduction
in FEVi in children in Israel near a coal-fired power plant. In a cross-sectional study of
32,712 adults in England, Forbes et al. (2009c') reported associations of FEVi with
exposure to SO2, PM10, and NO2, but not O3. Dales et al. (2008) found a weak decline in
FEVi and FVC related to long-term SO2 exposure in school children in Windsor, ON
using a cross-sectional prevalence design.
The majority of the recent studies and earlier studies used cross-sectional designs. Some
studies took into account potentially confounding covariates detailed in the Supplemental
Table 5S-14 (U.S. EPA. 2017c). Neophytou et al. (2016) controlled for age, height, and
calendar time, allowing for nonlinear effects, indicator variables for sex, race/ethnicity,
and continuous variables for SES (composite score variable), and numbers of smokers in
the household and also assessed effect modification by sex, obesity, SES, atopy, and
parental asthma. The designs used in most of the recent studies (i.e., ecological,
cross-sectional, single measure) limit the possible inferences about the relationship
between long-term SO2 exposure and lung function. The evidence does not include
studies evaluating concentration-responses. The one study conducting a copollutant
analysis found attenuation of the effect with adjustment for PM10. Thus, recent studies
do not add information that changes conclusions made in the 2008 SOx ISA (U.S.
EPA, 2008d) that there is not clear evidence that long-term SO2 exposure is related
to lung function changes.
Animal Toxicological Studies
A single long-term study with SO2 exposure concentrations at or below 2 ppm was
discussed in the 2008 SOx ISA (U.S. EPA. 2008d). Study characteristics are summarized
in Table 5-23. Smith et al. (1989) found that rats exposed to 1 ppm SO2 for 4 months had
decreased residual volume and quasi-static compliance when treated with saline (control).
Rats treated with elastase (a model of emphysema) and exposed to 1 ppm SO2 for
4 months had a decreased ratio of residual volume to total lung capacity and decreased
5-171
-------
alveolar plateau of the single-breath nitrogen (N2) washout (N2-slope), indicating a
worsening of the emphysema. However, Smith et al. (1989) concluded that the effects of
SO2 on lung function measurements were very minor in the saline (control) group and
likely due to chance alone (residual volume) or to unusually high control values
(quasi-static compliance).
Summary of Lung Function
Several studies evaluated the relationship between long-term SO2 exposure and
decrements in lung function. Evidence supporting this relationship is limited because
associations were inconsistent and because both PM and SO2 were at high concentrations
in the same areas, which does not allow determination of individual SO2 effects. Potential
confounding of long-term SO2 exposure-related decrements in lung function and lung
development by other pollutants, especially PM, was evaluated in only one study. This
study found an attenuation of the effect in copollutant analyses. No changes in lung
function were found in long-term animal toxicological studies at relevant SO2
concentrations. The recent studies support conclusions made in the 2008 SOx ISA (U.S.
EPA. 2008d) that evidence does not strongly support an effect of long-term SO2 exposure
on decreases in lung function in children.
5.2.2.4 Respiratory Infection
Epidemiologic Studies
Studies have also examined the association of long-term exposure to SO2 with infant
bronchiolitis, otitis media, and pneumonia in children, hospital admission for
community-acquired pneumonia in adults aged 65 years or more, and tuberculosis in
adults. Infant bronchiolitis was examined in British Columbia by Karr et al. (2009).
These study authors observed an association with lifetime exposure to SO2 after
adjustment for an array of confounders [Supplemental Table 5S-13 (U.S. EPA. 2017cVI.
The largest associations were observed with NO2 and CO concentrations. Maclntyre et al.
(2011) found no increased risk for otitis media in relation to long-term SO2 exposure in a
study of children up to the age of 2 years in British Columbia, while Bhattacharwa and
Shapiro (2010) found a strong relationship with long-term SO2 exposure in the U.S.
National Health Interview Survey of 126,060 children ages 3-6 years. Lu et al. (2014)
observed that the prevalence of pneumonia in children 3 to 6 years old was related to
long-term SO2 exposure. Liu et al. (2016) reported that doctor-diagnosed pneumonia in
children 4-6 years old was related to SO2 exposure during the first year of life. Neupane
5-172
-------
et al. (2010) estimated long-term SO2 exposure at the residence (2-yr avg for 2001-2002)
for both the case and control subjects using both bicubic spline and inverse distance
weighting methods, obtaining means of 4.65 and 5.80 ppb, respectively, but with a
twofold greater range for the bicubic spline method. Adjusted estimates of associations
for SO2 with hospitalization from community-acquired pneumonia were positive for the
bicubic spline method but not for inverse distance weighting. The incidence of
tuberculosis was associated with an increase of SO2 in adult males (Hwang et al.. 2014)
but not in a study in California (Smith et al.. 2016). Although limited in number, by
inconsistency, and by their cross-sectional design, these studies suggest a potential
relationship between long-term exposure to SO2 and respiratory infections due to various
infectious agents.
Animal Toxicological Studies
No new animal studies of the effects of long-term SO2 exposure on lung host defense
have been conducted since the previous review. Several studies of short- and long-term
exposure to SO2 were reported in the 1982 AQCD (U.S. EPA. 1982a) and discussed in
the 2008 SOx ISA (U.S. EPA. 2008d). Short-term exposure studies found some effects of
0.1-1 ppm SO2 on the clearance of labeled particles. Long-term exposure studies found
decreased tracheal mucus flow at a concentration of 1 ppm SO2, but no effects on
susceptibility to bacterial infection or alterations in the pulmonary immune system at
concentrations of 2 ppm or less.
Summary of Respiratory Infection
Evidence for prevalence of infant bronchiolitis and/or respiratory infections consists of
generally positive associations found in cross-sectional studies. Thus, they provide a
limited evidence base in number and design. While some animal toxicological studies
reported alterations in specific host defense mechanisms, there is no evidence to support
increases in bacterial or viral infections in animals exposed to SO2 at relevant
concentrations.
5.2.2.5 Development of Other Respiratory Diseases: Chronic Bronchitis, Chronic
Obstructive Pulmonary Disease, and Acute Respiratory Distress Syndrome
Chronic bronchitis consists of symptoms, including daily cough and/or congestion or
phlegm for 3 months in a row. While these symptoms may have started with acute
exacerbation, they are likely to represent chronic indolent symptoms. As discussed in the
2008 SOx ISA (U.S. EPA. 2008d). earlier cross-sectional studies observed positive
5-173
-------
relationships between long-term SO2 exposure estimates derived from fixed-site monitors
and chronic bronchitis as presented in Supplemental Table 5S-13 (U.S. EPA. 2017c).
Recent cross-sectional studies of the association of long-term exposure to SO2 with the
prevalence of bronchitis also observed positive relationships after adjustment for
potential confounders. In addition, a recent COPD incidence study in a national English
cohort (Atkinson et al.. 2015). discussed in Supplemental Table 5S-13 (U.S. EPA.
2017c). reported a positive association in an adjusted HR model with SO2 exposure
averaged over 3 years determined by dispersion models. Assessment of model validity
using national network sites and separate verification sites yielded poor If values for SO2
of 0 and 0.39, respectively. Other limitations of this study include a short follow-up time
and the failure to confirm the 36% of incident hospital admissions for COPD by a general
practitioner diagnosis.
A relationship between Acute Respiratory Distress Syndrome (ARDS) and long-term SO2
exposure has recently been studied (Ware et al.. 1986) as discussed in Supplementary
Table 5S-13 (U.S. EPA. 2017c). SO2 and PM2 5 were not associated with ARDS.
5.2.2.6 Respiratory Mortality
Recent studies provide some evidence that respiratory mortality may be more
consistently associated with long-term exposure to SO2 than other cause-specific
mortality outcomes (Section 5.5.2 and Figure 5-27). There is uncertainty in the small,
positive associations between long-term exposure to SO2 and respiratory mortality
observed in these studies because the exposure assessment and statistical methods are not
adequate for studying a highly spatially and temporally heterogeneous pollutant like SO2.
Additionally, there is little evidence of respiratory health effects in adults in relation to
long-term SO2 exposure that could provide coherence with the observed associations with
respiratory mortalities.
5.2.2.7 Summary and Causal Determination
Overall, the evidence is suggestive of, but not sufficient to infer, a causal relationship
between long-term SO2 exposure and respiratory effects, based on the evidence
across disciplines for development of asthma in children. This conclusion represents a
change from "inadequate to infer a causal association" for respiratory effects as stated in
the 2008 SOx ISA (U.S. EPA. 2008d). Evidence for development of other respiratory
diseases, lung function decrements, and respiratory mortality is relatively weak and does
not substantially influence the causal determination.
5-174
-------
The change in causal determination is based on evidence from recent epidemiologic
studies and support from animal toxicological studies. A limited number of longitudinal
epidemiologic studies report associations between asthma incidence among children and
long-term SO2 exposures. The longitudinal studies help to reduce the uncertainty
associated with temporality that is inherent in the previous cross-sectional studies. The
evidence from longitudinal studies is coherent with animal toxicological evidence of
allergic sensitization, airway remodeling, and increased airway responsiveness, which are
key events (or endpoints) in the proposed mode of action for the development of asthma.
The animal toxicological evidence provides support for an independent effect of SO2 and
a possible relationship between long-term exposure to SO2 and the development of
asthma in children. Some evidence of a link between long-term exposure to SO2 and
respiratory symptoms and/or respiratory allergies among children further supports this
relationship. The potential for SO2 to serve as an indicator for other pollutants or mixture
related to PM is an uncertainty that applies to the new body of epidemiologic evidence
across the respiratory effects examined.
The key evidence supporting the causal determination is detailed below using the
framework described in Table I of the Preamble to the ISAs (U.S. EPA. 2015b) and is
presented in Table 5-24.
5-175
-------
Table 5-24 Summary of evidence for suggestive of, but not sufficient to infer, a
causal relationship between long-term sulfur dioxide exposure and
respiratory effects.
so2
Concentrations
Rationale for Causal Associated
Determination3 Key Evidence13 Key References'3 with Effects0
Development and severity of asthma
Evidence from epidemiologic Evidence for increases in Nishimura et al. (2013) Mean (SD)
studies is generally supportive asthma incidence in cohorts of Clark etal (2010) across five cities
but not entirely consistent children in U.S. and Canada. 4.0 (3.4) ppb
Adequate adjustment for 1 gg (g 97) ppb
confounding by asthma risk
factors. Some inconsistency
regarding time window
Supporting cross-sectional Section 5.2.2.1
studies of asthma prevalence
among children but uncertainty
regarding the temporal
sequence between exposure
and the development of asthma
Supporting evidence for Sections 5.2.2.1 and 5.2.2.2
respiratory symptoms and
markers of respiratory allergies
among children in
cross-sectional studies
Supporting evidence from Section 5.2.2.1
intervention studies and natural
experiments
Evidence for increases in Section 5.2.2.1
asthma severity as indicated by
asthma severity score, degree
of asthma control, and AHR
Uncertainty regarding potential
for measurement error in
exposure estimates
SO2 concentrations assigned to
subjects based on combining
monitoring data by IDW in
asthma incidence studies and
monitors in cross-sectional
studies may not adequately
represent exposure
Nishimura et al. (2013)
Clark etal. (2010)
Section 3.5
Uncertainty regarding potential
confounding by copollutants
No copollutant models or
correlations analyzed in
asthma incidence studies. In
cross-sectional study, SO2
association persists with PM10
adjustment, not NO2. rwith SO2
across lifestages: 0.60-0.83.
Table 5-22 and Deng et al.
(2015a)
Section 3.4.3
5-176
-------
Table 5-24 (Continued): Summary of evidence for suggestive of, but not sufficient
to infer, a causal relationship between long term sulfur
dioxide exposure and respiratory effects
SO2
Concentrations
Rationale for Causal
Associated
Determination3
Key Evidence13
Key References'3
with Effects0
Limited animal toxicological
Th2 polarization (or other
Sona etal. (2012)
2,000 ppb
evidence provides coherence
Type 2 immune responses)
and biological plausibility
and airway inflammation
following repeated exposure of
naive newborn rats for 28 days
Evidence for enhanced
inflammation, airway
remodeling, and increased
airway responsiveness
following repeated exposure of
allergic newborn rats for
28 days
Coherence with evidence from
Inflammation and morphologic
Li et al. (2007)
2,000 ppb
short-term animal toxicological
responses indicative of airway
Li etal. (2014)
studies
remodeling following repeated
exposures of naive rats over
several days
Enhancement of allergic
Riedel et al. (1988)
100 ppb
sensitization, allergic
Park etal. (2001)
100 ppb
inflammation, and bronchial
obstruction in guinea pigs
exposed repeatedly over
several days and subsequently
sensitized and challenged with
an allergen
Enhanced inflammation and
Li et al. (2007)
2,000 ppb
allergic responses in rats
Li etal. (2014)
previously sensitized with an
allergen and then repeatedly
exposed
Some evidence for key events
Inflammation, allergic
Section 4.3.6
in proposed mode of action
sensitization, AHR, airway
remodeling
Development of allergy
Limited epidemiologic evidence
Generally positive associations
Section 5.2.2.2
but uncertainty regarding SO2
with different markers for
independent effects
allergies in cross-sectional
studies in children. Uncertainty
in temporality and exposures
estimated from fixed-site
monitors; copollutant
confounding examined on a
limited basis remains uncertain
5-177
-------
Table 5-24 (Continued): Summary of evidence for suggestive of, but not sufficient
to infer, a causal relationship between long term sulfur
dioxide exposure and respiratory effects
Rationale for Causal
Determination3
Key Evidence13
Key References'3
SO2
Concentrations
Associated
with Effects0
Lung function
Inconsistent epidemiologic
evidence among children from
quality studies and uncertainty
regarding SO2 independent
effects
In cohort studies, associations Neophvtou et al. (2016)
inconsistent with adjustment for Jedrvchowski etal. (1999)
Frischer et al. (1999)
PM and by season
Inconsistent results from
cross-sectional studies
Dockerv et al. (1989)
Schwartz (1989)
Ackermann-Liebrich et al.
(1997)
Frve et al. (2003)
Respiratory infection
Limited epidemiologic Generally positive associations Section 5.2.2.4
evidence; uncertainty regarding in cross-sectional studies.
SO2 independent effects Uncertainty in temporality,
exposures estimated from
monitors in the community, and
copollutant confounding
Limited animal toxicological
Altered clearance of particles U.S. EPA (1982a)
0.1-1 ppm
evidence
and decreased tracheal mucus
flow
Lack of evidence for key events
Changes in specific host
in proposed mode of action
defense mechanisms but no
evidence of greater infectivity
Development of other respiratory diseases
Limited epidemiologic evidence Generally positive associations Section 5.2.2.5
but uncertainty regarding SO2 for chronic bronchitis in
independent effects cross-sectional studies.
Uncertainty in temporality,
exposures estimated from
monitors in the community, and
copollutant confounding
5-178
-------
Table 5-24 (Continued): Summary of evidence for suggestive of, but not sufficient
to infer, a causal relationship between long term sulfur
dioxide exposure and respiratory effects
so2
Concentrations
Associated
Key References'3 with Effects0
Respiratory mortality
Rationale for Causal
Determination3 Key Evidence13
Generally consistent Small, positive associations Hart et al. (2011). Nafstad et al. 2.4-41.4
epidemiologic evidence between long-term exposure to (2004). Elliott et al. (2007). Cao
SO2 and respiratory mortality in et al. (2011), Carey et al.
several cohorts, even after (2013). Dong et al. (2012),
adjustment for common Katanoda etal. (2011)
potential confounders
No coherence between No evidence for a relationship Section 5.2.2.6
respiratory morbidity and between long-term exposure
respiratory mortality and respiratory mortality to
support the observed
associations with respiratory
morbidity
AHR = airway hyperresponsiveness; IDW = inverse distance weighting; N02 = nitrogen dioxide; PM = particulate matter;
PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm; r= correlation coefficient;
SD = standard deviation; S02 = sulfur dioxide Th2 = T-derived lymphocyte helper 2.
aBased on aspects considered in judgments of causality and weight of evidence in causal framework in Tables I and II of the
Preamble to the ISAs (U.S. EPA. 2015b).
bDescribes the key evidence and references, supporting or contradicting, contributing most heavily to causal determination, and
where applicable, to uncertainties or inconsistencies. References to earlier sections indicate where the full body of evidence is
described.
°Describes the S02 concentrations with which the evidence is substantiated (for experimental studies, <2,000 ppb).
Evidence for the Development of Asthma
A limited number of longitudinal studies demonstrate associations between ambient SO2
concentrations measured in the first year of life and/or over the first 3 years of life in
children and asthma incidence such as Clark et al. (2010) and Nishimura et al. (2013)
(Section 5.5.2.1). Results are fairly consistent between studies with one based on several
different locations across the U.S., another over a large area in Canada, and one in
Taiwan, involving a large number of participants. Uncertainties and the potential for
exposure measurement error related to the use of inverse distance weighting and area
comparisons in these studies may limit inferences that can be made (Section 3.3.2.3).
Additional supportive evidence for a link between long-term SO2 exposure and the
development of asthma is provided by cross-sectional studies of asthma prevalence,
respiratory symptoms, and markers of respiratory allergies among children
(Section 5.2.2.2). Findings of studies evaluating respiratory symptoms are supportive of
the development of asthma; however, they may also reflect other respiratory conditions.
5-179
-------
Intervention and natural experiment studies also indicate a possible relationship between
long-term exposure to SO2 and the development of asthma.
Epidemiologic studies of asthma development in children have not clearly characterized
potential confounding by other pollutants or mixtures of pollutants. This uncertainty was
present in the previous review, and there is no new information from incidence studies to
help reduce this uncertainty. No studies of asthma incidence have evaluated copollutant
models to address copollutant confounding, making it difficult to evaluate the
independent effect of SO2 within the epidemiologic evidence base for incidence. A recent
cross-sectional study of asthma prevalence provides inconsistent results from copollutant
models, persistence of the SO2 association with PM10 not NO2. In studies that examined
both SO2 and PM2 5, positive associations were observed between PM2 5 concentrations
and asthma development; the effects were similar in magnitude to those for SO2
(Nishimura et al.. 2013; Clark et al.. 2010). Correlations between SO2 and PM2 5 were not
reported in these studies. Thus, results from these two studies do not reduce the
uncertainty related to potential copollutant confounding.
The uncertainties in the epidemiologic evidence base is reduced, in part, by the biological
plausibility provided by findings from experimental studies that demonstrate
SCh-induced effects on key events or endpoints that are part of the proposed mode of
action for the development of asthma [i.e., allergic sensitization, airway remodeling and
increased airway responsiveness (Section 4.3.6)1. An experimental study in newborn rats,
which were not previously sensitized and challenged with an allergen (i.e., naive
animals), found that repeated acute SO2 exposures over several weeks led to airway
inflammation and Th2 polarization (or other Type 2 immune responses), important steps
in allergic sensitization IYSong et al.. 2012); see Section 5.2.2.11. Repeated SO2 exposure
in the newborn rats, which were previously sensitized and challenged with an allergen
(i.e., allergic animals), resulted in enhanced allergic airway inflammation and some
evidence of airway remodeling and increased airway responsiveness. Additional evidence
comes from experimental studies in adult animals involving short-term exposure to SO2
over several days. In naive rats, airway inflammation and morphologic responses
indicative of airway remodeling were seen (Section 5.2.1.7). Furthermore, enhancement
of allergic sensitization and other inflammatory responses were observed along with
evidence for increased airway responsiveness in guinea pigs exposed repeatedly to SO2
for several days and subsequently sensitized and challenged with an allergen
(Section 5.2.1.7). Similarly, SO2 exposure enhanced airway inflammation in rats
previously sensitized with an allergen (Section 5.2.1.2).
5-180
-------
Evidence for the Severity of Asthma
A few studies provide evidence for a potential relationship between long-term SO2
exposure and the severity of asthma, as indicated by asthma severity scores, asthma
control, and AHR (Section 5.2.2.1V
Evidence for the Development of Allergies
Epidemiologic evidence from a few long-term studies provides a link between long-term
SO2 exposure and respiratory allergies and allergic rhinitis among children
(Section 5.2.2.2). However, uncertainties remain given the cross-sectional design of these
studies. Two-pollutant models have begun to address the role of SO2 exposure in the
development of allergic rhinitis.
Evidence for Lung Function
Several studies evaluated the relationship between long-term SO2 exposure and
decrements in lung function (Section 5.2.2.3V Evidence supporting this relationship is
limited because associations were inconsistent and because both PM and SO2 were at
high concentrations in the same areas, precluding determination of individual SO2 effects.
Potential confounding of long-term SO2 exposure-related decrements in lung function
and lung development by other pollutants, especially PM, was evaluated in only one
study. This study found an attenuation of the effect in two-pollutant analyses. No changes
in lung function were found in long-term animal toxicological studies at relevant SO2
concentrations. The recent studies support conclusions made in the 2008 SOx ISA (U.S.
EPA. 2008d') that the available evidence was inadequate to infer a causal relationship
between long-term exposure to SO2 at ambient concentrations and changes in lung
function.
Evidence for Respiratory Infection
Respiratory infection related to long-term SO2 exposure is discussed in Section 5.2.2.4. A
limited number of the cross-sectional studies examined indicate associations between
long-term SO2 exposure and bronchitis or respiratory infection due to various infectious
agents; findings were generally positive. While some animal toxicological studies
reported alterations in specific host defense mechanisms, there is no evidence to support
increases in bacterial or viral infections in animals exposed to SO2 at relevant
concentrations.
5-181
-------
Evidence for the Development of Other Respiratory Diseases
Evidence for prevalence of bronchitis and/or COPD consists of generally positive
associations found in cross-sectional studies (Section 5.2.2.5).
Evidence for Respiratory Mortality
Small positive associations between long-term exposure to SO2 and respiratory mortality
among adults were found in several cohort studies after adjustment for common potential
confounders (Section 5.2.2.6V There is little evidence of respiratory health effects in
adults in relation to long-term SO2 exposure that could provide coherence with the
observed associations with respiratory mortality among adults.
Conclusion
Taken together, epidemiologic and animal toxicological studies provide evidence
that is suggestive of, but not sufficient to infer, a causal relationship between
long-term SO2 exposure and respiratory effects (see Table 5-24V The strongest
evidence is provided by coherence of findings of epidemiologic studies showing
associations between long-term SO2 exposure and increases in asthma incidence among
children and findings of animal toxicological studies that provide a pathophysiologic
basis for the development of asthma. These latter studies demonstrated that repeated SO2
exposure over several weeks resulted in Th2 polarization (or other Type 2 immune
responses) and airway inflammation, key steps in allergic sensitization, in naive newborn
animals. In addition, repeated SO2 exposure over several weeks resulted in enhanced
airway inflammation and some evidence of airway remodeling and increased airway
responsiveness in allergic newborn animals. Toxicological studies involving repeated
exposure to SO2 over several days provide additional evidence of these effects. However,
because the toxicological evidence in animals is limited, particularly for long-term
exposure, some uncertainty remains regarding an independent effect of long-term SO2
exposure on the development of asthma. In addition, potential confounding by other
pollutants is unexamined, and largely unavailable, for epidemiologic studies of asthma
among children. However, multiple lines of evidence suggest that long-term SO2
exposure results in a coherent and biologically plausible sequence of events that
culminates in the development of asthma, especially allergic asthma, in children.
5-182
-------
5.3
Cardiovascular Effects
5.3.1 Short-Term Exposure
5.3.1.1 Introduction
The 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d) reviewed studies published through
2006 and concluded that "the evidence as a whole is inadequate to infer a causal
relationship" between short-term exposure to SO2 and cardiovascular health effects.
Specifically, the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d') found a lack of
consistency with regard to short-term exposure to SO2 and markers of heart rate
variability (HRV), cardiac repolarization, discharges of implantable cardioverter
defibrillators (ICDs), blood pressure (BP), blood markers of cardiovascular disease risk,
the triggering of a myocardial infarction, or ED visits or hospital admission for
cardiovascular diseases. Recent epidemiologic studies add to the evidence for effects of
SO2 exposure on a broader array of cardiovascular effects and mortality. Still, substantial
uncertainties remain concerning exposure measurement error, the limited mechanistic
evidence to describe a role for SO2 in the initiation of key events in a proposed mode of
action, and potential confounding by copollutants.
This section reviews the published studies pertaining to the cardiovascular effects of
short-term exposure (i.e., up to 1 month) to SO2 in humans and animals. Emphasis has
been placed on studies published since the 2008 ISA for Sulfur Oxides, with the existing
body of evidence serving as the foundation. The majority of the recent evidence is from
epidemiologic studies, which examined the association of SO2 exposure with MI,
cerebrovascular disease, and other cardiovascular effects. With few exceptions, most
epidemiologic studies model the association of 24-h avg SO2 concentration with
cardiovascular outcomes. There are no toxicological studies evaluating cardiovascular
effects following 5-10 minute exposures to SO2.
The previous ISA included a small number of animal toxicological studies of blood
pressure (Section 5.3.1.5). HR and HRV (Section 5.3.1.10). arrhythmia frequency
(Section 5.3.1.3) and controlled human exposure studies that examined effects on the
autonomic nervous system (Section 5.3.1.10) from short-term exposure to SO2. Since the
2008 ISA for Sulfur Oxides (U.S. EPA. 2008d). no controlled human exposure studies
and few animal toxicological studies have investigated the effects of short-term SO2
exposure on the cardiovascular system. Results from the experimental studies included in
the past and current reviews that evaluated cardiovascular effects of short-term SO2
5-183
-------
exposures of less than 2,000 ppb are summarized in the relevant outcome section and
additional study details are summarized in Supplemental Table 5S-15 (U.S. EPA. 2017c).
Studies examining cardiovascular effects of sulfite exposure (via intraperitoneal injection,
intravenous injection, etc.) are not included in this section because these studies generally
involve exposures to sulfite that are higher than what is expected to occur following
inhalation of SO2 at ambient relevant concentrations. Some studies using prolonged
exposures to 300 ppb and higher concentrations of SO2 reported measurable changes in
the concentrations of sulfite/S-sulfonate in plasma and tissues. A positive correlation was
found between the concentration of inhaled SO2 and plasma sulfite/S-sulfonate levels in
humans exposed continuously to SO2 (300-6,000 ppb) (Gunnison and Palmes. 1974).
Similarly, a recent report in mice exposed to 5,000-20,000 ppb SO2 for 7 days found a
concentration-dependent increase in sulfite/S-sulfonate levels in lung, heart, and brain
compared to controls (Meng et al.. 2005b'). These studies suggest that prolonged exposure
to SO2 at concentrations higher than typically found in ambient air may increase
circulating sulfite, but these changes would be expected to be far less following ambient
exposures of shorter duration. The literature on the distribution and metabolism of sulfite
is discussed in Sections 4.2.3 and 4.2.4. The potential role of sulfite in the induction of
systemic effects, such as effects of the cardiovascular system, is discussed in
Section 4.3.4.
To clearly characterize the evidence underlying causality, the discussion of the evidence
is organized into groups of related outcomes [myocardial infarction and ischemic heart
disease (IHD) (Section 5.3.1.2). arrhythmia and cardiac arrest (Section 5.3.1.3).
cerebrovascular disease (Section 5.3.1.4). hypertension (Section 5.3.1.5). venous
thromboembolism (VTE) (Section 5.3.1.6). heart failure (Section 5.3.1.7). aggregated
cardiovascular disease (Section 5.3.1.8). and cardiovascular mortality (Section 5.3.1.9)1.
Evidence for subclinical effects (e.g., heart rate variability, blood biomarkers of
cardiovascular effects) of short-term exposure to SO2 that potentially underlie the
triggering or indication of various clinical events are discussed in Section 5.3.1.10. and
may provide biological plausibility for multiple outcomes.
5.3.1.2 Myocardial Infarction and Ischemic Heart Disease
Several lines of evidence are discussed in evaluating the relationship between short-term
SO2 exposure and MI. An MI, or heart attack, occurs as a consequence of IHD, resulting
in insufficient blood flow to the heart that overwhelms myocardial repair mechanisms
and leads to muscle tissue death. ICD codes for MI are classified within the group of
IHDs; thus, studies in which IHD is evaluated will include any patients diagnosed with an
5-184
-------
MI. Finally, acute MI may be characterized by ST-segment depression, a nonspecific
marker of myocardial ischemia. The evaluation of evidence supporting a relationship
between short-term SO2 exposure and the triggering of an MI includes hospitalization and
ED visits for MI or IHD and ST-segment amplitude changes.
The epidemiologic data available for review by the 2008 ISA for Sulfur Oxides (U.S.
EPA. 2008d') did not indicate an association between SO2 and risk of MI. A number of
additional studies based on administrative data of hospital admissions or ED visits or on
clinical data are now available in Figure 5-12. The air quality characteristics of the city,
or across all cities, and the exposure assignment approach used in each Mi-related
hospital admission and ED visit study evaluated in this section are presented in
Table 5-25. The recent clinical registry studies provide inconsistent evidence for an
association between MI and ambient SO2, while multicity and single-city hospital
admission and ED visit studies provide generally consistent evidence of an association.
However, potential copollutant confounding and limited mechanistic evidence are still
key uncertainties that make it difficult to interpret the results of these studies.
Additionally, most studies examined 24-h avg exposure metrics for SO2, which may not
adequately capture the spatial and temporal variability in SO2 concentrations
(Section 3.4.2).
5-185
-------
Study Outcome
fBhaskaran et al. (2011) Ml
fMilojevic et al. (2014) Ml
"(Turin et al. (2012) Ml
Ballesteretal. (2006) IHD
tRich et al. (2010) Ml
fCheng etal. (2009) Ml
tHsieh etal. (2010) Ml
fSteib et al. (2009) Ml/Angina
Cendon et al. (2006) Ml
tThach etal. (2010) IHD
tTsai etal. (2013) Ml
tQui etal. (2013) IHD
tTam et al. (2015) IHD
fBell et al. (2008) IHD
Lag Notes
1-6 h
0-4
1
0-1
0
0-2 > 25° C
0-2 < 25° C
0-2 2 23° C
0-2 < 23° C
1
0-7
0-1
0-2 Warm Days
0-2 Cool Days
0-3 All Year
0-3 Warm Days
0-3 Cool Days
0-3
0-3 All Monitors
City Monitors
Correlated Monitors
0.5
i
f
i—
1 1
1 1.5
Risk or Odds Ratio (95% CI)
CI = confidence interval; IHD = ischemic heart disease; Ml = myocardial infarction.
Note: Studies in red are recent studies. Studies in black were included in the 2008 ISA for Sulfur Oxides. All-year
associations = circles; summer/warm-days associations = diamonds; winter/cold-days associations = squares. Relative risks are
standardized to a 10-ppb or 40-ppb increase in sulfur dioxide for 24-hour avg and 1-hour max metrics, respectively. Lag times are
reported in days, unless otherwise noted. Corresponding quantitative results are reported in Supplemental Table 5S-16 (U.S. EPA.
2017c). All results are from single pollutant models.
Figure 5-12 Results of studies of short-term sulfur dioxide exposure and
hospital admissions for ischemic heart disease.
5-186
-------
Table 5-25 Mean and upper percentile concentrations of sulfur dioxide from
ischemic heart disease hospital admission and emergency
department visit studies.
Mean/Median Upper Percentile
Location Exposure Concentration of Concentrations
Study Years Assignment Metric ppb ppb
tBhaskaran et al.
(2011)
15 conurbations
in England and
Wales
(2003-2006)
Fixed-site monitor 1-h max
from each
conurbation
(aggregated when
more than one
monitor)
Mean: 1.9
75th: 3.4
tMiloievic et al.
(2014)
230 acute
hospitals in
England and
Wales
(2003-2009)
Nearest monitor
within 50-km
distance from
residence location
24-h avg
Median: 1.2
75th: 2.3
tTurin et al. (2012)
Takashima
County, Japan
(1988-2004)
Nearest monitor to
Takashima County
(20 km)
24-h avg
Mean: 3.9
75th: 4.8
Ballester et al.
(2006)
14 Spanish
cities
(1995-1999)
Citywide average
for each city
24-h avg
Mean: 2.9-15.6
across cities
90th: 4.8-28.8
across cities
tRich et al. (2010)
New Jersey
(2004-2006)
Closest of
14 monitor (those
>10 km from
monitor excluded)
24-h avg
NR
NR
tChenq et al.
(2009)
Kaohsiung,
Taiwan
(1996-2006)
Average across six
monitoring stations
24-h avg
Mean: 9.33
75th: 11.69
Max: 31.26
tHsieh et al. (2010)
Taipei, Taiwan
(1996-2006)
Average across six
monitoring stations
24-h avg
Mean: 4.36
75th: 5.48
Max: 17.82
tStieb et al. (2009)
Seven Canadian
cities
(1992-2003)
Citywide average
for each city
24-h avg
Mean: 2.6-10.0
across cities
75th: 3.3-13.4
across cities
Cendon et al.
(2006)
Sao Paulo,
Brazil
(1998-1999)
Average across
13 monitoring
stations
24-h avg
Mean: 5.6
95th: 12.1
tThach et al. (2010)
Hong Kong,
China
(1996-2002)
Average across
eight monitoring
stations
24-h avg
Mean: 6.8
NR
tTsaietal. (2012)
Taipei, Taiwan
(1999-2009)
Average across six
monitoring stations
24-h avg
Mean: 3.94
75th: 5.01
Max: 12.7
5-187
-------
Table 5-25 (Continued): Mean and upper percentile concentrations of sulfur
dioxide from ischemic heart disease hospital admission
and emergency department visit studies
Study
Location
Years
Exposure
Assignment
Metric
Mean/Median
Concentration
PPb
Upper Percentile
of Concentrations
PPb
+Qiu et al. (2013a)
Hong Kong,
China
(1998, 2007)
Average across
14 monitoring
stations
24-h avg
Mean: 7.4
NR
tSan Tam et al.
(2015)
Hong Kong,
China
(2001-2010)
Average across
13 monitoring
stations
24-h avg
Mean: 7.6
75th: 9.3
Max: 51.9
tBell et al. (2008)
Taipei, Taiwan
(1995-2002)
Average across
13 monitoring
stations; 5 within
city limits; or 6 with
correlations >0.75
24-h avg
Mean: 4.7
Max: 26.9
Avg = average; max = maximum; NR = not reported.
fStudies published since the 2008 ISA for Sulfur Oxides.
Some studies rely on clinical registries, which are generally less susceptible to
misclassification of the outcome. Using data from the Myocardial Ischemia National
Audit Project (MINAP) clinical registry, Bhaskaran et al. (2011) reported that hourly
ambient SO2 concentrations were not associated with risk of MI in a case-crossover study
of 15 conurbations in England and Wales between 2003 and 2006. While no associations
were reported in the population overall, there was some evidence of an association in
subgroup analyses within older age groups (60-69, 70-79, and 80+) at inconsistent lag
times. This study is unique because it included detailed data on the timing of MI onset in
more than 79,000 patients, which allowed examination of the association with ambient
SO2 in the hours preceding MI. Miloievic et al. (2014) also used data from MINAP, from
2003 to 2009, and observed stronger evidence of an association between SO2
concentrations and MI [4.3% (95% CI: -0.25, 8.8%) increase in risk of MI per 10-ppb
increase in 24-h avg SO2 at lag 0-4], Turin et al. (2012) did not observe any association
using data from the Takashima County Stroke and Acute Myocardial Infarction Registry
in central Japan, although this study was likely underpowered to detect an association of
the expected magnitude. None of the clinical registry studies examined copollutant
models.
One prominent study from the previous 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d)
was conducted in 14 cities across Spain and found a 4.5% (95% CI: 1.3, 8.1%) increase
in hospital admissions per 10-ppb shift in SO2 for the composite endpoint of IHD,
arrhythmias, and heart failure (Ballester et al.. 2006). This association was still positive,
but attenuated and no longer statistically significant after adjustment for CO or NO2. It
5-188
-------
was lessened in magnitude, but more precise (i.e., narrowed confidence intervals), with
adjustment for TSP or O3 in copollutant models (no quantitative results; results presented
graphically). Several additional ED visit and hospital admission studies are now
available. In a study of hospitalization in New Jersey, Rich et al. (2010) did not report
strong evidence for an association between SO2 and risk of hospital admissions for MI
[OR: 1.05 (95% CI: 0.84, 1.29) per 10-ppb increase in 24-h avg SO2 on the same day].
The inclusion of PM2 5 in a copollutant model did not reveal a positive association for SO2
[OR: 0.91 (95% CI: 0.69, 1.21)]. In Kaohsiung, Taiwan, Cheng et al. (2009) reported an
association between SO2 concentrations and hospital admissions for MI, but only on days
when the mean ambient temperature was <25°C. However, in copollutant models
adjusting for PM10, NO2, or CO, SO2 was no longer associated with increased admissions.
Conversely, in Taipei, Taiwan, Hsieh et al. (2010) only observed an association between
SO2 and MI on warm days (>23°C). Similar to the findings of Cheng et al. (2009). this
association was no longer positive after adjustment for PM10, NO2, O3, or CO in
copollutant models. Most other studies have not considered copollutant models.
A study using data from 14 hospitals in seven Canadian cities found a 4.2% (95% CI: 0.4,
8.0%) increase in risk of ED visits for the composite endpoint of acute MI or angina per
10-ppb increase in SO2 on the previous day (Stieb et al.. 2009). Most (San Tam et al..
2015; Qiu et al.. 2013a; Tsai et al.. 2012; Thach et al.. 2010; Cendon et al.. 2006; Martins
et al.. 2006) but not all (Bell et al.. 2008) studies using data from individual cities have
found associations between SO2 concentrations and risk of hospital admissions or ED
visits for ischemic heart disease or MI. None of the single-city studies evaluated potential
copollutant confounding, and all of the studies in this section used fixed-site monitors to
measure ambient SO2. The limitations of these monitors in capturing spatial variation in
SO2 has been noted previously (Section 3.4.2).
ST-Segment Changes
ST-segment changes (either ST-segment elevation or depression) on the
electrocardiogram are considered a nonspecific marker of myocardial ischemia. While
the 2008 ISA for Sulfur Oxides did not review any epidemiologic studies of ambient SO2
concentrations and markers of myocardial ischemia, one subsequent study reported an
association. Chuang et al. (2008) conducted a repeated-measures study in adults with a
history of coronary heart disease (CHD) and examined the association between ambient
pollutants and ST-segment level changes. This study found an odds ratio of 3.0 (95% CI:
1.8, 5.5) for ST-segment depression of>0.1 mm per 10-ppb increase in SO2 over the
previous 24 hours. This finding was generally unchanged after additional control for
PM2 5 and BC in copollutant models.
5-189
-------
Summary of Myocardial Infarction and Ischemic Heart Disease
In summary, while evidence from epidemiologic studies suggests a potential association
between ambient SO2 concentrations and rates of hospital admissions or ED visits for MI
or ischemic heart diseases in single-pollutant models, these associations may be the result
of confounding by other pollutants. While three studies based on clinical data report
inconsistent evidence regarding associations between ambient SO2 concentrations and
risk of MI, the majority of studies relying on MI hospital admission and ED visit data
observed either seasonal or year-round associations with SO2. However, some of these
associations were either attenuated or no longer present after controlling for potential
copollutant confounding (Hsieh et al.. 2010; Cheng et al.. 2009; Ballester et al.. 2006).
leaving uncertainties regarding the independent effect of short-term SO2 exposure. In
congruence with the evidence from hospital admission and ED visit studies, there was
limited evidence from a single study indicating that SO2 may be associated with
ST-segment changes on the electrocardiogram in patients with a history of coronary heart
disease. Most studies examined 24-h avg exposure metrics for SO2, which may not
adequately capture the spatial and temporal variability in SO2 concentrations
(Section 5.2.1.2). No experimental studies have been conducted to evaluate measures of
ischemic heart disease or MI following short-term SO2 exposure. Overall, despite some
epidemiologic evidence of an association between short-term exposure to SO2 and
hospital admissions and ED visits for ischemic heart disease and MI, uncertainties
regarding copollutant confounding continue to impede the determination of an
independent SO2 effect.
5.3.1.3 Arrhythmias and Cardiac Arrest
The 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d) concluded that the evidence available
at the time did not suggest that SO2 has an effect on cardiac arrhythmias. There continues
to be essentially no epidemiologic or toxicological evidence suggestive of such a
relationship.
Metzger et al. (2007) examined 518 patients with ICDs with 6,287 tachyarrhythmic
event-days over a 10-year period in Atlanta, GA and found no association between SO2
concentrations and the risk of tachyarrhythmias, either overall or in analyses limited to
more severe tachyarrhythmic events, or stratified by season or the presence of a recent
past arrhythmic event (results for this study and other studies in this section can be found
in Table 5-26). A similar study in London, England also found limited evidence of an
association between SO2 concentrations and arrhythmic risk (Anderson et al.. 2010).
Anderson etal. (2010) reported an increase in risk of ICD activations corresponding to an
5-190
-------
increase in ambient SO2, but the confidence interval was wide [OR: 1.35 (95% CI: 0.75,
2.41) per 10-ppb increase in SO2 at lag days 0-1], Similarly, a study in Boston, MA of
ICD activations was even less precise, offering no evidence of an association [32.0%
(95% CI: -48.5, 336.2%) increase in ICD activations per 10-ppb increase in SO2
concentrations at lag 1] (Link et al.. 2013). Additionally, a multicity study in Canada
(Stieb et al.. 2009) and a large single-city study in Taipei, Taiwan (Tsai et al.. 2009) have
reported finding no association between SO2 and ED visits for arrhythmias, while a large
single-city study in Shanghai, China reported a positive association that was attenuated
and no longer positive in a copollutant model adjusted for NO2 (Zhao et al.. 2014).
The majority of out-of-hospital cardiac arrests (OHCA) are due to cardiac arrhythmias.
Dennekamp et al. (2010) considered the association between ambient pollutants and
OHCA among 8,434 cases identified through the Victorian Cardiac Arrest Registry in
Melbourne, Australia and found null and/or imprecise associations (i.e., wide 95% CIs)
between SO2 concentrations and risk of OHCA. A similar approach was used by
Silverman etal. (2010) with data from 8,216 OHCAs in New York City. Quantitative
results for SO2 were not provided, but graphs showed a null association between OHCA
and year-round SO2 concentrations. Silverman et al. (2010) also presented
season-specific analyses graphically, demonstrating that out-of-hospital cardiac arrests
were positively but imprecisely (i.e., wide 95% CI) associated with SO2 concentrations
during the warm season. Two additional case-crossover studies of OHCA in Perth,
Australia (Stranev et al.. 2014) and Helsinki, Finland (Rosenthal et al.. 2013) observed
null associations with ambient SO2. In contrast, Kang et al. (2016) observed an
association between 24-h avg SO2 and OHCA in Seoul, South Korea at individual lag
days 0 through 3 (no quantitative results; results presented graphically).
One animal toxicological study (Nadzieiko et al.. 2004) evaluated arrhythmia frequency
in rats following short-term SO2 exposure and reported no significant changes in
spontaneous arrhythmias (irregular, delayed, or premature beats).
5-191
-------
Table 5-26 Epidemiologic studies of arrhythmia and cardiac arrest.
Study
Location and
Years
Sample Size
Mean and Upper
Concentration
SO2 (ppb)
Exposure
Assessment
Selected Effect Estimates3
95% CI
tMetzqer et al.
(2007)
Atlanta, GA
1993-2002
(n = 518)
1-h max: 15.5
90th percentile:
36
Max: 149
Central All tachyarrhythmic events (OR); year
monitor round
Lag 0: 1.00 (0.94, 1.08)
Warm season
Lag 0: 1.06 (0.98, 1.25)
Cold season
Lag 0: 0.97 (0.91, 1.05)
Cardiac pacing or defibrillation (OR): Lag
0: 0.98 (0.88, 1.09)
Defibrillation (OR):
Lag 0: 1.01 (0.98, 1.24)
tAnderson et al.
(2010)
London, U.K.
1995-2003
[n = 705
(5,462 device
activations)]
24-h avg: 1.03
75th percentile:
1.15
Max: 2.67
Citywide avg ICD activations (OR);
Lag 01: 1.35 (0.75, 2.41)
Lag 05: 1.71 (0.69, 4.27)
Correlations: PM10: 0.48, PM2.5: 0.42, BS:
0.35, SO42": 0.19, PNC: 0.29, NO2: 0.60,
NO: 0.44, NOx: 0.49, O3: -0.36
tLinketal. (2013)
Boston, MA
24-h avg: 3.2
Citywide avg
ICD activations (percent change);
2006-2010
75th percentile: 4
Lag 1: 32.0 (-48.5, 336.2)
[n = 176
Correlations: CO:-0.06 to 0.75, NO2: 0.05
(328 atrial
to 0.69, Os: -0.52 to -0.18, PM10: 0.27 to
fibrillation
0.55, PM2.5: 0.01 to 0.67
episodes
>30 sec)]
tStieb et al. (2009)
Seven
24-h avg: 2.6 to
Citywide avg
Dysrhythmia ED visits (percent change);
Canadian
10 across cities
for each city
Lag 0: -1.4 (-6.0, 3.4)
cities
75th percentile:
Lag 1: 0.8 (-6.4, 8.6)
1992-2003
3.3 to 13.4 across
Lag 2: -5.0 (-9.2, -0.6)
(n = 45,160 ED
visits)
cities
Correlations: PM10: 0.52, NO2: 0.43, CO:
0.24, Os: 0.09
tTsai et al. (2009)
Taipei, Taiwan
2000-2006
(n = 21,581
ED visits)
24-h avg: 3.93
75th percentile:
5.02
Max: 12.7
Citywide avg Arrhythmia ED visits (OR);
>23°C: 1.04 (0.88, 1.23)
<23°C: 1.04 (0.88, 1.27)
Correlations: PM10: 0.52, NO2: 0.43, CO:
0.24, Os: 0.09
tZhao et al. (2014)
Shanghai,
China
2010-2011
(n = 56,940
outpatient
visits)
24-h avg: 11.1
75th percentile:
14.1
Max: 49.6
Central Arrhythmia outpatient visits (percent
monitor change);
Lag 0: 1.06 (1.04, 1.07)
5-192
-------
Table 5-26 (Continued): Epidemiologic studies of arrhythmia and cardiac arrest.
Study
Location and
Years
Sample Size
Mean and Upper
Concentration
SO2 (ppb)
Exposure Selected Effect Estimates3
Assessment 95% CI
tDennekamp et al.
(2010)
Melbourne,
Australia
2003-2006
(n = 8,434
OHCA)
24-h avg: 0.49 Central OHCA (percent change);
75th percentile: monitor Lag 0:-10.0 (-40.3, 64.0)
0.76 Lag 1: 6.9 (-34.9, 75.6)
Lag 2: 0.8 (-39.0, 66.7)
Lag 01: -0.7 (-34.9, 75.6)
tSilverman et al.
New York City,
24-h avg: 6.3
Citywide avg No quantitative results; results presented
(2010)
NY
(median)
graphically. Null association between
2003-2006
75th percentile:
OHCA and year-round SO2
(n = 8,216
OHCA)
9.6
95th percentile:
concentrations. OHCA positively but
imprecisely (i.e., wide 95% CI) associated
18
with ambient SO2 during the warm season
tStranev et al.
(2014)
Perth,
Australia
2000-2010
(n = 8,551
OHCA)
1-h avg: 0.4
(median)
75th percentile:
0.9
95th: 3.5
Nearest OHCA (OR);
monitor Lag 0: 0.91 (0.71, 1.17)
tRosenthal et al.
(2013)
Helsinki,
Finland
1998-2006
(n = 2,134
OHCA)
24-h avg: 1.5
Citywide avg OHCA (OR);
Lag 0
Lag 1
Lag 2
Lag 3
0.93 (0.58, 1.44)
0.68 (0.42, 1.08)
1.08 (0.68, 1.66)
1.00 (0.63, 1.55)
Lag 03: 0.86 (0.42, 1.55)
tKanq et al. (2016)
Seoul,
South Korea
2006-2013
(n = 28,315
OHCA)
24-h avg: 2.1
75th percentile:
2.5
Max:
.1
No quantitative results; results presented
graphically. Positive, statistically
significant associations at single day lags
0 through 3. Null associations at lags 4
and 5.
Avg = average; BS = black smoke; CI = confidence interval; CO = carbon monoxide; ED = emergency department;
ICD = implantable cardioverter defibrillators; max = maximum; n = sample size; N = population number; NO = nitric oxide;
N02 = nitrogen dioxide; NOx = the sum of NO and N02; 03 = ozone; OHCA = out-of-hospital cardiac arrests; OR = odds ratio;
PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm; PM2.5 = particulate matter with a
nominal aerodynamic diameter less than or equal to 2.5 |jm; PNC = particle number concentration; S02 = sulfur dioxide;
aEffect estimates are standardized to a 10-ppb or 40-ppb increase in S02 concentration for 24-h avg and 1-h max metrics,
respectively.
All Lag times are in days, unless otherwise noted.
fStudies published since the 2008 ISA for Sulfur Oxides.
5-193
-------
In summary, studies of patients with implantable cardioverter defibrillators, hospital
admissions for arrhythmias, and out-of-hospital cardiac arrests do not provide evidence to
support the presence of an association between ambient SO2 concentrations and
arrhythmias. Most of these studies have been focused on other pollutants and, therefore,
have not explored whether such an association might exist in certain subgroups.
Additionally, the majority of studies used fixed-site monitors to estimate ambient SO2
exposure, which have noted limitations in capturing spatial variation in SO2 that generally
lead to attenuation and loss of precision in the effect estimates (Section 3.4.4). One
toxicological study also found no evidence for arrhythmias following short-term SO2
exposure.
5.3.1.4 Cerebrovascular Diseases and Stroke
Results among the studies reviewed in the 2008 ISA for Sulfur Oxides were inconsistent
with regard to the association between ambient SO2 concentrations and hospital
admissions or ED visits for cerebrovascular diseases or stroke (a specific form of
cerebrovascular disease). Many additional studies are now available for consideration
(study details and results presented in Table 5-27 and Figure 5-13). In Edmonton, AB,
Szvszkowicz (2008) reported that risk of ED visits for ischemic stroke was linked to SO2
concentrations, but this association was observed only in subgroup analyses stratified by
sex, season, and age. A subsequent study in Vancouver, BC, found that SO2 was
associated with risk of ED visits for ischemic stroke in the population overall [OR: 2.09
(95% CI: 1.23, 3.52) per 10-ppb increase in SO2 at lag 3] (Szvszkowicz et al.. 2012a).
The association was generally unchanged after adjustment for O3 in a copollutant model,
and attenuated, although still positive, after adjustment for CO [OR: 1.73 (95% CI: 1.00,
3.10)]. Chen et al. (2014b) also observed an association between SO2 and ischemic stroke
at longer lags in Edmonton, AB. In Brazil, Costa Nascimento et al. (2012) observed a
7.8% (95% CI: 0.0, 16.5%) increase in risk of hospital admissions of stroke per 10-ppb
increase in 24-h avg SO2 at lag 0. Zheng et al. (2013) reported a small association
between SO2 concentrations and risk of hospital admission for cerebrovascular disease
[1.7% increase (95% CI: 0.5, 2.8%) per 10-ppb increase in 24-h avg SO2 at lag 2] in
Lanzhou, a heavily polluted city in China with a high observed mean daily concentration
of SO2 (30.19 ppb) over the 5-year study period. The association was as strong, or
stronger, after adjustment for PM10 [1.8% increase (95% CI: 0.4, 3.2%)] or NO2 [2.6%
increase (95% CI: 1.4, 3.7%)] in copollutant models. In central Japan, Turin et al. (2012)
found that the risk of hemorrhagic stroke was associated with SO2 concentrations, but
found no association with other types of stroke. However, the 95% CI for the
5-194
-------
hemorrhagic stroke association was wide, and copollutant confounding was not
considered.
Table 5-27
Mean and upper percentile concentrations of sulfur dioxide from
cerebrovascular disease and stroke-related hospital admission and
emergency department visit studies.
Study
Location
Years
Exposure
Assignment
Metric
Mean/Median
Concentration
PPb
Upper Percentile of
Concentrations
PPb
tZhenq et al. (
'2013) Lanzhou. China
(2001-2005)
Average across
four monitoring
stations
24-h avg
Mean: 30.19
75th: 40.46
Max: 141.60
tThach et al. (2010) Hong Kong, Average across 24-h avg Mean: 6.79 NR
China eight monitoring
(1996-2002) stations
tBell et al. (2008) Taipei, Taiwan Average across 24-h avg Mean: 4.7 Max: 26.9
(1995-2002) 13 monitoring
stations; 5 within
city limits; or
6 with correlations
>0.75
tTurin et al. (2012) Takashima Nearest monitor 24-h avg Mean: 3.9 75th: 4.8
County, Japan to Takashima
(1988-2004) County (20 km)
Henrotin et al.
(2007)
Dijon, France
(1994-2004)
Fixed-site monitor
24-h avg
Mean:
2.63
75th: 3.44
Max: 24.81
tSzvszkowicz
(2008)
Edmonton, AB
(1992-2002)
Average across
three monitoring
stations
24-h avg
Mean:
2.6
NR
tSzvszkowicz et al.
(2012a)
Vancouver, BC
(1999-2003)
Average across
11 monitoring
stations
24-h avg
Mean:
2.5
NR
tMechtouff et al.
(2012)
Lyon, France
(2006-2007)
Average across
five monitoring
stations
24-h avg
Mean:
2.02
75th: 2.67
Max: 22.52
tChen et al.
(2014b)
Edmonton, AB
(1998-2002)
Average across
three monitoring
stations
1-h avg
Mean:
2.0
95th: 6.7
tVilleneuve et al.
(2012)
Edmonton, AB
(2003-2009)
Average across
three monitoring
stations
24-h avg
Mean:
1.5
75th: 1.9
5-195
-------
Table 5-27 (Continued) Mean and upper percentile concentrations of sulfur
dioxide from cerebrovascular disease and stroke related
hospital admission and emergency department visit
studies.
Study
Location
Years
Exposure
Assignment Metric
Mean/Median
Concentration
PPb
Upper Percentile of
Concentrations
PPb
tCosta Nascimento
Sao Paulo,
Fixed-site monitor 24-h avg
NR
NR
etal. (2012)
Brazil
(2007-2008)
Avg = average; max = maximum; NR = not reported.
fStudies published since the 2008 ISA for Sulfur Oxides.
5-196
-------
Study
tZhengetal. (2013)
¦fThach et al. (2010)
fBelletal. (2008)
-(Turin etal. (2012)
fHenrotin et al. (2007)
tSzyszkowicz et al. (2008)
tSzyszkowiczetal. (2012)
fMechtouff et al. (2012)
tChen et al. (2014)
•fVilleneuve et al. (2012)
Outcome
Cerebrovascular Disease
Stroke
Cerebrovascular Disease
Stroke
Cerebral Infarction
Intracerebral Hemorrhage
Subarachnoid Hemorrhage
Ischemic Stroke
Hemorrhagic Stroke
Ischemic Stroke
Ischemic Stroke
Ischemic Stroke
Acute Ischemic Stroke
Acute Ischemic Stroke
Stroke
Ischemic Stroke
Transient Ischemic Stroke
tNascimentoet al. (2012) Stroke
Lag
0-3
0-1
0-3
0-3
0-3
0
0
0
0
1
1
Notes
All Monitors
City Monitors
Correlated Monitors
65-100 years; Cold Season
NR
1-24 h
25-28 h
-•-r-
~
0-2
0-2
0-2
All Year
Warm Days
Cool Days
-+-
0.5 1
Risk Odds Ratio (95% CI)
2.5
CI = confidence interval; NR = not reported.
Note: Studies in red are recent studies. Studies in black were included in the 2008 ISA for Sulfur Oxides. All-year
associations = circles; summer/warm-days associations = diamonds; winter/cold-days associations = squares. Relative risks are
standardized to a 10 or 40-ppb increase in sulfur dioxide for 24-hour avg and 1-hour max metrics, respectively, but not standardized
for other metrics [e.g., (Chen et al.. 2014b)l. Lag times are reported in days, unless otherwise noted. Corresponding quantitative
results are reported in Supplemental Table 5S-17 (U.S. EPA. 2017c). All results are from single pollutant models.
Figure 5-13 Results of studies of short-term sulfur dioxide exposure and
hospital admissions for cerebrovascular disease and stroke.
In contrast to the studies that reported some evidence of an association between SO2
concentrations and cerebrovascular disease, a number of studies observed null or
imprecise associations. In an effort to reduce uncertainty related to the use of fixed-site
monitors, Bell et al. (2008) estimated SO2 exposure over the entire Taipei, Taiwan area
(average of 13 monitors), within Taipei City only (average of 5 monitors), and using a
subset of monitors where all pairs of monitors had SO2 correlations greater than 0.75
(6 monitors). Using three exposure metrics, the authors did not observe an association
between SO2 and risk of hospital admission for cerebrovascular diseases. Contrary to
other studies that reported associations between SO2 concentrations and hospital
5-197
-------
admissions and ED visits for stroke in Canada (Chen et al.. 2014b; Szvszkowicz et al..
2012a; Szvszkowicz. 2008). Villeneuve et al. (2012) reported null and/or imprecise
(i.e., wide 95% CI) associations between SO2 and all stroke, ischemic stroke, and
hemorrhagic stroke in Edmonton, AB. Studies in Hong Kong (Thach et al.. 2010). Dijon,
France (Henrotin et al.. 2007). and Lyon, France (Mechtouff et al.. 2012) also observed
null associations between SO2 concentrations and rates of hospital admission for stroke.
Overall, findings for the association between SO2 and cerebrovascular diseases continue
to be inconsistent across studies. As for other outcomes, associations reported from single
pollutant models in some locations may be at least partly due to confounding by other
pollutants.
5.3.1.5 Blood Pressure and Hypertension
Based on the data available at the time, the 2008 ISA for Sulfur Oxides (U.S. EPA.
2008d) concluded that the overall evidence was insufficient to determine that SO2 has an
effect on blood pressure. Recent evidence provides limited and inconsistent evidence for
changes in blood pressure associated with short-term exposure to SO2.
Epidemiologic Studies
A number of longitudinal studies measured BP in subjects in Beijing before, during, and
after the 2008 Beijing Olympics when citywide air pollution control measures
substantially reduced ambient levels of most criteria pollutants. Huang et al. (2012)
measured blood pressure repeatedly on up to four occasions in 40 participants with
pre-existing cardiovascular disease in Beijing, including one measurement during the
2008 Beijing Olympics when citywide air pollution control measures reduced ambient
SO2 concentrations by up to 50%. Huang et al. (2012) found a small decrement in
diastolic blood pressure per IQR increase in prior 30-minute exposure to SO2 [-0.9 mm
Hg (95% CI: -2.0, 0.2 mm Hg)], but observed a null association between ambient SO2
and systolic blood pressure. Focusing on healthy young adults, Rich et al. (2012) and
Zhang et al. (2013) observed associations between SO2 and blood pressure in
repeated-measures studies conducted before, during, and after the 2008 Beijing Olympics
(no quantitative results; results presented graphically). Using the same protocol, Zhang et
al. (2013) and Rich et al. (2012) observed a positive association between 24-h avg SO2
and systolic blood pressure, but an inverse association between 24-h avg SO2 and
diastolic blood pressure. The negative association between SO2 and diastolic blood
pressure was relatively unchanged after adjustment for PM2 5, EC, or sulfate, while the
5-198
-------
association between SO2 and systolic blood pressure was also robust to sulfate, but
attenuated, although still positive, after adjustment for PM2 5 or EC (Zhang et al.. 2013).
In another repeated measures study, Kim et al. (2016b) observed positive associations
between short-term SO2 concentrations and systolic blood pressure, diastolic blood
pressure, and mean arterial pressure among 560 older adults living in Seoul, South Korea.
A pair of cross-sectional studies reported conflicting evidence of an association.
Examining data from 7,578 participants in the Taiwanese Survey on Prevalence of
Hyperglycemia, Hyperlipidemia, and Hypertension, Chuang et al. (2010) concluded that
there is "no significant association" between SO2 concentrations and blood pressure (no
quantitative results presented). However, in a cross-sectional analysis of data from
9,238 participants in the Taiwan Community-based Integrated Screening program, Chen
et al. (2012d) found a 4.0 mm Hg (95% CI: 3.0 to 5.0 mm Hg) increase in diastolic blood
pressure per 10-ppb increase in SO2 concentrations 2 days earlier, and a 1.6 mm Hg (95%
CI: 0.15, 3.1 mm Hg) decrease in systolic blood pressure related to SO2 concentrations
3 days earlier.
In addition to longitudinal and cross-sectional studies, a few new studies examined ED
visits for hypertension. In Beijing, Guo etal. (2010) observed a 10.0% (95% CI: 1.1,
19.7%) increase in risk of ED visits for hypertension per 10-ppb increase in 24-h avg SO2
on the same day. The association was attenuated, but still positive, in a copollutant model
adjusting for PM10 [6.7% (95% CI: -3.4, 17.9%) increase at lag 0] and no longer present
in a copollutant model adjusting for NO2 [-0.8% (95% CI: -12.8, 13.0%) change at
lag 0]. Inconsistent results were reported in two studies of ED visits for hypertension in
Canada. In a case-crossover study in Calgary and Edmonton, Brook and Kousha (2015)
reported positive associations between ED visits for hypertension and 24-h avg SO2
concentrations for males [OR: 2.50 (95% CI: 1.00, 5.87) per 10-ppb increase] and
females [OR: 2.59 (95% CI: 1.12, 5.61) per 10-ppb increase]. Conversely, in Edmonton,
Szvszkowicz et al. (2012b) observed that ED visits for hypertension were both positively
and negatively associated with SO2 depending on the lag time examined.
Experimental Studies
Several experimental studies examined hypertension and blood pressure following SO2
exposure. Study characteristics are summarized in Supplemental Table 5S-15 (U.S. EPA.
2017c). One controlled human exposure study reported no change in mean arterial
pressure following SO2 exposure (Routledge et al.. 2006). Two animal toxicological
studies have examined blood pressure following SO2 exposure (Halinen et al.. 2000b;
Halinen et al.. 2000a). In both studies, SO2 was administered intratracheally to
hyperventilated guinea pigs in cold, dry air. These studies reported increases in blood
5-199
-------
pressure following cold, dry air exposure with and without SO2 and did not determine
whether there were any effects on blood pressure caused by SO2 that may not be
attributable to cold, dry air exposure.
Summary of Blood Pressure
In summary, epidemiologic studies evaluating the association between ambient SO2
concentrations and blood pressure remain inconsistent with most relying on fixed-site
monitors and few examining the potential for copollutant confounding. Experimental
studies provide no additional evidence for SC>2-induced changes in blood pressure. The
most informative studies to date found no evidence of within-person changes in blood
pressure despite relatively large changes in SO2 concentrations during the Beijing
Olympics. Experimental studies do not demonstrate effects of SO2 on blood pressure. As
such, the current evidence does not support the presence of an association between
ambient SO2 and blood pressure.
5.3.1.6 Venous Thromboembolism
VTE is a term that includes both deep vein thrombosis (DVT) and pulmonary embolism
(PE). DVT occurs when a blood clot develops in the deep veins, most commonly in the
lower extremities. A part of the clot can break off and travel to the lungs, causing a PE,
which can be life threatening.
There were no epidemiologic studies of VTE available for the 2008 ISA for Sulfur
Oxides. One recent study covering the metropolitan region of Santiago, Chile, found a
10.8% (95% CI: 3.3, 15.7%) and 8.5% (95% CI: 4.0, 13.2%) increased rate of hospital
admission for venous thrombosis and pulmonary embolism, respectively, per 10-ppb
increase in 24-h avg SO2 concentrations (Dales et al.. 2010). Copollutant models were not
evaluated. Given the limited epidemiologic evidence, the association between ambient
SO2 concentrations and venous thromboembolism is unclear.
5.3.1.7 Heart Failure
Results among the studies reviewed in the 2008 ISA for Sulfur Oxides (U.S. EPA.
2008d') were inconsistent with regard to the association between ambient SO2
concentrations and hospital admissions or ED visits for heart failure. A small number of
additional studies are now available, including a multicity study of seven Canadian cities
(Stieb et al.. 2009). Stieb et al. (2009) observed an imprecise association (i.e., wide 95%
5-200
-------
CI) between 24-h avg SO2 concentrations on the previous day and ED visits for heart
failure [3.0 % (95% CI: —1.9, 8.2%) increase in risk of ED visits per 10-ppb increase in
SO2]. Similarly, in Guangzhou, China, Yang et al. (2014a) observed a 14.5% increase
(95% CI: 6.1, 23.2%) in emergency ambulance dispatches for heart failure per 10-ppb
increase in 24-h avg SO2 concentrations on the same day. This association was slightly
attenuated, but still positive and statistically significant in copollutant models adjusting
for PM10 [13.1% (95% CI: 3.3, 23.4%)] andN02 [11.3% (95% CI: 1.7, 21.5%)]. In
contrast, Yang (2008) did not observe evidence of a positive association between ambient
SO2 exposure and heart failure in Taipei, Taiwan.
In summary, the available epidemiologic evidence is limited and inconsistent and,
therefore, does not support the presence of an association between ambient SO2
concentrations and hospital admissions or ED visits for heart failure.
5.3.1.8 Aggregated Cardiovascular Disease
Many epidemiologic studies consider the composite endpoint of all cardiovascular
diseases, which typically includes all diseases of the circulatory system (e.g., heart
diseases and cerebrovascular diseases). This section summarizes the results of
epidemiologic studies evaluating the association between ambient SO2 concentrations
and ED visits or hospitalizations for all cardiovascular diseases. Table 5-28 presents
study details and air quality characteristics of the city, or across all cities, from the U.S.
and Canadian cardiovascular-related hospital admission and ED visit studies evaluated in
the 2008 ISA for Sulfur Oxides and those more recent.
5-201
-------
Table 5-28 Mean and upper percentile concentrations of sulfur dioxide from
cardiovascular-related hospital admission and emergency
department visit studies: U.S. and Canadian studies from the 2008
ISA for Sulfur Oxides and recent studies.
Mean
Upper Percentile of
Location
Type of Visit
Concentration
Concentrations
Study
Years
(ICD9/10)
Metric
PPb
PPb
U.S.
Gwvnn et al.
Buffalo and
Hospital admissions:
24-h avg
12.2
Max: 37.7
(2000)
Rochester, NY
circulatory (401-405,
(1988-1990)
410-417)
tlto et al.
New York
Hypertensive diseases
24-h avg
7.4
(2011)
City, NY
(402, 111); Ml (410,
(2000-2006)
121-122); IHD (414,
I25); dysrhythmias
(427, I48); heart failure
(428, I50); and stroke
(430-439, I60-I69)
Koken et al.
Denver, CO
Discharge data from
24-h avg
5.7
Max: 18.9
(2003)
(1993-1997)
Agency for Healthcare
Research and Quality
database:
Acute Ml
(410.00-410.92),
atherosclerosis
(414.00-414.05),
pulmonary heart failure
(416.0-416.9),
dysrhythmia
(427.0-427.9), CHF
(428.0)
Low et al.
New York
Ischemic stroke
24 h avg
10.98
Max: 96.0
(2006)
City, NY
(433-434),
(1995-2003)
undetermined stroke
(436); monitored intake
in 11 hospitals (ED or
clinic visits). Excluded
stroke patients
admitted for
rehabilitation
Metzaer et al.
Atlanta, GA
ED visits:
1-h max:
11.0 (median)
90th: 39
(2004)
(1993-2000)
IHD (410-414); acute
Ml (410); dysrhythmias
(427); cardiac arrest
(427.5); CHF (428);
peripheral and
cerebrovascular
disease (433-437, 440,
443-444, 451-453);
atherosclerosis (440);
stroke (436)
5-202
-------
Table 5-28 (Continued): Mean and upper percentile concentrations of sulfur
dioxide from cardiovascular related hospital admission
and emergency department visit studies: U.S. and
Canadian studies from the 2008 ISA for Sulfur Oxides and
recent studies.
Michaud et al.
(2004)
Hilo, HI
(1997-2001)
ED visits
Heart (410-414,
425-429)
24-h avg
1.92 (all hourly
measurements)
Max: 447 (all hourly
measurements)
Moolaavkar
(2003)
Moolaavkar
(2000)
Cook County, Hospital admissions: 24-h avg
IL; Los
Angeles
County, CA;
Maricopa
County, AZ
(1987-1995)
CVD (390-429);
cerebrovascular
disease (430-448)
Cook: 6 (median)
Los Angeles: 2
(median)
Maricopa:
2 (median)
Cook: Max: 36
Los Angeles: Max: 16
Maricopa: Max: 14
Morris et al.
(1995)
Los Angeles,
CA; Chicago,
IL;
Philadelphia,
PA; New York
City, NY;
Detroit, Ml;
Houston, TX;
Milwaukee, Wl
(1986-1989)
Hospital admissions:
CHF (428)
1-h max Los Angeles: 10 NR
Chicago: 25
Philadelphia: 29
New York City: 32
Detroit: 25
Houston: 18
Milwaukee: 17
Peel et al. Atlanta, GA ED visits: 1-h max 16.5(17.1) 90th: 39
(2007} (1993-2000) IHD (410-414),
dysrhythmia (427),
CHF (428), peripheral
vascular and
cerebrovascular
disease (433-437, 440,
443, 444, 451-453)
tRich et al. New Jersey Hospital Admissions: 24-h avg NR NR
(2010) (2004-2006) transmural infarction
(410.0,410.1,410.2,
410.3, 410.4, 410.5,
410.6), nontransmural
infarction (410.7)
Schwartz and Detroit, Ml Hospital discharge: IHD 24-h avg 25.4 90th: 44.0
Morris (1995) (1986-1989) (410-414), CHF (428),
dysrhythmia (427)
Schwartz Tuscon, AZ Hospital discharge: 24-h avg 4.6 90th: 10.1
(1997) (1988-1990) CVD (390-429)
Tolbert et al. Atlanta, GA ED visits: 1-h max 14.9 Max: 149.0
(2Q07} (1993-2004) CVD (410-414, 427,
428, 433-437, 440,
443-445, 451-453)
5-203
-------
Table 5-28 (Continued): Mean and upper percentile concentrations of sulfur
dioxide from cardiovascular related hospital admission
and emergency department visit studies: U.S. and
Canadian studies from the 2008 ISA for Sulfur Oxides and
recent studies.
Ulirsch et al.
Southeast
Hospital admissions
NR
3.0
90th: 7.9, 7.7
(2007)
Idaho
and medical visits:
Max: 30.3, 30.3
(1994-2000)
CVD (390-429)
(two time series
examined)
Wellenius et
Birmingham,
Hospital admissions:
24-h avg
6.22 (median)
90th: 16.17
al. (2005b)
AL; Chicago,
ischemic stroke,
IL; Cleveland,
primary diagnosis of
OH; Detroit,
acute but ill-defined
Ml;
cerebrovascular
Minneapolis,
disease or occlusion of
MN; New
the cerebral arteries;
Haven, CT;
HS, primary diagnosis
Pittsburgh,
of intracerebral
PA; Seattle,
hemorrhage. (ICD
WA
codes not provided)
(1986-1999)
Wellenius et
Allegheny
Hospital admissions:
24-h avg
14.78 (9.88)
95th: 33.93
al. (2005a)
County, PA
CHF (428)
(1987-1999)
Canada
Burnett et al.
Metropolitan
Hospital discharge:
1-h max
7.9
Max: 26
(1997)
Toronto (East
IHD (410-414);
York,
cardiac dysrhythmias
Etobicoke,
(427); heart failure
North York,
(428); all cardiac
Scarborough,
(410-414, 427, 428)
Toronto, York)
(1992-1994)
Burnett et al.
Metropolitan
IHD (410-414); cardiac
24-h avg
5.35
Max: 57
(1999)
Toronto (East
dysrhythmias (427);
York,
CHF (428); all cardiac
Etobicoke,
(410-414, 427, 428)
North York,
Scarborough,
Toronto, York)
(1980-1994)
Funa et al.
Windsor, ON
CHF (428), IHD
1-h max
27.5 (16.5)
Max: 129
(2005)
(1995-2000)
(410-414),
dysrhythmias (427)
and all cardiac
Stieb et al.
Saint John,
ED visits:
24-h avg
6.7 (5.6)
95th: 18
(2000)
NB
angina pectoris, Ml,
Max: 60
(1992-1996)
dysrhythmia/conduction
disturbance, CHF, all
cardiac
5-204
-------
Table 5-28 (Continued): Mean and upper percentile concentrations of sulfur
dioxide from cardiovascular related hospital admission
and emergency department visit studies: U.S. and
Canadian studies from the 2008 ISA for Sulfur Oxides and
recent studies.
tSzvszkowicz Edmonton, AB ED visits: 24-h avg 2.6 NR
(2008) (1992-2002) acute ischemic stroke
(434 and 436)
tSzvszkowicz Vancouver, ED visits (discharge 24-h avg 2.5 NR
et al. (2012a) BC diagnosis):
(1999-2003) transient ischemic
attack, cerebrovascular
incident, seizure
tSzvszkowicz Edmonton, AB ED visits: hypertension 24 h avg 2.6 Max: 16.3
et al. (2012b) (1992-2002) (401.9)
Villeneuve et Edmonton, AB ED visits: 24-h avg All year: 2.6 (1.9) All year 75th: 4.0
al. (2006a) (1992-2002) stroke
Avg = average; CHF = congestive heart failure; CVD = cardiovascular disease; ED = emergency department; HS = hemorrhagic
stroke; ICD = International Classification of Diseases; IHD = ischemic heart disease; ISA = Integrated Science Assessment;
max = maximum; Ml = myocardial infarction; NR = not reported.
tStudies published since the 2008 ISA for Sulfur Oxides.
The majority of epidemiologic studies reviewed in the 2008 ISA for Sulfur Oxides (U.S.
EPA. 2008d') found a positive association between ambient SO2 concentrations and rates
of hospital admission or ED visits for all cardiovascular diseases. One prominent study
from the previous ISA was a study conducted in 14 cities across Spain, which observed a
3.5% (95% CI: 0.5, 6.7%) increased risk of hospital admissions for all cardiovascular
diseases per 10-ppb increase in SO2 at lag 0-1 lYBallester et al.. 2006); study details and
results for this and other studies in this section are presented in Table 5-29. and
Figure 5-141. The study authors indicate (results not reported) that the association with
SO2 was attenuated after adjustment for CO or NO2 in copollutant models. Most studies
published since the 2008 ISA for Sulfur Oxides also observed positive associations
between SO2 and ED visits or hospitalizations for all CVD, although only a few
considered potential copollutant confounding. For example, a case-crossover study in
Beijing found that SO2 averaged over eight monitoring sites was associated with risk of
ED visits for all cardiovascular diseases in a single-pollutant model [OR: 1.04 (95% CI:
1.01, 1.06) per 10-ppb increase in SO2 on the same day] (Guo et al.. 2009). The
association remained comparable in copollutant models adjusting for either PM2 5 [OR:
1.03 (95% CI: 0.99, 1.06)] orN02 [OR: 1.03 (95% CI: 1.00, 1.07)]. Similarly, in
Shanghai, Chen etal. (2010b) reported a small increase in risk of hospital admissions for
CVD per 10-ppb increase in 24-h avg SO2 at lag 5 [1.7% (95% CI: 0.5, 3.0%)] and lag
5-205
-------
0-6 [1.3% (5% CI: 0.0, 3.2%)]. The association at lag 5 was similar after adjusting for
NO2 or PM10, while copollutant models for lag 0-6 were not presented.
A number of other studies considering single-pollutant models also reported generally
consistent associations between SO2 concentrations and hospital admissions or ED visits
for CVD. A study in New York City (Ito etal.. 2011) observed an association between
SO2 concentrations that was stronger in the warm season [OR: 1.026 (95% CI: 1.021,
1.031) per 10-ppb increase in 24-h avg SO2] than in the cold season [OR: 1.018 (95% CI:
0.998, 1.049)]. Two studies in Sao Paolo, Brazil (Filho et al.. 2008; Martins et al.. 2006)
also found associations in single pollutant models (no quantitative results; results
presented graphically). Another study found an increase in the risk of daily hospital
admissions per IQR increase in 24-h avg SO2 in the heavily polluted city of Lanzhou,
China (Zheng et al.. 2013). However, this association was less clinically relevant when
standardized to a 10-ppb increase in 24-h avg SO2. In contrast, a large study in Beijing,
China reported that CVD ED visits were not associated with SO2 concentrations on the
same day (Zhang et al.. 2015b). The authors also examined a number of other single-day
lags and cumulative lags and found little evidence of an association. One important
limitation is that these studies did not examine potential copollutant confounding
particularly for PM25 and sulfate.
Table 5-29 Mean and upper percentile concentrations of sulfur dioxide from
cardiovascular-related hospital admission and emergency
department visit studies.
Study
Location
Years
Exposure
Assignment Metric
Mean/Median
Concentration
PPb
Upper Percentile
of Concentrations
PPb
tlto etal. (2011)
New York City,
NY
(2000-2006)
Average across 24-h avg
five monitoring
sites
Mean: 7.4
NR
Metzqer et al. (2004)
Atlanta, GA
(1993-2000)
Fixed-site monitor 1-h max
Median: 11
90th: 39
Moolqavkar (2003)
Los Angeles, CA
(1987-1995)
Fixed-site monitor 24-h avg
NR
NR
Schwartz (1997)
Tuscon, AZ
(1998-1990)
Fixed-site monitor 24-h avg
Mean: 4.6
75th: 5.9
90th: 10.1
Burnett et al. (1997)
Toronto, Canada
(summer
1992-1994)
Average across 1-h max
four to six
monitoring sites
Mean: 7.9
75th: 11
Max: 26
5-206
-------
Table 5-29 (Continued): Mean and upper percentile concentrations of sulfur
dioxide from cardiovascular related hospital admission
and emergency department visit studies.
Study
Location
Years
Exposure
Assignment
Metric
Mean/Median
Concentration
PPb
Upper Percentile
of Concentrations
PPb
Sunver et al. (2003)
Seven European
cities
(1990-1996)
Fixed-site
monitors in each
city
24-h avg
Median: 1.9-8.0
across cities
90th: 5.3-29.4
across cities
Ballester et al.
(2006)
14 Spanish
cities
(1995-1999)
Citywide average
for each city
24-h avg
Mean: 2.9-15.6
across cities
90th: 4.8-28.8
across cities
Atkinson et al.
(1999)
London,
England
(1992-1994)
Average across
five monitoring
sites
24-h avg
Mean: 8.1
90th: 11.8
Max: 31.4
Poloniecki et al.
(1997)
London,
England
(1987-1994)
Fixed-site monitor
24-h avg
Median: 6
90th: 21
Max: 114
Anderson et al.
(2001)
Birmingham,
England
(1994-1996)
Average across
five monitoring
sites
24-h avg
Mean: 7.2
90th: 12.3
Max: 59.8
Ballester et al.
(2001)
Valencia, Spain
(1994-1996)
Average across
14 monitoring
sites
24-h avg
Mean: 9.8
Max: 26.1
Llorca et al. (2005)
Torrelavega,
Spain
(1992-1995)
Average across
three monitoring
sites
24-h avg
Mean: 5.1
NR
tFilho et al. (2008)
Sao Paulo,
Brazil
(2001-2003)
Average across
13 monitoring
sites
24-h avg
Mean: 5.3
Max: 16.4
tMartins et al.
(2006)
Sao Paulo,
Brazil
(1996-2001)
Average across
six monitoring
sites
24-h avg
Mean: 6.5
Max: 28.7
tZhenq et al. (2013)
Lanzhou, China
(2001-2005)
Average across
four monitoring
sites
24-h avg
Mean: 30.2
75th: 40.5
Max: 141.6
tZhana et al.
(2015b)
Beijing, China
(2009-2011)
Average across
11 monitoring
stations
24-h avg
Mean: 10.7
75th: 13.4
Max: 89.5
+Guo et al. (2009)
Beijing, China
(2004-2006)
Average across
eight monitoring
sites
24-h avg
Mean: 18.8
75th: 23.7
Max: 111.8
5-207
-------
Table 5-29 (Continued): Mean and upper percentile concentrations of sulfur
dioxide from cardiovascular related hospital admission
and emergency department visit studies.
Study
Location
Years
Exposure
Assignment
Metric
Mean/Median
Concentration
PPb
Upper Percentile
of Concentrations
PPb
tChen etal. (2010b)
Shanghai, China
(2005-2007)
Average across
six monitoring
sites
24-h avg
Mean: 21.4
75th: 27.5
Max: 89.7
Wonaetal. (1999)
Hong Kong,
China
(1994-1995)
Average across
seven monitoring
sites
24-h avg
Median: 6.5
75th: 9.5
Max: 26.1
Chana et al. (2005)
Taipei, Taiwan
(1997-2001)
Average across
six monitoring
sites
24-h avg
Mean: 4.3
75th: 5.5
Max: 14.6
Jalaludin et al.
(2006)
Sydney,
Australia
(1997-2001)
Average across
14 monitoring
sites
24-h avg
Mean: 1.07
75th: 1.39
Max: 3.94
Petroeschevskv et
al. (2001)
Brisbane,
Australia
(1987-1994)
Average across
two monitoring
sites
24-h avg
Mean: 13.9
Max: 49.7
Avg = average; max = maximum; NR = not reported.
fStudies published since the 2008 ISA for Sulfur Oxides.
5-208
-------
Study
Outcome
Lag
tlto etal. (2011)
CVD
0
Metzger et al. (2004)
CVD
0-2
Moolgavkar et al. (2003)
CVD
0
Schwartz et al. (1997)
CVD
0-2
Burnett etal. (1997)
CVD
0-3
Sunyer et al. (2003)
CVD
0-1
Ballesteret al. (2006)
CVD
0-1
Atkinson et al. (1999)
CVD
0
Poloniecki et al. (1997)
CVD
1
Anderson et al. (2001)
CVD
0-1
Ballesteret al. (2001)
CVD
2
Llorca et al. (2005)
CVD
0
tZheng et al. (2013)
CVD
0-3
tZhang et al. (2015)
CVD
0
fGuo etal. (2009)
CVD
0
tChen etal. (2010b)
CVD
5
Wonget al. (1999)
CVD
0-1
Chang et al. (2005)
CVD
0-2
Jalaludinetal. (2006)
Petroeschevsky et al.
(2001)
CVD
CVD
0
§
Notes
WarmS eason
Cold Season
All Ages
65+ Tears Old
All Ages
0-64 Years Old
65+ Years Old
«IrsOld
m s
:
!-•-
!~
All Ages
5-64 Years Old
5+ Years Old
-+-
0.75 1 1.25
Risk or Odds Ratio (95% CI)
1.5
CI = confidence interval; CVD = cardiovascular disease.
Note: Studies in red are recent studies. Studies in black were included in the 2008 ISA for Sulfur Oxides. All-year
associations = circles; summer/warm-days associations = diamonds; winter/cold-days associations = squares. Relative risks are
standardized to a 10 or 40-ppb increase in sulfur dioxide for 24-hour avg and 1-hour max metrics, respectively. Lag times are
reported in days, unless otherwise noted. Corresponding quantitative results are reported in Supplemental Table 5S-18 (U.S. EPA.
2017c). All results are from single pollutant models.
Figure 5-14 Studies of hospital admissions and emergency department visits
for all cardiovascular disease.
Overall, consistent associations between ambient SO2 concentrations and rates of hospital
admissions or ED visits for all cardiovascular diseases have been observed. Although
associations are evident in single-pollutant models in many locations, there was limited
assessment of potential copollutant confounding. Therefore, this association may at least
partly be the result of confounding by correlated pollutants. Additionally, most studies
examined 24-h avg exposure metrics for SO2, which may not adequately capture the
spatial and temporal variability in SO2 concentrations (Section 5.2.1.2).
5-209
-------
5.3.1.9 Cardiovascular Mortality
Studies evaluated in the 2008 SOx ISA that examined the association between short-term
SO2 exposure and cause-specific mortality found consistent positive associations with
cardiovascular mortality using a 24-h avg exposure metric. Across studies, there was
evidence that the magnitude of the SC>2-cardiovascular mortality relationship was similar
or slightly larger than total mortality. Recent multicity studies conducted in Asia (Chen et
al.. 2012b; Kan et al.. 2010) and Italy (Bellini et al.. 2007). and a meta-analysis of studies
conducted in Asia (Atkinson et al.. 2012) provide evidence that is consistent with those
studies evaluated in the 2008 SOx ISA (Section 5.5.1.3. Figure 5-18). The associations
between short-term SO2 concentrations and cardiovascular mortality are further supported
by studies focusing on stroke mortality (Yang etal.. 2014b; Chen et al.. 2013). In a study
conducted in eight of the CAPES cities, Chen et al. (2013) reported associations for SO2
and stroke similar to those for all cardiovascular mortality across all of the CAPES cities
(Section 5.5.1.3. Figure 5-18). The magnitude of the association for stroke mortality
observed in Chen et al. (2013) is supported by multiple systematic reviews and
meta-analyses of stroke mortality (Shah et al.. 2015; Yang et al.. 2014b). Both studies
reported similar results, with Yang et al. (2014b) reporting a 2.5% increase in stroke
mortality (95% CI: 1.8, 3.1) for a 10-ppb increase in 24-h avg SO2 concentrations in a
meta-analysis of mortality studies conducted in Asia, Europe, and North America and
Shah et al. (2015) reporting a 2.2% increase in stroke mortality (95% CI: 1.4, 3.1) for a
10-ppb increase in SO2 concentrations (averaging time was not reported) in a
meta-analysis of studies conducted worldwide. However, when interpreting the results of
Yang et al. (2014b). it is important to note that when examining regional associations in
SCh-related stroke (i.e., Asia vs. Europe and North America), which combined both
mortality and hospital admission outcomes, the magnitude of the association was much
smaller, 0.8% (95% CI: -0.2, 1.7), than those observed in studies conducted in Asia,
2.1% (95% CI: 1.2, 3.2). This could be attributed to the relatively low variability and
overall low SO2 concentrations observed in both Europe and North America compared to
Asia (Section 5.5.1.2 and Table 5-39).
Previous studies evaluated in and prior to the 2008 SOx ISA that examined the
association between short-term SO2 exposures and cardiovascular mortality focused
exclusively on single-pollutant analyses. Therefore, questions arose with regard to the
independent effect of SO2 on cardiovascular mortality and whether associations remained
robust in copollutant models. A few recent multicity studies conducted in China (Chen et
al.. 2012b) and across Asia (Kan et al.. 2010) examined both of these questions. Chen et
al. (2012b) found that the S02-cardiovascular mortality association was attenuated, but
remained positive in copollutant models with PM10 [1.0% (95% CI: 0.08, 1.9) for a
10-ppb increase in 24-h avg SO2 concentrations at lag 0-1] and NO2 [0.5% (95% CI:
5-210
-------
-0.5, 1.4)]. These results are similar to those reported by Chen et al. (2012b) when
examining the SCh-total mortality association in models with NO2 (i.e., -80% reduction),
but a larger degree of attenuation was observed in models with PM10 for cardiovascular
mortality (i.e., -40% reduction for total mortality and 50% reduction for cardiovascular
mortality) (Section 5.5.1.4). Kan et al. (2010). as part of the PAPA study, also examined
potential copollutant confounding (i.e., NO2, PM10, and O3) but only in each city
individually. The authors found that, although the SCh-cardiovascular mortality
association remained positive in copollutant models, there was evidence of an attenuation
of the association in models with PM10 and NO2 (Figure 5-19). In an analysis of stroke
mortality in eight of the CAPES cities, Chen et al. (2013) reported a pattern of
associations similar to that of Chen et al. (2012b) and Kan et al. (2010) in copollutant
models with PM10 and NO2. In single-pollutant models, the authors reported a 2.3% (95%
CI: 1.4, 3.2) increase in stroke mortality for a 10 ppb increase in 24-h avg SO2
concentrations at lag 0-1. However, in copollutant models, Chen et al. (2013) observed
that SC>2-stroke mortality associations were attenuated in models with PM10, -40%
reduction [1.9% (95% CI: 0.3, 3.5)] andN02, -80% reduction [0.0% (95% CI: -1.8,
1.9)]. Overall, the studies that examined potential copollutant confounding on the
SCh-cardiovascular mortality relationship report results consistent with what was
observed for total mortality. However, the overall assessment of copollutant confounding
remains limited, and it is unclear how the results observed in Asia translate to other
locations, specifically due to the unique air pollution mixture and higher concentrations
observed in Asian cities.
Of the multicity studies evaluated, potential seasonal differences in S02-cardiovascular
mortality associations were only assessed in a study conducted in Italy (Bellini et al..
2007) with additional information from U.S.-based single-city studies conducted in
Philadelphia (Sacks et al.. 2012) and New York City (Ito et al.. 2011). In a study of
15 Italian cities, Bellini et al. (2007) reported larger S02-cardiovascular mortality
associations in the summer [9.4% increase (April-September)], compared to both winter
[1.6% increase (October-March)] and all-year analyses (2.9% increase), which are
consistent with the pattern of associations observed for total and respiratory mortality.
These results are supported by Ito etal. (2011) in a study conducted in New York City
that found that when examining single-day lags of 0 to 3 days, the S02-cardiovascular
mortality association was consistently positive during the warm season, ranging from a
1.2 to 3.5% increase across lags. The authors reported no evidence of an association in
winter and all-year analyses. Within this analysis, Ito etal. (2011) reported rather poor
monitor-to-monitor temporal correlations for SO2, which would indicate potential
exposure error and subsequently attenuation and imprecision in the risk estimate
(Sections 3.4.2. 3.4.4). Sacks et al. (2012) provide additional support to the limited
evidence indicating differences in the seasonal pattern of S02-cardiovascular mortality
5-211
-------
associations. However, as detailed in Section 5.5.1.4. Sacks et al. (2012) demonstrated
that across models that use various approaches to control for seasonality and the potential
confounding effects of weather, the magnitude of seasonal SC>2-cardiovascular mortality
associations may vary depending on the modeling approach employed. Therefore,
although Bellini et al. (2007) and Ito et al. (2011) provide initial evidence indicating
potentially larger cardiovascular mortality associations in the summer, the results of
Sacks et al. (2012) suggest that the evidence remains unclear whether the seasonal pattern
of SCh-cardiovascular mortality associations is consistent across statistical modeling
choices and study locations.
An uncertainty that often arises when evaluating studies that examine the relationship
between short-term air pollution exposures and cause-specific mortality is whether
analyses of statistical modeling parameters, the lag structure of associations, and the C-R
relationship provide results that are consistent with what is observed for total mortality.
Chen et al. (2013) examined each of these issues in a study of stroke mortality, with
additional supporting evidence from the full CAPES (Chen et al.. 2012b). When
examining alternative approaches to controlling for seasonality, Chen et al. (2013) found
that increasing the df employed from 4 to 10 df per year did not substantially change the
SCh-stroke mortality association. However, when altering the lag structure of the
temperature term included to control for the potential confounding effects of weather,
Chen et al. (2012b) reported an attenuation of the association, although it did remain
positive. As detailed in Section 5.5.1.4. this could be the result of including only one
temperature term in the model.
When examining the lag structure of associations, Chen et al. (2013) reported results for
stroke mortality that are consistent with those observed for all cardiovascular mortality.
As depicted in Figure 5-15. there is evidence of a steady decline in the SC>2-stroke
mortality association at longer individual lag days, with the strongest association
occurring for a moving average of lag 0-1 day. A similar pattern of associations was
observed for cardiovascular mortality by Chen et al. (2012b) in the full CAPES
(Figure 5-20). as well as the PAPA study (Kan et al.. 2010) (Figure 5-21). These results
are further confirmed in a systematic review and meta-analysis of studies of stroke
mortality conducted by Yang et al. (2014b). which found the strongest associations at
lag 0 and 1 in a subgroup analysis of single-day lags of 0 to 2 days.
5-212
-------
2.5 r
i I
S 0.5
•E 0.0
o>
£ -0.5
Lag
-1.0 L
01 = lag 0-1 days; 04 = lag 0-4 days; S02 = sulfur dioxide.
Source: Adapted from Chen et al. (20131. ReprintedwithpermissionolWoltersKluwerHealth.
Figure 5-15 Percent increase in stroke mortality associated with a 10 |jg/m3
(3.62 ppb) increase in sulfur dioxide concentrations using
different lag structures.
Chen et al. (2013) also examined the shape of the SC^-stroke mortality C-R relationship.
To examine the assumption of linearity, the authors fit both a linear and spline model to
the SCh-stroke mortality relationship. Chen et al. (2013) then computed the deviance
between the two models to determine any evidence of nonlinearity. An examination of
the deviance did not indicate that the spline model improved the overall fit of the
SCh-stroke mortality relationship (Figure 5-16).
5-213
-------
1 1 1 1 1
0 20 40 eo 80
SC2 concert aticr
S02 = sulfur dioxide.
Note: The solid line represents the mean estimate and the dotted lines are 95% confidence intervals.
Source: Adapted from Chen et al. (20131. Reprinted with permission of Wolters Kluwer Health.
Figure 5-16 Pooled concentration-response curves for sulfur dioxide and
daily stroke mortality in eight Chinese cities for a 10 |jg/m3
(3.62 ppb) increase in 24-hour avg concentrations at lag 0-1 day.
Overall, recent multicity studies report evidence of consistent positive associations
between short-term SO2 concentrations and cardiovascular mortality, which is consistent
with those studies evaluated in the 2008 SOx ISA. Unlike studies evaluated in the 2008
SOx ISA, recent studies examined whether copollutants confound the relationship
between short-term SO2 concentrations and cardiovascular mortality. Overall, these
studies reported evidence that the S02-cardiovascular mortality association was
attenuated in models with NO2 and PM10, but the analyses are limited to Asian cities
where the air pollution mixture and concentrations are different from those reported in
other areas of the world. A few studies examined potential seasonal patterns in
associations, and found initial evidence of larger S02-cardiovascular mortality
associations in the summer/warm season. However, seasonal associations may be
influenced by study location and the statistical modeling choice employed. In examining
other uncertainties, a limited number of analyses suggest that: (1) when examining model
specification, associations remain robust when alternating the df used to control for
5-214
-------
seasonality; (2) when examining lag structure, associations are larger and more precise
(i.e., narrower confidence intervals) within the first few days after exposure in the range
of 0 and 1 days; and (3) when examining the C-R relationship, there is a log-linear, no
threshold C-R relationship. However, for both total and cause-specific mortality, studies
have conducted a rather limited exploration of potential alternatives to linearity when
examining the shape of the C-R relationship, which in combination with the potential
measurement error due to uncharacterized spatial and temporal variability in SO2
concentrations, complicates the interpretation of the SCh-mortality C-R relationship
(Sections 3.4.2.2 and 3.4.2.3).
5.3.1.10 Subclinical Effects Underlying Cardiovascular Diseases
The following subsections review studies of subclinical effects that serve as useful
measures of physiological and biochemical responses that could provide mechanistic
evidence to describe a role for SO2 in the manifestation of cardiovascular diseases. These
subclinical effects are not widely validated markers of specific clinical cardiovascular
outcomes, but could potentially underlie the development, progression, or indication of
various clinical events and provide biological plausibility for multiple outcomes.
Heart Rate and Heart Rate Variability
The 2008 ISA for Sulfur Oxides concluded that the overall evidence available at the time
was insufficient to conclude that SO2 has an effect on cardiac autonomic control as
assessed by indices of HRV. HRV provides a noninvasive marker of cardiac autonomic
nervous system function. The rhythmic variation in the intervals between heart beats can
be quantified in either the time domain or the frequency domain (TFESC and NASPE.
1996). Common time-domain measures of HRV include the standard deviation of all
normal-to-normal intervals (SDNN, an index of total HRV) and the root-mean-square of
successive differences (rMSSD, an index influenced mainly by the parasympathetic
nervous system). In the frequency domain, HRV is usually divided into the high
frequency (HF) and low frequency (LF) components, as well as the ratio of the LF to HF
components (LF:HF) (TFESC and NASPE. 1996). Changes in indices of HRV have been
associated with increased risk of cardiovascular events in prospective cohort studies
(Eguchi et al.. 2010; La Rovere et al.. 2003; Kikuva et al.. 2000; Tsuii et al.. 1996; Tsuii
etal.. 1994V
5-215
-------
Epidemiology
A number of additional epidemiologic studies are now available for review. In a
cross-sectional study in South Korea, Min et al. (2009) reported negative associations
between ambient SO2 concentrations and indices of HRV (SDNN, and the LF and HF
components) among 256 smokers, but no association among the 767 nonsmokers (no
quantitative results; result presented graphically). In another cross-sectional study, Min et
al. (2008b) reported a -7.6% (95% CI: -14.7, 0.1%) change in SDNN and a -23.1%
(95% CI: -35.4, -6.5%) change in LF per 10-ppb increase in 24-h avg SO2 among
1,349 participants in South Korea. The amount of overlapping participants between these
two studies is unclear.
The above studies are limited by their cross-sectional approach that compares measures
of HRV across individuals assessed on different days. In contrast, longitudinal or
repeated-measure study provides an estimate of the average association between SO2 and
measures of HRV within individuals. Huang et al. (2012) measured HRV repeatedly in
40 participants with pre-existing cardiovascular disease in Beijing in the summer of 2007
and again in the summer of 2008, including one measurement period during the 2008
Beijing Olympics when citywide air pollution control measures substantially reduced
ambient concentrations of most criteria pollutants. In this study, SO2 concentrations
during the Olympics were reduced by nearly 30% versus the previous month and nearly
50% versus the same period the previous summer (Huang et al.. 2012). Despite these
large changes in SO2 concentrations, overall only small associations were observed
between SO2 concentrations and HRV indices, limited to a 4.8% reduction (95% CI:
-9.1, -0.3%) in the LF component and an unexpected 4.1% increase (95% CI: -2.2,
10.9%) in the HF component of HRV per interquartile range (NR) increase in SO2 in the
previous 12 hours (Huang et al.. 2012). In subgroup analyses, SDNN was significantly
positively associated with SO2 concentrations among those with higher levels of
C-reactive protein (CRP; a marker of inflammation), those with diabetes, and males.
These results are difficult to understand given that a higher SDNN is generally thought to
be associated with lower risk of cardiovascular events. The findings were also
inconsistent with another study that observed a negative association between SDNN and
ambient SO2 concentrations. A repeated measure study in Shanghai, China reported a
4.36% reduction (95% CI: -5.85, -2.86%) in SDNN per IQR increase (NR) in 4-hour
moving average exposure to SO2 (Sun et al.. 2015). This association was attenuated but
still statistically significant in copollutant models adjusting for BC [-2.91% (95% CI:
-4.66, -1.13%)] and 03 [-3.24% (95% CI: -4.83, -1.62%)]. The association was
attenuated and no longer statistically significant but still negative in copollutant models
adjusting for N02 [-0.56% (95% CI: -2.38, 1.30%)] and CO [-1.25% (95% CI: -3.02,
0.55%)]. In another study in Beijing before, during, and after the 2008 Olympics, Rich et
5-216
-------
al. (2012) observed small but statistically significant increases in heart rate associated
with ambient SO2 concentrations on the previous day (no quantitative results; result
presented graphically). In expanded results from the same protocol, Zhang et al. (2013)
found that the association was similar in copollutants models adjusting for CO, NO2, O3,
EC, or OC, but was attenuated and no longer positive after adjustment for PM2 5 or SO42 .
Zhang et al. (2013) also reported a strong association between LF:HF and ambient SO2
concentrations on the previous day. This association was relatively unchanged after
adjustment for CO, NO2, O3, EC, OC, or PM2 5 in copollutant models and attenuated but
still positive after adjustment for SO42 . In contrast, a panel study in Taipei, Taiwan used
Holter monitors to continuously monitor HRV in 46 participants and observed no
associations between ambient SO2 and SDNN, r-MSSD, LF component, or HF
component (quantitative results not reported) (Chuang et al.. 2007). Although new studies
are available, findings are mixed and they do not support the presence of an association
between ambient SO2 and measures of HRV.
Experimental Studies
Several experimental studies examined heart rate and HRV following SO2 exposure.
Study characteristics are summarized in Supplemental Table 5S-15 (U.S. EPA. 2017c).
Animal studies have reported no changes in heart rate following SO2 exposures of
1,000-5,000 ppb in guinea pigs and 1,200 ppb in rats (Nadzieiko et al.. 2004; Halinen et
al.. 2000b; Halinen et al.. 2000a).
Controlled human exposure studies evaluated heart rate during and following SO2
exposure. Tunnicliffe et al. (2001) reported no change in heart rate in healthy adults or
adults with asthma during exposure to 200 ppb SO2 for 1 hour at rest. However,
Routledge etal. (2006) found an increase in heart rate measured by a statistically
significant decrease in the RR interval from electrocardiographic (ECG) recordings
4 hours after SO2 exposure in healthy adults. Statistically significant changes in heart rate
were not observed in S02-exposed older adults with stable angina and coronary artery
disease immediately after and 4 hours after exposure. Tunnicliffe et al. (2001) did not
obtain ECG measures following exposure and, thus, would not have been able to observe
the increase in heart rate reported by Routledge etal. (2006).
Tunnicliffe et al. (2001) and Routledge et al. (2006) reported changes in different
measures of HRV in adults either during or shortly after SO2 exposure. Tunnicliffe et al.
(2001) reported that HF power, LF power, and total power were higher with SO2
exposures compared to air exposure in the healthy subjects, but that these indices were
reduced during SO2 exposure in the subjects with asthma (statistical significance only in
total power in healthy adults). The LF:HF ratios were unchanged in both groups.
Routledge etal. (2006) reported a reduction in SDNN, rMSSD, percentage of successive
5-217
-------
RR interval differences exceeding 50 ms (pNNso), and HF power, the latter of which did
not reach statistical significance, in healthy adults 4 hours after SO2 exposure. Baroreflex
sensitivity was also reduced 4 hours after SO2 exposure determined by changes in a-HF
and a-LF. There were no changes in HRV among the patients with coronary heart
disease; however, this lack of response may be due to a drug treatment effect because a
large portion of these patients were taking beta-blockers. The changes in HRV observed
in Tunnicliffe et al. (2001) and Routledge et al. (2006) indicate the potential for SO2 to
affect the autonomic nervous system (see Section 4.3.1).
Summary of Heart Rate and Heart Rate Variability
The current epidemiologic evidence does not support the presence of an association
between ambient SO2 and measures of HRV. No changes in heart rate were observed in
experimental animal studies while changes in HRV observed in human clinical studies
may indicate the potential for SO2 to affect the autonomic nervous system (see
Section 4.3.1). Overall, studies evaluating the effect of ambient SO2 concentrations and
measures of HRV and heart rate remain limited.
QT Interval Duration
The time between the start of the Q wave and the end of the T wave on an ECG (QT
interval) provides an electrocardiographic marker of ventricular repolarization.
Prolongation of the QT interval is associated with increased risk of life-threatening
ventricular arrhythmias. In an analysis of data from the Boston-area Normative Aging
Study, Baiaetal. (2010) observed a small and imprecise (i.e., wide confidence intervals)
association between heart-rate-corrected QT interval and 10-hour moving average of SO2
concentrations among older, generally white men (no quantitative results; result
presented graphically). The only prior study available for comparison from the 2008 ISA
for Sulfur Oxides (U.S. EPA. 2008d) also found that SO2 concentrations were positively
associated with increased QT interval duration amongst a small sample of 56 men in
Erfurt, Germany [3.75 ms increase (95% CI: 1.21, 6.28 ms) per 0.61-ppb increase in 24-h
avg SO2] (Henneberger et al.. 2005). There was little variability between daily measured
SO2 concentrations, so the effect estimate is not standardized to prevent inflation of the
confidence interval.
The two reviewed studies provide limited evidence of association between short-term
SO2 exposure and markers of ventricular repolarization. Neither of these studies
evaluated potential copollutant confounding and coherence for an association between
SO2 exposure and arrhythmias is not provided by experimental studies (Section 5.3.1.3).
5-218
-------
Insulin Resistance
There were no epidemiologic studies of diabetes or insulin deficiency available for the
2008 ISA for Sulfur Oxides. Recent studies reported contrasting findings regarding
short-term associations between air pollutants and measures of insulin resistance and
fasting glucose, which play key roles in the development of Type 2 diabetes mellitus. In a
panel study of older adults in Korea, Kim and Hong (2012) observed 0.94 (95% CI:
-0.02, 1.88) and 0.94 (95% CI: 0.01, 1.81) mean increases in the homeostatic model
assessment index of insulin resistance [fasting insulin x (fasting glucose ^ 22.5)] per
10-ppb increase in 24-h avg SO2 at lags 3 and 4, respectively. There were imprecise
(i.e., wide 95% CI) or null associations at all other individual lag days examined, from 0
to 10. Another panel study, conducted in the heavily polluted Tangshan, China, reported
an association between 24-h avg SO2 concentrations and fasting glucose levels (Chen et
al.. 2015b). However, this association is unlikely to be clinically relevant when
standardized to a 10-ppb increase in 24-h avg SO2 [0.045 mmol/L (95% CI: 0.039,
0.050 mmol/L) increase at lag 0-3], Conversely, Kelishadi et al. (2009) reported the lack
of an association between 24-h avg SO2 and insulin resistance in a cross-sectional study
of 374 Iranian children aged 10-18 years.
In summary, the available epidemiologic evidence is limited and inconsistent, and does
not support the presence of an association between ambient SO2 concentrations and
measures of insulin resistance.
Biomarkers of Cardiovascular Risk
Several epidemiologic and toxicological studies have explored the potential relationship
between SO2 and biomarkers of cardiovascular risk. In particular, markers of
inflammation have been evaluated in a number of epidemiologic and toxicological
studies published since the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d') (Table 5-30).
Relatively few studies have evaluated the potential link between SO2 and other
circulating markers of cardiovascular risk, including markers of coagulation, vascular
injury, or lipid oxidation.
5-219
-------
Table 5-30 Epidemiologic studies of biomarkers of cardiovascular effects.
Study
Location and
Years
Sample Size
Mean and Upper
Concentration SO2
PPb
Exposure
Assessment
Selected Effect Estimates3
(95% CI)
tDubowskv et al.
(2006)
St. Louis, MO
Mar-Jun 2002
(n = 44)
24-h avg: 6.7
75th percentile: 7.4
Max: 27
Fixed-site CRP (percent change)
Lag 04: -36.1 (-65.2, -2.8)
IL-6 (percent change)
Lag 04: -16.5 (-38.7, 6.5)
White blood cells (cells/pL)
Lag 04: 10.0 (0.4, 19.6)
tSteinvil et al. Tel Aviv, Israel 24-h avg: 2.8 City wide avg CRP (percent WBC (cells/pL)
(2008) 2002-2006 ^5th percentile: 3.5 change) men; men; women
women Lag 0:
LagO: 231 (-419,
0 (-38, 38); 875)]
-13 (56,28) -169 (-1,000,
Lag 1: 656)
-19 (-50, 25); Lag 1:
-13 (-63, 38) 44 (-631, 713);
Lag 2: -544 (-1,381,
6 (-38, 44); 294)
-25 (-69, 31) Lag 2;
Fibrinogen -125 (-819,
(mg/dL) 563);
men; women -481 (-1,356,
Lag 0: 388)
-20.0 (-40.0, 0.6);
-23.8 (-51.3, 3.8)
Lag 1:
-21.3 (-42.5, 0.0);
-13.1 (-41.3,
14.4)
Lag 2:
-15.0 (-37.5, 6.9);
17.5 (-11.9, 46.9)
tThompson et al. Toronto, ON
(2010) 1999-2003
(n =45)
24-h avg: 3.57 Fixed-site
No quantitative results; results
presented graphically. Increase in
IL-6 associated with 4- and 5-day
moving avg SO2 concentrations. Null
association between SO2 and
fibrinogen
Correlations: CO: 0.43, NO2: 0.44,
Os: -0.19, PM2.5: 0.45
tGandhi et al. Piscataway, NJ
(2014) 2005-2009
(n = 49)
24 h avg: 2.4 Fixed-site
75th percentile: 3.2
Max: 13.8
Change in plasma nitrate (nM):
Lag 0: 53.6 (-4.5, 111.4)
Lag 1: 45.0 (0.9, 90.9)
Lag 2: 48.2 (-13.2, 110.0)
5-220
-------
Table 5-30 (Continued): Epidemiologic studies of biomarkers of cardiovascular
effects.
Study
Location and
Years
Sample Size
Mean and Upper
Concentration SO2
PPb
Exposure
Assessment
Selected Effect Estimates3
(95% CI)
tLee etal. (2011b)
Allegheny
County, PA
1997-2001
(n = 1,696)
7-day avg: 8.4
75th percentile: 10.1
Max: 25.4
City wide avg
No quantitative results presented.
"...SO2... associations (with CRP)
were negligible for both the entire
population and nonsmokers only."
tHildebrandt et al.
(2009)
Erfurt,
Germany
2001-2002
(n = 38)
24-h avg: 1.35
Max: 14.2
Fixed-site
No quantitative results presented.
"No significant associations"
between SO2 and inflammatory
(fibrinogen, E-selectin) or
coagulation (D-dimer, prothrombin)
markers.
Baccarelli et al.
(2007a)
Lombardia,
Italy
1995-2005
(n = 1,218)
24-h avg median:
2.4
75th percentile: 4.5
Max: 96.7
City wide avg
Effect estimates not provided. SO2
not correlated with anticoagulation
proteins (plasma fibrinogen,
functional AT, functional protein C,
protein C antigen, functional
protein S, or free protein S).
Baccarelli et al.
(2007b)
Lombardia,
Italy
1995-2005
(n = 1,213)
24-h avg Median:
2.4
75th percentile: 4.5
Max: 96.7
City wide avg
Homocysteine difference, fasting
(percent change).
Lag 24 h: 0.2 (-6.3, 6.7)
Lag 0-6 days: 0.2 (-4.3, 4.7)
Homocysteine difference,
post-methionine-load (percent
change)
Lag 24 h: 2.6 (-3.2, 8.6)
Lag 0-6 days: 2.6 (-1.5, 6.7)
Wellenius et al.
(2007)
Boston, MA
2002-2003
(n = 28)
24-h avg: 4.8
City wide avg
No quantitative results presented.
"No significant associations were
observed between (NO2) and B-type
natriuretic peptide levels at any of
the lags examined."
tGoldbera et al.
(2008)
Montreal, QC
2002-2003
(n = 31)
NR
Fixed-site
Oxygen saturation (mean difference)
Lag 0: -0.104 (-0.320, 0.110)
Lag 1: -0.277 (-0.497, -0.058)
Lag 0-2: -0.210 (-0.536, 0.116)
tBruske et al.
(2011)
Augsburg,
Germany
2003-2004
(n = 200)
24-h avg: 1.15
75th percentile:
1.26
Max: 2.4
Fixed-site
No quantitative results; results
presented graphically. Inverse
associations were observed for SO2
with Lp-PLA2 at lag days 2 and 3
and positive associations were
estimated with Lp-PLA2 lag days 4
and 5.
Correlations: PNC: 0.77, PM2.5:
0.42, PM10: 0.43, CO: 0.63, NO2:
0.51, NO: 0.60, O3: -0.45.
5-221
-------
Table 5-30 (Continued): Epidemiologic studies of biomarkers of cardiovascular
effects.
Study
Location and
Years
Sample Size
Mean and Upper
Concentration SO2
PPb
Exposure
Assessment
Selected Effect Estimates3
(95% CI)
tZhanq et al. (2013)
Beijing, China
Jun-Oct, 2008
(n = 125)
24-h avg Fixed-site No quantitative results; results
Before: 7.45 presented graphically. Positive
During: 2.97 association between SO2 and
After: 6.81 fibrinogen (lag 6). Inverse
association between SO2 and WBC
count (lag 5).
tLinetal. (2015)
Beijing, China
2007-2008
(n = 36 school
children)
NR
Monitor located
nearby school
Urinary 8-oxodG
(Geometric mean ratio by SO2
exposure percentile)
<30th (<2.1 ppb):
referent
30th-60th (2.1-6.4 ppb):
1.26 (0.93, 1.70)
60th-90th (6.4-49.1 ppb):
1.66 (1.15, 2.41)
>90th (>49.1 ppb):
2.31 (1.54, 3.46)
Urinary Malondialdehyde
<30th: referent
30th-60th: 1.21 (1.05, 1.40)
60th-90th: 1.40 (1.15, 1.69)
>90th: 1.40 (1.08, 1.83)
tKhafaie et al. Pune City,
(2013) India
2005-2007
(n = 1,392)
24-h avg: 8.3 Citywide avg
No quantitative results; results
presented graphically. SO2 was
associated with increases in CRP at
lags 0, 1, 2, 4, 5, 0-7, 0-14, and
0-30.
AT = atascadero; avg = average; CI = confidence interval; CO = carbon monoxide; CRP = C-reactive protein; IL-6 = interleukin-6;
Lp-PLA2 = lipoprotein-associated phospholipase A2; max = maximum; n = sample size; N = population number; NO = nitric oxide;
N02 = nitrogen dioxide; NR = not reported; 03 = ozone; PM2.5 = particulate matter with a nominal aerodynamic diameter less than
or equal to 2.5 |jm; PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm; PNC = particle
number concentration; S02 = sulfur dioxide; WBC = white blood cell.
aEffect estimates are standardized to a 10-ppb or 40-ppb increase in S02 concentration for 24-h avg and 1-h max metrics,
respectively.
fStudies published since the 2008 ISA for Sulfur Oxides.
Note: All lag times are in days, unless otherwise noted.
5-222
-------
Epidemiologic Studies
The epidemiologic data available for review by the 2008 ISA for Sulfur Oxides (U.S.
EPA. 2008d) did not suggest a consistent link between SO2 and biomarkers of
cardiovascular risk, including markers of inflammation and coagulation. Results from
more recent studies continue to be inconsistent. Dubowskv et al. (2006) investigated
associations between ambient pollutants and markers of systemic inflammation in a panel
(repeated-measures) study of 44 seniors in St. Louis, MO and found that higher ambient
SO2 concentrations were associated with lower levels of CRP and white blood cells, but
not IL-6 (results for this study, and other studies in this section can be found in
Table 5-30). Similarly, during the Beijing Olympics, SO2 was inversely associated with
white blood cell counts, although positively associated with fibrinogen (Zhang et al..
2013). The negative associations observed in these two studies are unexpected and
difficult to explain. In contrast, among 45 nonsmoking adults, Thompson et al. (2010)
found a positive association between SO2 and IL-6, but not fibrinogen. In another panel
study examining pollutant levels before, during, and after the Beijing Olympics, Lin et al.
(2015) reported positive associations between SO2 concentrations and urinary markers of
oxidative stress, malondialdehyde and 8-oxodG, in children.
In a cross-sectional analysis of data from a panel study of 49 young adults in New Jersey,
Gandhi et al. (2014) observed that plasma nitrite levels, a marker for endothelial
dysfunction, were associated with an increase in 24-h avg SO2 concentrations on the
same day. Khafaie et al. (2013) observed a positive association between SO2 and CRP in
a cross-sectional study of patients with Type 2 diabetes in Pune City, India, whereas a
study of 1,696 pregnant women (Lee etal.. 2011b). and one of 38 male patients with
chronic pulmonary disease (Hildebrandt et al.. 2009) observed null associations between
SO2 and CRP. In a cross-sectional analysis of 3,659 participants in Tel-Aviv, Steinvil et
al. (2008) observed inconsistent and/or imprecise (i.e., wide 95% CI) associations
between SO2 and CRP, white blood cells, or fibrinogen among men and women.
Observed associations were both positive and negative depending on the length of the
lags, making interpretation of the results difficult.
Ambient SO2 concentrations are reportedly not associated with blood coagulation
(Baccarelli et al.. 2007a). plasma homocysteine (Baccarelli et al.. 2007b). markers of
vascular injury (Hildebrandt et al.. 2009). or markers of functional status in patients with
heart failure (Wellenius et al.. 2007). Conversely, SO2 concentrations were inversely
associated with blood oxygen saturation in patients with heart failure (Goldberg et al..
2008) and positively associated with lipoprotein-associated phospholipase A2 (LP-PLA2)
in survivors of myocardial infarction (Briiske et al.. 2011).
5-223
-------
Experimental Studies
Experimental studies examined biomarkers of cardiovascular risk following SO2
exposure, including markers of inflammation, coagulation, and oxidative injury. Study
characteristics are summarized in Supplemental Table 5S-15 (U.S. EPA. 2017c). No
changes were reported in serum C-reactive protein or markers of coagulation (fibrinogen,
D-dimer, platelet aggregation, blood count, or differential white blood cell count) in
healthy humans and patients with stable angina and coronary artery disease exposed to
SO2 (Routlcdgc et al.. 2006). An animal toxicological study examined the hematological
effects of short-term SO2 exposure on blood biomarkers. Acute exposure of rats to
870 ppb SO2 for 24 hours resulted in increased hematocrit, sulfhemoglobin, and osmotic
fragility as well as decreased whole blood and packed cell viscosities (Baskurt. 1988).
These results indicate a systemic effect of inhaled SO2 and are consistent with an
oxidative injury to red blood cells. Furthermore, prolonged exposure to as low as 0.3 ppm
SO2 results in measurable amounts of circulating sulfite and its metabolite S-sulfonate in
humans (see Section 4.3.4). The relationship between circulating sulfite/S-sulfonate and
cardiovascular effects of inhaled SO2 has not yet been explored in human subjects.
Summary of Blood Markers of Cardiovascular Risk
There is inconsistent evidence from epidemiological studies regarding any potential link
between SO2 and other circulating markers of cardiovascular risk. Few experimental
studies of markers of inflammation or oxidative stress in humans or animals were
available; however, the presence of circulating sulfite/S-sulfonate has been reported. In
addition, there is limited evidence for systemic oxidative stress under conditions of
prolonged exposure. Overall, evidence from available studies is not coherent between
disciplines and does not clearly support an effect of ambient SO2 concentrations and
markers of cardiovascular disease, including inflammation.
5.3.1.11 Summary and Causal Determination
Overall, the available evidence is inadequate to infer a causal relationship between
short-term exposure to SO2 and cardiovascular health effects. Multiple epidemiologic
studies report positive associations between short-term ambient SO2 concentrations and
cardiovascular outcomes, such as cardiovascular mortality, myocardial infarction and
ischemic heart disease, and aggregated cardiovascular outcomes; however, substantial
uncertainties remain regarding exposure measurement error and copollutant confounding.
Specifically, the majority of studies reporting positive associations evaluated averaged
SO2 concentrations over multiple monitors and used a 24-h avg exposure metric, which
may not adequately capture the spatial and temporal variability in S02 concentrations
5-224
-------
(Sections 3.4.2.2 and 3.4.2.3). Further, among studies adjusting for copollutants, the
observed associations are generally attenuated, complicating the determination of an
independent SO2 association. Support for observed epidemiological associations from
experimental data is weak, particularly as some experimental evidence for cardiovascular
effects relied on high concentrations and prolonged exposures. These experimental
studies also provide some evidence to support key events in a proposed mode of action
for cardiovascular effects from SO2 exposure, including changes in the autonomic
nervous system and oxidative stress. Overall, these experimental studies provide only
limited evidence of biological plausibility and key events in a proposed mode of action;
however, important limitations remain, as well as issues of coherence across disciplines.
This determination is consistent with that of the 2008 ISA for Sulfur Oxides (U.S. EPA.
2008d). The majority of epidemiologic studies reviewed in the 2008 ISA for Sulfur
Oxides examined hospital admissions or ED visits for aggregated categories of
cardiovascular disease or for mortality from cardiovascular causes. These studies
generally reported positive associations in single pollutant models, but analyses designed
to assess copollutant confounding were limited. Relatively few studies evaluated specific
cardiovascular outcomes such as MI, arrhythmia, cerebrovascular disease, and heart
failure, and those that were available did not support an association with short-term SO2
exposure. Controlled human exposure studies demonstrated the potential for SO2
exposure to exert an effect on the autonomic nervous system but there was a lack of
supporting animal toxicological data. The available animal toxicological studies did not
report effects on HR, HRV, arrhythmia, or blood pressure following short-term SO2
exposures [Table 5S-15 (U.S. EPA. 2017cYI. In addition, limited and inconsistent
mechanistic evidence, including evidence pertaining to key events in a proposed mode of
action, offered only limited insight for the role of SO2 in the triggering of cardiovascular
diseases. Although multiple recent epidemiologic studies add to the evidence available
for the current review, the additional studies do not substantially reduce uncertainties
related to copollutant confounding. Moreover, there continues to be a limited
experimental evidence to provide biological plausibility to strengthen the inference of
causality for S02-related cardiovascular effects.
The evidence for cardiovascular effects, with respect to the causal determination for
short-term exposure to SO2 is detailed below using the framework described in the
Preamble to the IS As IYU.S. EPA. 2015b'). Tables I and II]. The key evidence, supporting
or contradicting, as it relates to the causal framework is summarized in Table 5-31.
5-225
-------
Table 5-31 Summary of evidence, which is inadequate to infer a causal
relationship between short-term sulfur dioxide exposure and
cardiovascular effects.
Rationale for Causal
Determination3
Key Evidence13
Key References'3
SO2 Concentrations
Associated with
Effects0
Triggering a myocardial infarction
Although most
epidemiologic studies
examining Ml or all
CVD report positive
associations, results
are generally
attenuated after
adjustment for
copollutant
confounding
Increases in hospital admissions and
ED visits for IHD, Ml, and all CVD in
adults in multiple studies, including
multicity studies
However, a number of studies report
associations with ED visits and
hospital admissions were attenuated
after adjustment with CO, NO2, or
PM10.
Section 5.3.1.2
Section 5.3.1.8
Supplemental
Figures 5S-3, 5S-4, and
5S-5 (U.S. EPA. 2017c)
24-h avg: 1.2-15.6 ppb
24-h avg: 1.9-30.2 ppb
Uncertainty regarding Studies examining the association Section 3.4.2.2
exposure between short-term SO2 exposures
measurement error and cardiovascular effects generally
rely on single or the average of
multiple monitors. Because SO2
generally has low to moderate
spatial correlations across urban
geographical scales such studies are
subject to exposure measurement
error.
Uncertainty due to Lack of evidence from epidemiologic
lack of coherence with panel studies and experimental
other lines of evidence studies for clinical cardiovascular
effects
Lack of evidence to
identify key events in
the proposed mode of
action
Lack of mechanistic evidence for key Section 4.3
events leading to extrapulmonary
effects
Limited and inconsistent evidence of Section 5.3.1.10
increased systemic inflammation in
epidemiologic studies
5-226
-------
Table 5-31 (Continued): Summary of evidence, which is inadequate to infer a
causal relationship between short term sulfur dioxide
exposure and cardiovascular effects.
Rationale for Causal
Determination3
Key Evidence13
Key References'3
SO2 Concentrations
Associated with
Effects0
Other cardiovascular effects
Inconclusive evidence
from epidemiologic,
controlled human
exposure and
toxicological studies
Epidemiologic studies report
generally null associations between
SO2 and risk of cardiac arrest and
arrhythmias. One experimental study
provides no evidence of arrhythmia.
Section 5.3.1.3
Inconsistent epidemiologic evidence Sections 5.3.1.4 and
for an association between SO2 and 5.3.1.5
risk of cerebrovascular disease and
stroke, and increased blood
pressure and hypertension
Insufficient quantity of studies Sections 5.3.1.6 and
evaluating decompensation of heart 5.3.1.7
failure and venous thrombosis and
pulmonary embolism
Changes in HR and HRV reported in Tunnicliffe et al. (2001) 200 ppb, 1 h at rest
controlled human exposure but Routledae et al (2006) (humans)
coherence with animal toxicological _ .. ^ „ . ..
, ¦ . ¦ 1 4. .¦ |. .< 1 Section 5.3.1.10
and epidemiologic studies is limited
Some evidence to
identify key events in
the proposed mode of
action
Some evidence for activation of Section 4.3.1
neural reflexes in humans leading to Fiqure 4-2
altered HRV
200 ppb, 1 h at rest
(humans)
Some evidence of systemic oxidative (Baskurt. 1988)
stress based on measured
sulfhemoglobin and mitochondrial
changes
Section 4.3.4
0.87 ppm, 24 h
(rats)
1.34 ppm, 4 h/day for
30 days
(rats)
5-227
-------
Table 5-31 (Continued): Summary of evidence, which is inadequate to infer a
causal relationship between short term sulfur dioxide
exposure and cardiovascular effects.
Key References'3
SO2 Concentrations
Associated with
Effects0
Rationale for Causal
Determination3
Key Evidence13
Cardiovascular mortality
Consistent
epidemiologic
evidence but
uncertainty regarding
SO2 independent
effect
Multicity studies consistently observe Section 5.3.1.9
24-h avg: 2.5-38.2
associations with cardiovascular
mortality, including stroke with 24-h
avg SO2 at lags primarily of 0-1
days.
Analysis of potential confounding by
copollutants primarily limited to PM10
and NO2 reported evidence of
attenuation of associations. No
studies included copollutant
analyses with PM2.5.
Chen etal. (2012b)
Chen etal. (2013)
Kan etal. (2010)
Bellini etal. (2007)
Atkinson et al. (2012)
Avg = average; CO = carbon monoxide; CVD = cardiovascular disease; ED = emergency department; HR = heart rate;
HRV = heart rate variability; IHD = ischemic heart disease; Ml = myocardial infarction; N02 = nitrogen dioxide; PM10 = particulate
matter with a nominal aerodynamic diameter less than or equal to 10 |jm; PM2.5 = particulate matter with a nominal aerodynamic
diameter less than or equal to 2.5 |jm; S02 = sulfur dioxide.
aBased on aspects considered in judgments of causality and weight of evidence in causal framework in Tables I and II of the
Preamble to the ISAs (U.S. EPA. 2015b).
bDescribes the key evidence and references, supporting or contradicting, that contribute most heavily to causal determination.
References to earlier sections indicate where full body of evidence is described.
°Describes the S02 concentrations with which the evidence is substantiated.
Recent epidemiologic studies of specific cardiovascular outcomes add to the overall
evidence for the effect of short-term SO2 exposure on the cardiovascular system with a
number of these studies evaluating effects related to triggering an MI (Section 5.3.1.2V
Several recent epidemiologic studies of MI hospitalizations and ED visits consistently
report associations in single pollutant models but associations are not always robust in
copollutant models indicating that the associations may be due to confounding (Hsich et
al.. 2010; Cheng et al.. 2009; Ballester et al.. 2006). The small number of studies based
on clinical MI data, rather than hospitalizations, report inconsistent evidence regarding
associations between ambient SO2 concentrations and risk of MI (Miloicvic et al.. 2014;
Turin et al.. 2012; Bhaskaran et al.. 2011). The only study that examined the association
of hourly ambient SO2 concentrations prior to MI onset reported no association, although
there was some evidence of a positive association in a sensitivity analysis of older adults
(Bhaskaran et al.. 201IV Although Chuang et al. (2008) reported an association between
short-term SO2 exposure and ST-segment changes, a nonspecific marker of myocardial
ischemia, in patients with a history of coronary heart disease that generally remained
unchanged after additional control for PM2 5 and BC in copollutant models; the evidence
overall, was not generally consistent.
5-228
-------
Findings from recent studies of the association of short-term exposure to SO2 with
hospital admissions or ED visits for cerebrovascular diseases or stroke are inconsistent
and, associations reported from single pollutant models in some locations may be due to
confounding by copollutants (Section 5.3.1.4V Epidemiologic studies evaluating the
association between ambient SO2 concentrations and blood pressure remain inconsistent
with most relying on centrally located monitors that do not capture the spatial variability
of SO2 and few examining the potential for copollutant confounding (Section 5.3.1.5).
Although a small number of studies were conducted to examine the association of
short-term SO2 exposure with other clinical outcomes, including heart failure
(Section 5.3.1.7) and VTE (Section 5.3.1.6). findings from these studies do not support an
effect of short-term exposure to SO2. There is also a lack of epidemiologic evidence
supporting an effect of short-term SO2 exposure on arrhythmia (Section 5.3.1.3).
although associations between short-term SO2 exposure and markers of ventricular
repolarization abnormalities that are risk factors for arrhythmia have been observed (Baia
et al.. 2010; Henneberger et al.. 2005) (Section 5.3.1.10).
Consistently positive associations have been reported in epidemiologic studies of
short-term SO2 exposure and cardiovascular mortality (Section 5.3.1.9). These include
studies reviewed in the 2008 ISA for Sulfur Oxides and recent multicity studies that
generally report an association similar or slightly larger in magnitude for cardiovascular
mortality compared to total mortality. Studies that report results from copollutants models
generally report attenuation of the association between short-term SO2 exposure and
cardiovascular mortality after adjustment for PM10 and NO2.
Few experimental studies have evaluated the effects of SO2 exposure on the
cardiovascular system. The strongest evidence comes from controlled human exposure
studies, for which copollutant confounding is not a concern, that short-term exposure to
SO2 can affect the autonomic nervous system of healthy adults and adults with asthma
(Routledge et al.. 2006; Tunnicliffe et al.. 2001) (Section 5.3.1.10). These studies report
changes in HR and HRV following SO2 exposure in adults, which is indicative of
potential cardiovascular effects being mediated by the neural reflex pathway; however,
these changes were not reported in animal studies, nor did epidemiological evidence
support the presence of associations, particularly after adjusting for copollutant
confounding (Section 5.3.1.10). Animal studies also provide limited evidence for the role
of systemic oxidative stress as a key event in a proposed mode action (Section 4.3). In
particular, studies observed changes in sulfhemoglobin (Baskurt. 1988). In addition,
studies of long-term exposure found lipid peroxidation in the brain (Oin et al.. 2012). and
mitochondrial changes in the heart and brain (Qin et al.. 2016; Qin et al.. 2012)
(Section 4.3.4). While evidence of the diffusion of sulfite into the circulation and tissues
following exposure to SO2 has been reported and could play a role in the induction of
5-229
-------
systemic effects, this evidence is limited by the lack of studies directly investigating the
relationship between circulated sulfite and cardiovascular effects (Section 4.3.4V
Furthermore, changes observed in animal studies were not reported in human
experimental data, or epidemiological studies, and the available data relied on prolonged
exposure periods and higher concentrations (0.87-1.34 ppm). Overall, these experimental
studies provide limited evidence of biological plausibility and key events in a proposed
mode of action; however, there remain important issues, such as limited coherence
between disciplines, as well as limitations in the data based on exposure concentration
and duration.
Despite numerous additional epidemiologic studies reporting positive associations
between short-term SO2 exposure and cardiovascular effects, a key uncertainty that
remains since the 2008 ISA for Sulfur Oxides is the potential for confounding by other
pollutants, specifically those from a common source that are highly correlated with SO2.
The majority of hospital admission or ED visit studies have not evaluated whether the
reported associations with SO2 are robust to adjustment for other pollutants. Those
studies that do examine associations with SO2 adjusted for PM [Figure 5S-3 and
Table 5S-19 (U.S. EPA. 2017c)l. N02 [Figure 5S-4 and Table 5S-20 (U.S. EPA. 2017c)l.
or other correlated pollutants [Figure 5S-5 and Table 5S-21 (U.S. EPA. 2017c)l report
that, in general, associations were either attenuated or no longer present after controlling
for potential copollutant confounding (Hsieh et al.. 2010; Cheng et al.. 2009; Ballester et
al.. 2006). A limited number of studies examined copollutant confounding on the
S02-cardiovascular mortality relationship, which included analyses on stroke mortality,
and provided evidence that the SO2 association was reduced in copollutant models with
N02and PM10 (Chen et al.. 2013; Chen et al.. 2012b; Kan et al.. 2010). Finally, while
copollutant models are a common statistical tool used to evaluate the potential for
copollutant confounding, their interpretation can be limited (Section 5.1.2V Without
consistent and reproducible experimental evidence that is coherent with the effects
observed in epidemiologic studies, uncertainty still exists concerning the role of
correlated pollutants in the associations observed with SO2. Thus, uncertainty remains
regarding the extent to which SO2 exposure is independently associated with
cardiovascular outcomes or if SO2 is a marker for the effects of another correlated
pollutant or mix of pollutants.
In conclusion, the evidence overall is inadequate to infer a causal relationship between
short-term SO2 exposure and cardiovascular health effects. This conclusion does not
represent a change from the conclusion of the 2008 ISA for Sulfur Oxides (U.S. EPA.
2008d). Multiple epidemiologic studies report positive associations between short-term
ambient SO2 concentrations and cardiovascular outcomes, but these associations are
generally attenuated after adjustment for copollutants. There is limited experimental
5-230
-------
evidence in humans or animals evaluating exposure to SO2 and the results of these studies
only provide limited coherence for the positive associations observed in the
epidemiologic studies. Further, while there is some evidence of key events in a proposed
mode of action, important limitations remain in the available evidence regarding the
biological plausibility of effects observed in epidemiological studies. The evidence from
epidemiologic and experimental studies is of insufficient consistency and continues to
have limited coherence, and thus, is inadequate to infer between short-term SO2 exposure
and cardiovascular effects.
5.3.2 Long-Term Exposure
5.3.2.1 Introduction
Studies of the effects of long-term exposure to SO2 on the cardiovascular system were not
available for inclusion in the 1982 AQCD (U.S. EPA. 1982a). The 2008 ISA for Sulfur
Oxides (U.S. EPA. 2008d) reviewed a limited body of toxicological and epidemiologic
studies published through 2006 and concluded that the available evidence was "too
limited to make any conclusions" between the effects of long-term exposure to SO2 and
cardiovascular health. These recent studies do not change the conclusion from the 2008
ISA for Sulfur Oxides (U.S. EPA. 2008d).
The 2008 ISA for Sulfur Oxides included one epidemiologic study, which reported an
increased risk of cardiovascular events in association with long-term exposure to SO2 in
post-menopausal women (50-79 years old) without previous CVD from 36 U.S.
metropolitan areas. In this study, Miller et al. (2007) found that PM2 5 was most strongly
associated with cardiovascular events [MI, revascularization, angina, congestive heart
failure (CHF), CHD death], compared to the other pollutants evaluated [HR: 1.24 (95%
CI: 1.04, 1.48) per 10 |ig/m3], followed by SO2 [1.07 (95% CI: 0.95, 1.20) per 5 ppb].
Exposures to air pollution were estimated by assigning the annual (for the year 2000)
mean air pollutant concentration measured at the monitor nearest to the subject's
five-digit residential ZIP code centroid. The effect estimate for SO2 was strengthened in a
multipollutant model that was adjusted for several other pollutants including PM2 5.
However, correlations among pollutants were not described and exposure measurement
error may have introduced a bias (Section 3.4.4.2). Consequently, the extent to which this
study supports an independent effect of SO2 on the cardiovascular system is limited.
Several recent epidemiologic studies of the association of long term SO2 exposure with
subclinical and clinical cardiovascular outcomes add to the available body of evidence.
5-231
-------
Experimental animal studies with long-term exposures below 2,000 ppb were not
available for inclusion in the 2008 ISA for Sulfur Oxides. Although a small number of
studies using exposures above 2,000 ppb were included, they did not contribute
substantially to conclusions because the concentrations of SO2 used in these studies were
unlikely to be relevant to ambient concentrations of SO2. No new toxicological studies in
humans or animals have been published since the 2008 ISA for Sulfur Oxides. Overall,
the biological plausibility and independence of the SO2 effect observed in epidemiologic
studies remains an important uncertainty.
This section reviews the published studies of the cardiovascular effects of long-term
exposure to SO2 (i.e., longer than 1 month). To clearly characterize the evidence
underlying causality, the discussion of the evidence is organized into groups of related
outcomes [ischemic heart disease and myocardial infarction (Section 5.3.2.2).
cerebrovascular disease and stroke (Section 5.3.2.3). hypertension (Section 5.3.2.4). other
cardiovascular effects (Section 5.3.2.5). and cardiovascular mortality (Section 5.3.2.6)1.
Evidence for subclinical effects (e.g., blood biomarkers of cardiovascular effects) of
long-term exposure to SO2 are discussed in Section 5.3.2.7 and serve to inform biological
plausibility across multiple clinical cardiovascular events and outcomes.
Similar to Section 5.3.1. studies examining cardiovascular effects of sulfite exposure (via
intraperitoneal injection, intravenous injection, etc.) are not included in this section
because these studies generally involve exposures to sulfite that are higher than what is
expected to occur following inhalation of SO2 at ambient relevant concentrations. Studies
in humans and animals suggest that prolonged exposure to SO2 may result in measurable
changes in the concentrations of sulfite in plasma and tissues, but these changes would be
expected to be far less following concentrations of SO2 typically found in ambient air.
The literature describing the distribution and metabolism of sulfite is discussed in
Sections 4.2.3 and 4.2.4. The potential role of sulfite in the induction of systemic effects,
such as effects of the cardiovascular system, is discussed in Section 4.2.4.
5.3.2.2 Myocardial Infarction and Ischemic Heart Disease
IHD generally develops due to a buildup of plaques in the arterial walls
(i.e., atherosclerosis) that impede the blood flow and oxygen delivery to the heart. This
restricted oxygen delivery or ischemia from excess plaque, plaque rupture, and clot
formation can lead to an MI. Several epidemiologic studies provide evidence of a
relationship between long-term exposure to SO2 and ischemic heart disease and incident
or fatal MI (Table 5-32). However, uncertainty remains regarding the influence of
exposure measurement error on the effect estimates observed in epidemiologic studies
5-232
-------
(Section 3.4.4.2) and the ability of these studies to distinguish the independent effect of
long-term SO2 exposure from the effect of correlated copollutant exposures
(Section 3.4.3).
Table 5-32 Epidemiologic studies of the association of long-term exposure to
sulfur dioxide with cardiovascular disease.
Study
Cohort,
Location, and
Study Period
Mean
PPb
Exposure
Assessment
Effect Estimates (95% Cl)a
tLipsett et al. (2011)
California
Teachers Study
Cohort
N = 124,614
(n = 43 Ml
events)
California
Jun1996-
Dec 2005
S02
Geocoded
IQR: 0.43
residential address
mean:
linked to pollutant
1.72
surface developed
using IDW
(fixed-site monitors
concentrations
from 1995-2005
used to model
exposure as a
time-dependent
function)
Correlation of SO2
with: ozone,
r= -0.17
PM2.5, r= 0.02
PM10, r= 0.54
NO2, r= 0.67
CO, r= 0.80
IQR:
Annual average
0,83
SO2 concentration
mean
for 2002 at a 1 by
(SD):
1-km resolution
1.47
derived from
dispersion models
and linked to
residential post
codes
Correlation of SO2
with: NO2, r= 0.86
Ml incidence
SO2: HR 1.97 (0.07, 60)
Covariates: age, race,
smoking, second-hand
smoke, BMI, lifetime physical
activity, nutritional factors,
alcohol, marital status,
menopausal status, hormone
replacement therapy,
hypertension medication and
aspirin, and family history of
Ml/stroke
Copollutant adjustment:
none
tAtkinson etal. (2013)
National GP
Patient Cohort
N = 810,686
(n = 13,956 Ml
events)
England
2003
Ml incidence
HR: 1.34 (1.13, 1.50)
Covariates: age, sex,
smoking, BMI, diabetes,
hypertension, and index of
multiple deprivation
Copollutant adjustment:
none
5-233
-------
Table 5-32 (Continued): Epidemiologic studies of the association of long term
exposure to sulfur dioxide with cardiovascular disease.
Study
Cohort,
Location, and
Study Period
Mean
PPb
Exposure
Assessment
Effect Estimates (95% Cl)a
tRosenlund et al. (2006)
SHEEP cohort
n = 1,397 cases
and
1,870 controls
Stockholm,
Sweden
1992-1994
Cases Dispersion models
median: to estimate SO2
9.6 from heating at
5th—95th: residential
2.6-18.2 address.
Controls Residential history
median' available for 30 yr
g 3 exposure estimate.
5th—95th: Correlation of 30 yr
7.7-17.5 SO2 with:
30 yrN02,r= 0.73
30 yr CO, r= 0.49
First Ml
OR: 0.99 (0.9,1.1) per 5 ppb
Covariate adjustment: age,
sex, hospital catchment
area, smoking, diabetes,
physical inactivity, and SES
Copollutant adjustment:
none
tAncona et al. (2015)
Rome, Italy
2001-2010
N = 85559
(n = 491 men
and
356 women)
2.5 |jg/m3
SOx
SD: 0.9
Lagrangian particle
dispersion model
(SPRAY Ver. 5)
used SOx as
exposure marker
for petrochemical
refinery emissions
IHDb
HR men: 0.87 (0.74, 1.02)
HR women: 0.83 (0.64,1.07)
CVDb
HR men: 1.01 (0.93, 1.0)
HR women: 1.02 (0.92, 1.12)
PM10: 0.81
H2S: 0.78
Miller et al. (2007)
WHI Cohort
U.S.
1994-1998
NR Annual avg (2000):
nearest monitor to
residence ZIP
code centroid
Cardiovascular events (Ml,
revascularization, angina,
CHF, CHD death)
HR: 1.07 (0.95, 1.20)
Covariates: age, ethnicity,
education, household
income, smoking, diabetes,
hypertension, systolic blood
pressure, BMI, and
hypercholesterolemia
HR: 1.13 (0.98, 1.30) after
simultaneous adjustment for
PM2.5, PM10-2.5, CO, NO2,
and O3
5-234
-------
Table 5-32 (Continued): Epidemiologic studies of the association of long term
exposure to sulfur dioxide with cardiovascular disease.
Study
Cohort,
Location, and
Study Period
Mean
PPb
Exposure
Assessment
Effect Estimates (95% Cl)a
tQin etal. (2015)
tDonq et al. (2013a)
N =24,845
Random
selection
(18-74 yr) from
households in
33 communities
in 11 districts of
northeastern
China
Mean:
20.3
Median:
18
IQR: 7.5
3-yr avg
(2006-2008) SO2
concentration for
each district
NO2, r= 0.38
O3, r= 0.87
PM10, r= 0.70
CVD
OR: 1.08 (0.93, 1.26)
Note: associations stronger
among males
OR BMI <25 kg/m2
1.11 (0.97, 1.27)
OR BMI 25 kg/m2
1.12 (0.99, 1.25)
Covariate adjustment: age,
race, education, income,
smoking, drinking, exercise,
diet, sugar, family history of
CVD or stroke, and district
Copollutant adjustment:
none
Avg = average; BMI = body mass index; CHD = coronary heart disease; CHF = congestive heart failure; CI = confidence interval;
CO = carbon monoxide; CVD = cardiovascular disease; GP = general practice; H2S = hydrogen sulfide; HR = heart rate;
IDW = inverse distance weighting; IHD = ischemic heart disease; IQR = interquartile range; Ml = myocardial infarction; n = sample
size; N = population number; N02 = nitrogen dioxide; NR = not reported; 03 = ozone; OR = odds ratio; PM25 = particulate matter
with a nominal aerodynamic diameter less than or equal to 2.5 |jm; PM10 = particulate matter with a nominal aerodynamic diameter
less than or equal to 10 |jm; PM10-2.5 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm and
greater than a nominal 2.5 |jm; r= correlation coefficient; SD = standard deviation; SES = socioeconomic status;
SHEEP = Stockholm Heart Epidemiology Programme; S02 = sulfur dioxide; SOx = sulfur oxides; WHI = Women's Health Initiative.
aEffect estimates are standardized per 5-ppb increase in S02 concentrations.
bEffect estimate per 2.88 |jg/m3 increase in SOxconcentration (as reported by author in original publication).
fStudies published since the 2008 ISA for Sulfur Oxides.
Lipsett et al. (2011) analyzed the association of incident MI with long-term exposure to
SO2, other gases (NO2, CO, O3), and PM. These authors studied a cohort of California
public school teachers aged 20-80 years old (N = 124,614). Each participant's geocoded
residential address was linked to pollutant surfaces that were determined by inverse
distance weighted interpolation of pollutant concentrations measured at fixed-site
monitors during the period 1996-2005. The average of monthly SO2 concentrations was
modeled as a time-dependent function for subjects with at least 12 months of exposure.
Those living outside the radial range for which the monitor was intended to provide
representative data were excluded from the analysis. This "representative range" was
3 km for neighborhood SO2 monitors and 5 km for the urban/regional SO2. There was no
evidence of an association between SO2 and incident MI [HR 1.97 (95% CI: 0.07, 60) per
5 ppb].
Atkinson et al. (2013) examined the association of incident cardiovascular disease with
SO2. These authors studied patients (aged 40-89 years) registered with 205 general
5-235
-------
practices across England. The authors report that approximately 98% of the population is
registered with a general practitioner minimizing the potential for selective participation.
Predicted annual average SO2 concentrations within 1-km x 1-km grids, estimated using
dispersion models, were assigned to participants based on their residential postal code.
Cardiovascular disease outcomes included in the analysis were MI, stroke, arrhythmias,
and heart failure. Authors reported an association of SO2 with MI in a fully adjusted
model [HR: 1.34 (95% CI: 1.13, 1.50) per 5 ppb]. The performance of the dispersion
model used to estimate SO2 concentration was characterized as moderate to poor
depending on the study year. Failure of the model to capture the spatial variability of SO2
could lead to bias toward or away from the null (Section 3.4.4.2).
Rosenlund et al. (2006) conducted a population case-control study to examine the
association of first MI with long-term exposure to air pollution in Stockholm, Sweden. In
this study, residential histories were used to estimate 30-yr avg SO2 concentration from
residential heating sources using dispersion models. Although a positive association of
SO2 and other pollutants (NO2, CO, PM10) with fatal MI was observed in this study, no
association between nonfatal MI and long-term SO2 exposure was reported. Panasevich et
al. (2013) reported higher TNF-a levels among those with a genetic polymorphism of a
TNF-a gene (TNF308G/A) as well as an increased risk of MI in the same population
(Section 5.3.2.5).
Weak or inverse associations of both cardiovascular and ischemic heart disease were
reported in a study relying on a Lagrangian particle dispersion model (see
Section 3.3.2.4) to estimate SOx emissions (gaseous and particulate component) from a
refinery (Ancona et al.. 2015). Exposure model performance statistics were not reported.
Null associations of cardiovascular hospitalizations with PM10, which was highly
correlated with SOx (r = 0.81) in this study, were observed. Because SOx was used as a
marker for refinery emissions, which contains multiple toxics including VOCs, the study
was not designed to evaluate the independent effect of SO2. In addition to the study by
Miller et al. (2007). which was included in the previous review, two analyses examined
the association of long-term SO2 exposure with a relatively broadly defined outcome that
included several cardiovascular diseases (Qin et al.. 2015; Dong et al.. 2013a). These
studies, which were conducted among Chinese adults, reported imprecise increases in the
risk of cardiovascular disease and results suggest the potential for age and body weight to
modify the association with long-term SO2 exposure. Neither of these analyses adjusted
for copollutant confounding despite the potentially high correlation of SO2 with PM2 5
(correlations reported with PM10 were 0.70). Further, the district-level SO2 concentrations
used to indicate exposure may not have adequately captured the spatial variability of
long-term SO2 exposure.
5-236
-------
Overall, these epidemiologic data do not provide support for an association of long-term
SO2 exposure with IHD or more broadly defined categories of cardiovascular disease.
There is uncertainty related the independent effect of SO2 on the cardiovascular system.
Further, the exposure assessment techniques applied in the studies were subject to
varying degrees of error depending on the method. Uncharacterized spatial variability in
the exposure estimate has the potential to bias the health effect estimate (Section 3.4.2V
5.3.2.3 Cerebrovascular Diseases and Stroke
Lipsett et al. (2011) evaluated the association of incident stroke with long-term exposure
to SO2, other gases (NO2, NOx, CO, ozone), and PM (Table 5-33). The authors found no
evidence of an association between SO2 and incident stroke [HR: 6.21 (95% CI: 0.4, 88)].
A positive association of SO2 with incident stroke of 1.13 (95% CI: 1.00, 1.34) per 5 ppb
was reported by Atkinson et al. (2013) in patients across England (study methods in
Section 5.3.2.2).
Two analyses of a random selection of adults (N = 24,845) ranging from 18 to 74 years
old from households in 33 Chinese communities examined the association between
long-term SO2 exposure and stroke. Monitor concentrations within each district were
used to derive 3-yr avg concentrations that were assigned to participants. The mean
concentration among study participants was 20 ppb. Dong et al. (2013a) reported an
increased risk of stroke [OR: 1.09 (1.01, 1.18) per 5 ppb] with the strongest associations
in males. Oin et al. (2015) further evaluated effect modification by obesity and reported
an increased risk of stroke among participants with body mass index (BMI) greater or
equal to 25 kg/m2 [OR: 1.18 (1.05, 1.32) per 5 ppb]. Neither of these studies considered
copollutant confounding despite moderate to high (r = 0.38 to 0.87) correlations with the
other pollutants that were evaluated (PM10, NO2, or ozone). The district level SO2
concentrations may not have adequately captured the spatial variability of SO2.
5-237
-------
Table 5-33 Epidemiologic studies of the association of long-term exposure to
sulfur dioxide with stroke.
Study
Cohort, Location, and Mean
Study Period ppb
Exposure
Assessment
Effect Estimates
(95% CI)
tLipsett et al. (2011)
California Teachers
Study Cohort
N = 124,614
(n = 56 stroke events)
California
Jun1996-
Dec 2005
S02
Geocoded
IQR:
residential
0.43
address linked
mean:
to pollutant
1.72
surface
developed
using IDW
(fixed-site
monitors
concentrations
from
1995-2005
used to model
exposure as a
time-dependent
function)
Correlation of
SO2 with:
ozone,
r= -0.17
PM2.5, r= 0.02
PM10, r= 0.54
NO2, r= 0.67
CO, r= 0.80
IQR:
Annual average
0,83
SO2
mean
concentration
(SD):
for 2002 at a 1
1.47
by 1-km
resolution
derived from
dispersion
models and
linked to
residential post
codes
Correlation of
S02with: NO2,
r= 0.86
Stroke incidence
SO2: HR 6.21 (0.4, 88)
Covariates: age, race,
smoking, second-hand
smoke, BMI, lifetime
physical activity,
nutritional factors,
alcohol, marital status,
menopausal status,
hormone replacement
therapy, hypertension
medication and aspirin,
and family history of
Ml/stroke
Copollutant adjustment:
none
tAtkinson etal. (2013)
National GP Patient
Cohort
England
2003
N = 836,557
(n = 13,956 stroke
events)
Stroke incidence
HR: 1.13 (1.00, 1.34)
Covariates: age, sex,
smoking, BMI, diabetes,
hypertension, and index
of multiple deprivation
Copollutant adjustment:
none
5-238
-------
Table 5-33 (Continued): Epidemiologic studies of the association of long term
exposure to sulfur dioxide with stroke
Study
Cohort, Location, and Mean Exposure
Study Period ppb Assessment
Effect Estimates
(95% CI)
tDong et al. (2013a)
N = 24,845
Random selection
(18-74 yr) from
households in
33 communities in
11 districts of
northeastern China
Mean:
20
median:
18
IQR:
7.5
3-yr avg
(2006-2008)
SO2
concentration
for each district
NO2, r= 0.38
O3, r= 0.87
PM10, r= 0.70
Prevalent stroke
OR: 1.09 (1.01, 1.18)
Note: associations
stronger among males
Covariate adjustment:
age, sex, educational
level, occupation, family
income, BMI,
hypertension, family
history of stroke, family
history of CVD, smoking
status, drinking, diet, and
exercise
tQin et al. (2015)
N = 24,845
Random selection
(18-74 yr) from
households in
33 communities in
11 districts of
northeastern China
Mean:
20.3
IQR:
7.5
3-yr avg
(2006-2008)
SO2
concentration
for each district
NO2, r= 0.38
Os, r= 0.87
PM10, r= 0.70
Stroke
BMI <25 kg/m2:
OR: 1.03 (0.92, 1.14)
BMI 25 kg/m2:
OR: 1.18 (1.05, 1.32)
Sex-stratified analyses
also presented
Covariate adjustment:
age, race, education,
income, smoking,
drinking, exercise, diet,
sugar, family history of
CVD or stroke, and
district
5-239
-------
Table 5-33 (Continued): Epidemiologic studies of the association of long term
exposure to sulfur dioxide with stroke
Study
Cohort, Location, and Mean Exposure
Study Period ppb Assessment
Effect Estimates
(95% CI)
tJohnson et al. (2010)
Edmonton, Alberta
Canada
Jan 2003-
Dec 2007
SO2 IDW average
mean: monitor SO2
1.3 concentration
Ecological analysis of
stroke incidence rates:
Stroke ED visits
Q1 RR: 1.0 (reference)
Q2 RR: 0.91 (0.83, 1.00)
Q3 RR: 0.89 (0.81, 0.98)
Q4 RR: 0.84 (0.73, 0.96)
Q5 RR: 0.93 (0.89, 0.98)
aResults for HS, non-HS,
and TIA also presented
Covariate adjustment:
age, sex, and household
income
Copollutant adjustment:
none
assigned at
postal code
centroid level
Correlation of
5-yr avg SO2
with:
NO2, r= 0.40
O3, r= 0.41
CO, r= -0.19
Avg = average; BMI = body mass index; CI = confidence interval; CO = carbon monoxide; CVD = cardiovascular disease;
ED = emergency department; GP = general practice; HR = heart rate; HS = hemorrhagic stroke; IDW = inverse distance
weighting; IQR = interquartile range; Ml = myocardial infarction; n = sample size; N = population number; N02 = nitrogen dioxide;
non-HS = nonhemorrhagic e; 03 = ozone; OR = odds ratio; PM25 = particulate matter with a nominal aerodynamic diameter less
than or equal to 2.5 |jm; PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm; Q1 = 1st
quartile; Q2 = 2nd quartile; Q3 = 3rd quartile; Q4 = 4th quartile; Q5 = 5th quartile r = correlation coefficient; RR = relative risk;
SD = standard deviation; S02 = sulfur dioxide; TIA = transient ischemic attack.
aEffect estimates are standardized per 5-ppb increase in S02 concentrations.
fStudies published since the 2008 ISA for Sulfur Oxides.
An inverse association between SO2 concentration and stroke incidence was observed in
an ecological analysis of long-term exposure to ambient pollution conducted in
Edmonton (Johnson et al.. 2010). while an association of SO2 with stroke prevalence was
observed in a study of 33 Chinese communities [OR: 1.21 (95% CI 1.01, 1.46)] (Dong et
al.. 2013a).
In summary, the epidemiologic studies do not provide evidence to support an effect of
long-term SO2 exposure on stroke morbidity. Findings are not generally consistent across
studies and there are uncertainties related to the potential for exposure measurement error
and confounding by copollutants.
Several cross-sectional analyses conducted in China where the mean long-term SO2
concentration is 18.7 ppb report positive associations with hypertension and increased
blood pressure. Dong et al. (2013d) observed an association with hypertension [OR: 1.17
(95% CI: 1.06, 1.28) per 5-ppb increase in SO2 concentration] among adults greater than
55 years of age in 33 Chinese communities. The absolute change in diastolic and systolic
5.3.2.4 Blood Pressure and Hypertension
5-240
-------
blood pressure in the study population overall was 0.46 mm Hg (95% CI: 0.15, 0.75) and
1.18 mm Hg (95% CI: 0.68, 1.69) per 5-ppb increase in SO2 concentration, respectively.
Zhao et al. (2013) reported a greater effect of SO2 on blood pressure among the
overweight and obese in this population. A similar trend was also observed with other
pollutants (i.e., ozone and NO2). In a study of children 5-17 years old from elementary
schools in seven Chinese cities, Dong et al. (2014) reported associations with arterial
blood pressure hypertension in males [OR: 1.17 (95% CI 1.08, 1.27)] and females [OR
1.19 (95% CI 1.10, 1.28)] per 5-ppb increase in 4-yr avg SO2 concentration. In an
extended analysis of this cohort, Dong et al. (2015) reported large associations for SO2
concentration in overweight and obese children. Although an array of risk factors was
considered in the analysis as potential confounders (Table 5-34). no adjustment for
copollutants was presented nor were copollutant correlations reported.
Table 5-34 Epidemiologic studies of the association of long-term exposure to
sulfur dioxide with hypertension.
Cohort, Location,
Mean
Exposure
Study
and Study Period
PPb
Assessment
Effect Estimates (95% CI)
tDonq et al. (2013d)
N =24,845
Mean:
3-yr avg
OR: 1.07 (1.03, 1.12)
Cross-sectional
Random selection
20.3
(2006-2008)SO2
(18-74 yr) from
households in
IQR: 7.5
concentration for
each district
SBP: 0.21 mm Hg (0.07,
0.34)
33 communities in
11 districts of
NO2, r= 0.38
DBP: 0.53 mm Hg (0.31,
0.76)
Covariate adjustment: age,
race, education, income,
smoking, drinking, exercise,
diet, sugar, family history of
hypertension, and district
northeastern China
O3, r= 0.87
PM10, r= 0.70
tZhao et al. (2013)
N =24,845
Mean:
3-yr avg
OR normal: 1.03
Cross-sectional
Random selection
20.3
(2006-2008)SO2
(0.99-1.08)
(18-74 yr) from
IQR: 7.5
concentration for
OR overweight: 1.10
households in
each district
(1.05-1.15)
33 communities in
OR obese: 1.10 (0.99-1.23)
11 districts of
northeastern China
NO2, r= 0.38
O3, r= 0.87
PM10, r= 0.70
Covariate adjustment: race,
education, income,
smoking, drinking, exercise,
diet, sugar, family history of
hypertension, and district
5-241
-------
Table 5-34 (Continued): Epidemiologic studies of the association of long term
exposure to sulfur dioxide with hypertension
Cohort, Location, Mean Exposure
Study and Study Period ppb Assessment Effect Estimates (95% CI)
Hypertension in males:
OR 1.17(1.08, 1.27)
Hypertension in females:
OR 1.19 (1.10, 1.28)
per 5 ppb
DPB (all children)
0.43 (0.26, 0.61)
SBP (all children)
0.71 (0.50, 0.91)
Normal weight:
0.89 (0.83, 0.96)
Overweight:
1.36 (1.18, 1.56)
Obese:
1.66 (1.46, 1.89)
per 5 ppb
Covariate adjustment: age,
sex, BMI, parental
education, LBW, premature
birth, income, passive
smoking exposure, home
coal use, exercise time,
area residence per person,
family history of
hypertension, and district
Avg = average; BMI = body mass index; CI = confidence interval; DPB = diastolic blood pressure; IQR = interquartile range;
LBW = low birth weight; n = sample size; N = population number; N02 = nitrogen dioxide; NR = not reported; 03 = ozone;
OR = odds ratio; PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm; r= correlation
coefficient; SBP = systolic blood pressure; S02 = sulfur dioxide.
fStudies published since the 2008 ISA for Sulfur Oxides.
5.3.2.5 Other Cardiovascular Effects
Few studies have evaluated other cardiovascular effects associated with long-term SO2
concentrations. Atkinson et al. (2013) examined the association of arrhythmias and heart
failure with long-term SO2 exposure. Study methods are described in Section 5.3.2.2.
Authors reported a positive association of SO2 with heart failure [HR: 1.27 (95% CI:
1.06-1.59) per 5 ppb] and with arrhythmia [HR: 1.13 (95% CI 1.00, 1.27)] in the fully
adjusted model. No association of annual SO2 concentration with hospital admissions for
heart failure was reported in a study of county-level air pollution indicator concentrations
(Bennett et al.. 2014).
tDona et al. (2014)
tDona et al. (2015)
Cross-sectional
N = 9,354 Mean:
Children (5-17 yr) 18.7.
Seven cities
northeastern China
2012-2013
4-yr avg
concentration for
one fixed-site
monitor within
1 km of
participant's home
Correlations NR
5-242
-------
5.3.2.6
Cardiovascular Mortality
The recent evidence for associations between long-term SO2 exposure and total mortality
(Section 5.5.2) is generally consistent with the evidence in the 2008 ISA for Sulfur
Oxides. Several studies report associations between long-term SO2 exposure and
cardiovascular mortality (Figure 5-27); however, there is no consistent trend toward
positive associations for cardiopulmonary or cardiovascular causes of death overall.
Additionally, confounding by copollutants is not ruled out (Section 3.4.3) and
uncertainties remain regarding the influence of exposure measurement error
(Section 3.4.2). Together, these uncertainties limit the interpretation of the causal nature
of the associations observed in the available epidemiologic studies of long-term
mortality.
5.3.2.7 Subclinical Effects Underlying Cardiovascular Diseases
Carotid intima-media thickness (cIMT) is a measurement of thickness of the inner layers
of the wall of the artery and can be used to indicate the presence of subclinical
atherosclerosis. Other markers of preclinical atherosclerosis include pulse wave velocity
and augmentation index, both of which indicate arterial stiffening. In an analysis of the
Atherosclerosis Risk in Young Adults study, which is a prospective cohort study (Lenters
et al.. 2010). no association of SO2 concentration with cIMT was observed; however,
there was a weak increase in aortic pulse wave velocity reported. SO2 concentration at the
home address for the year 2000 was assigned to participants of this study. The
correlations of SO2 with NO2, black smoke, and PM2 5 reported in this study were low,
ranging from r = 0.09 to 0.12. The correlations of SO2 with metrics of traffic intensity
were also low (r = -0.06 to 0.06). In another study, Weng et al. (2015) reported that
annual average SO2 concentration was correlated with brachial-ankle pulse wave velocity
in univariate analyses but not after adjustment for PM10 and other potential confounders.
This study was based on data from 127 heart disease patients undergoing hemodialysis in
Taoyuan, Taiwan.
Inflammation and oxidative stress have been shown to play a role in the progression of
chronic cardiovascular disease. Forbes et al. (2009b) examined the association of
predicted annual average SO2 concentration with CRP and fibrinogen among the English
population. Multilevel linear regression models were used to determine pooled estimates
across three cross-sectional surveys conducted during different years. Each participant's
postal code of residence was linked to predicted annual average SO2 concentration
derived from dispersion models. SO2, was not associated with increased CRP or
fibrinogen in these data. A study conducted among men and women (45-70 years) in
5-243
-------
Stockholm reported an association of 30-yr avg source-specific heating-related SO2
concentration, estimated using dispersion models, with increases in IL-6; however, SO2
was not associated with CRP, TNF-a, fibrinogen, or plasminogen activator inhibitor-1 in
this study (Panasevich et al.. 2009). Associations between long-term NO2 concentration,
which were moderately correlated with SO2 (r = 0.53), and increased plasma IL-6 were
also observed in this study. A study conducted among older adults in Taiwan reported no
changes in blood pressure, total cholesterol, fasting glucose, hemoglobin Ale, IL-6, and
neutrophils in association with increasing SO2 concentration while associations between
these endpoints and other pollutants were observed (Chuang et al.. 2011).
Overall, the body of evidence is limited and there is no consistent positive trend in the
associations observed between SO2 and subclinical atherosclerosis or circulating markers
of inflammation. These findings are consistent with the general lack of mechanistic
evidence for key events in the proposed mode of action leading to extrapulmonary
effects.
5.3.2.8 Summary and Causal Determination
Overall, the evidence is inadequate to infer a causal relationship between long-term
exposure to SO2 and cardiovascular health effects.
Although a number of epidemiologic studies report positive associations between
long-term exposure to SO2 concentrations and cardiovascular disease and stroke
(Section 5.3.2.3). the evidence for any one outcome is limited and inconsistent. As
discussed in Section 3.4.2.2. fixed-site monitors may not capture the spatial variability in
SO2 concentration. Dispersion models generally capture SO2 variability on near-source
spatial scales (up to tens of km) but exposure estimates from such models are subject to
other uncertainties (Section 3.3.2.4). Bias stemming from exposure measurement error
can be either toward or away from the null. No studies corrected for such error,
complicating the interpretation of findings from studies of long-term exposure of SO2
(Section 3.4.4.2). There is also uncertainty regarding the potential for copollutant
confounding (Section 3.4.3). Primary pollutants such as NO2 and CO typically show
moderate to high correlations with SO2 (Tables 5-32. 5-33. and 5-34). and there is a lack
of experimental evidence to provide coherence or biological plausibility for an
independent effect of SO2 on cardiovascular health. Several epidemiologic studies
evaluated the association between SO2 concentration and subclinical atherosclerosis or
circulating markers of inflammation; however, there is no consistent positive trend in the
associations observed between SO2 and these potential key events in a mode of action.
5-244
-------
The available evidence examining the relationship between long-term exposure to SO2
and cardiovascular effects was evaluated using the framework described in Tables I and
II of the Preamble to the ISAs (U.S. EPA. 2015b). The key evidence, supporting or
contradicting, as it relates to the causal framework is summarized in Table 5-35. In
conclusion, the evidence lacks coherence and is of insufficient consistency, and thus,
is inadequate to infer a causal relationship between long-term exposure to S02 and
cardiovascular health effects.
Table 5-35 Summary of evidence, which is inadequate to infer a causal
relationship between long-term sulfur dioxide exposure and
cardiovascular effects.
Rationale for Causal
Determination3
Key Evidence13
Key References'3
SO2
Concentrations
Associated
with Effects0
Some epidemiologic studies
report positive associations but
results are not generally
consistent.
Positive associations of SO2
with Ml, CVD events, or stroke
events
tLipsett et al. (2011)
1.72 ppb (mean)
tAtkinson et al. (2013)
1.47 ppb (mean)
Miller etal. (2007)
NR
Null/inverse associations
observed with Ml and stroke
Rosenlund et al. (2006)
9.6 ppb
(median)
tJohnson et al. (2010)
1.3 ppb (mean)
Limited coherence with
evidence for cardiovascular
mortality
No consistent positive trend
observed in long term studies
of cardiovascular mortality.
Section 5.3.2.4
Uncertainty due to confounding
by correlated pollutants
Correlations of SO2 with CO
and NO2 vary by location but
are generally moderate to high.
Section 3.4.3
Tables 5-32. 5-33. 5-34
5-245
-------
Table 5-35 (Continued): Summary of evidence, which is inadequate to infer a
causal relationship between long term sulfur dioxide
exposure and cardiovascular effects.
SO2
Concentrations
Rationale for Causal
Associated
Determination3
Key Evidence13
Key References'3 with Effects0
Uncertainty due to exposure
Centrally located monitors may
Miller et al. (2007)
measurement error
not capture spatial variability of
Section 3.4.2
SO2 concentrations.
SO2 estimates from dispersion
tAtkinson et al. (2013)
models in specific studies show
tForbes et al. (2009a)
poor to moderate agreement
with measured concentrations.
Exposure measurement error
Section 3.4.4.2
can introduce bias toward or
away from the null in studies of
long-term exposure
Uncertainty due to lack of
Lack of experimental human or
coherence with other lines of
animal studies evaluating
evidence
cardiovascular effects of
long-term SO2 exposure
Weak evidence to identify key
Limited of mechanistic
Sections 4.3. 5.3.2.7
events in the mode of action
evidence for key events leading
to extrapulmonary effects.
Limited and inconsistent
evidence of increased
subclinical atherosclerosis and
systemic inflammation
(e.g., IL-6, CRP) in
epidemiologic studies.
CO = carbon monoxide; CRP = C-reactive protein; CVD = cardiovascular disease; IL-6 = interleukin-6; Ml = myocardial infarction;
N02 = nitrogen dioxide; NR = not reported; S02 = sulfur dioxide.
aBased on aspects considered in judgments of causality and weight of evidence in causal framework in Tables I and II of the
Preamble to the ISAs (U.S. EPA. 2015b).
bDescribes the key evidence and references, supporting or contradicting, that contribute most heavily to causal determination.
References to earlier sections indicate where full body of evidence is described.
°Describes the S02 concentrations with which the evidence is substantiated.
fStudies published since the 2008 ISA for Sulfur Oxides.
5.4 Reproductive and Developmental Effects
5.4.1 Introduction
This section covers studies of health endpoints with exposures to SO2 occurring during or
around pregnancy and/or the first years of life. This includes not only pregnancy and
5-246
-------
birth outcomes (including infant mortality) occurring close in time to the exposure, but
also developmental outcomes potentially occurring years later. Exposures occurring in
pregnancy and early life may alter development, and have effects not immediately
identifiable but evident at later points. These studies are characterized in this section as
they contribute to the weight of evidence for effects of SO2 on reproductive health and
development. Evidence regarding fertility, reproduction, and pregnancy are discussed in
Section 5.4.2. with a series of birth outcomes [fetal growth (Section 5.4.3.1). preterm
birth (PTB) (Section 5.4.3.2). birth weight (Section 5.4.3.3). birth defects
(Section 5.4.3.4). fetal mortality (Section 5.4.3.5). and infant mortality (Section 5.4.3.6)1
discussed in Section 5.4.3. Studies of developmental outcomes are discussed in
Section 5.4.4. with a focus on respiratory developmental outcomes in Section 5.4.4.1.
The 2008 SOx ISA (U.S. EPA. 2008d) concluded the evidence was inadequate to infer a
causal relationship with reproductive and developmental effects. Epidemiologic studies
included in the 2008 SOx ISA (U.S. EPA. 2008d) examined impacts on reproductive
outcomes including preterm birth, birth weight, intra-uterine growth retardation, birth
defects, infant mortality, and neonatal respiratory hospitalizations. While positive
associations were observed in the previous SOx ISA (U.S. EPA. 2008d). there was little
biologic plausibility for these associations provided by supporting toxicological literature.
Interpretation of those results was also limited by the lack of control for potential
confounding by copollutants, the small number of studies, and uncertainty regarding
exposure. Overall, the number of studies examining associations between exposure to
ambient SO2 and reproductive and developmental outcomes has increased substantially
since publication of the 2008 SOx ISA, yet evidence for an association with individual
outcomes remains relatively limited and key uncertainties have not been reduced.
The body of literature characterizing the reproductive health effects of exposure to SO2
has grown considerably since the 2008 SOx ISA (U.S. EPA. 2008d). with over 50 recent
epidemiologic studies. However, the number of studies for any particular outcome
remains relatively limited. Among the recent epidemiologic studies, birth outcomes
(e.g., small for gestational age, preterm birth, and birth weight) predominate. Several new
studies of congenital anomalies are now available in addition to the single study included
in the 2008 SOx ISA. Recent studies of other outcomes, such as fetal mortality, infant
mortality, fertility, and conditions related to pregnancy have also been published. Key
epidemiologic studies are summarized in Table 5-36. In toxicological research, a single
study published at relevant exposure levels (1,500 ppb or lower) investigated
reproductive and developmental changes in exposed female rats and their offspring,
finding altered estrus cyclicity with fewer cycles overtime, altered birth outcomes of
increased litter size, and decreased postnatal body weight in offspring whose dams were
exposed to S02. This study is summarized in Table 5-37. The majority of the remaining
5-247
-------
animal toxicological evidence for reproductive and development effects is for exposure at
5,000 ppb or greater, doses which are beyond the scope of this document.
Table 5-36 Key reproductive and developmental epidemiologic studies for sulfur
dioxide.
Study
Location
(Sample Size)
Years
Mean SO2
ppb
Exposure
Assessment
Selected Effect Estimates3
(95% CI)
Fetal growth
Liu et al. (2003)
Vancouver
(n = 229,085)
1986-1998
4.9 Monitors at census
subdivision level
IUGR (those with birth weight fall
below the 10th percentile, by sex
and gestational week, of all
singleton live births in Canada
between 1986 and 1998, term)
M1: 1.07 (1.01, 1.13)
Last mo: 1.00 (0.94, 1.
.06)
T1
T2
T3
: 1.07 (1.00, 1.14)
: 0.98 (0.91, 1.04)
: 1.03 (0.96, 1.10)
Robust to NO2, CO, and O3 in
copollutant models
Brauer et al. (2008) Vancouver
(n = 70,249)
1999-2002
2.2 Inverse distance
weighting of three
closest monitors
within 50 km, 14 SO2
monitors
SGA (those with birth weights below
the 10th percentile of the cohort,
stratified by sex, for each week of
gestation)
EP: 1.02 (1.00, 1.03)
No copollutant models'5
Rich et al. (2009)
New Jersey
(n = 178)
1999-2003
T1
T2
T3
5.7
5.6
5.5
Nearest monitor
(within 10 km)
VSGA (growth ratio <0.75)
T1: 1.00 (0.92, 1.08)
T2: 1.04 (0.96, 1.13)
T3: 1.05 (0.97, 1.14)
No copollutant models'5 for SO2
5-248
-------
Table 5-36 (Continued): Key reproductive and developmental epidemiologic
studies for sulfur dioxide
Study
Location
(Sample Size)
Years
Mean SO2
PPb
Exposure
Assessment
Selected Effect Estimates3
(95% CI)
tLe etal. (2012)
Detroit, Ml
(n = 112,609)
1990-2001
5.8 Nearest monitor (ZIP
code within 4 km of
one of three
monitors)
SGA (infants whose birth weights
fell below the 10th percentile by sex
and gestational week, based on
study population's distribution, term)
T1, adjusted for CO, NO2, and PM10
reference
1.18 (0.92, 1.51)
1.01 (0.83, 1.23)
1.05 (0.87, 1.28)
T2, adjusted for CO, NO2, and PM10
Q1: reference
1.30 (1.01, 1.69)
1.12 (0.91, 1.37)
1.11 (0.90, 1.36)
adjusted for CO, NO2, and PM10
reference
1.17 (0.94, 1.45)
1.24 (1.02, 1.50)
1.31 (1.06, 1.60)
No copollutant models'5
Q1
Q2
Q3
Q4
Q2
Q3
Q4
T3,
Q1
Q2
Q3
Q4
Preterm birth
Liu et al. (2003)
Vancouver, BC
(n = 229,085)
1986-1998
4.9 Monitors at census
subdivision level
M1: 0.95 (0.88, 1.03)
Last mo: 1.09 (1.01, 1.19)
Robust to NO2, CO, and O3 in
copollutant models
Saaiv et al. (2005)
Pennsylvania
(n = 187,997)
1997-2001
7.9 Monitors at county
level
Last 6 wk: 1.05 (1.00, 1.10)
3-day lag: 1.02 (0.99, 1.05)
No copollutant models'5
tZhao etal. (2011)
Guangzhou, China
(n = 7,836 preterm
births)
2007
20 City average from
monitors
Same day: 1.04 (1.02, 1.06)
1-day lag: 1.01 (0.99, 1.04)
2-day lag: 1.02 (0.99, 1.04)
3-day lag: 1.02 (0.99, 1.04)
Robust to NO2 and PM10 in
copollutant models
tMendola et al.
(2016a)
U.S.
(n = 223,502)
2002-2008
3.99 CMAQ-monitor fused
surface with inverse
distance weighting,
spatial modeling
scales not provided;
delivery hospital
referral region
Week 34
Asthma: 1.32 (1.05, 1.70)
No asthma: 1.02 (0.90, 1.14)
Week 35
Asthma: 1.17 (1.02, 1.34)
No asthma: 0.98 (0.92, 1.05)
Last 6 wk of pregnancy
Asthma: 0.90 (0.81, 1.00)
No asthma: 0.81 (0.77, 0.92)
EP
Asthma: 0.93 (0.83, 1.03)
No asthma: 0.92 (0.87, 0.97)
No copollutant models'5
5-249
-------
Table 5-36 (Continued): Key reproductive and developmental epidemiologic
studies for sulfur dioxide
Study
Location
(Sample Size)
Years
Mean SO2
PPb
Exposure
Assessment
Selected Effect Estimates3
(95% CI)
Low birth weight
Ha et al. (2001)
Seoul, South Korea
(n = 276,763)
1996-1997
T1: 13 Monitors averaged to
T3: 12 city
T1: 1.05 (1.02, 1.08)
T1, adjusted forT3:
1.06 (0.98, 1.13)
T3: 0.96 (0.92, 0.99)
T3, adjusted forT1:
1.02 (0.94, 1.10)
No copollutant models'5
Lee et al. (2003)
Seoul, South Korea
(n = 388,105)
1996-1998
12.1
Monitors averaged to
city
EP: 1.02 (0.99, 1.05)
T1: 1.05 (1.02, 1.09)
T2: 0.97 (0.92, 1.00)
T3: 1.12 (1.03, 1.20)
No copollutant models'5
Liu et al. (2003)
Vancouver, BC
(n = 229,085)
1986-1998
4.9
Monitors at census
subdivision level
M1: 1.11 (1.01, 1.22)
Last mo: 0.98 (0.89, 1.08)
Robust to NO2, CO, and O3 in
copollutant models
Duqandzic et al.
(2006)
Nova Scotia
(n = 74,284)
1988-2000
10
Nearest monitor
(postcode within
25 km)
T1: 1.20 (1.05, 1.38)
T2: 0.99 (0.91, 1.09)
T3: 0.95 (0.86, 1.04)
No copollutant models'5
tMorello-Frosch et
al. (2010)
California
(n = 3,545,177)
1996-2006
2.1
Nearest monitor
(census block
centroid within 3, 5,
or 10 km)
EP
3 km: 1.10 (0.95, 1.34)
5 km: 1.05 (0.95, 1.16)
10 km: 1.05 (1.00, 1.10)
No copollutant models'5
tEbisu and Bell
(2012)
Northeastern and
mid-Atlantic U.S.
(n = 1,207,800)
2000-2007
6.1
County average from
monitors
EP: 1.05 (1.01, 1.09)
No copollutant models'5
tKumar (2012)
Chicago, IL
(n = 398,120)
2000-2004
Nearest
monitor:
4.7
County
average:
4.6
Nearest monitor
(census tract within
3 miles)
County average from
monitors
Nearest monitor
EP: 1.19 (0.90, 1.57)
County average
EP: 1.05 (0.91, 1.20)
No copollutant models'5
5-250
-------
Table 5-36 (Continued): Key reproductive and developmental epidemiologic
studies for sulfur dioxide
Study
Location
(Sample Size)
Years
Mean SO2
PPb
Exposure
Assessment
Selected Effect Estimates3
(95% CI)
Birth Weight
tDarrow et al.
(2011)
Distributed lag, 1-h
max SO2
Atlanta, GA
(n = 400,556)
1994-2004
M1: 10.7
T3: 9.5
Population weighted
spatial model based
on monitors,
five-county area, 1-h
max
M1: 0.625 (-2.625, 3.875)
T3: -6.500 (-12.500, -0.667)
Non-Hispanic white
T3: -8.667 (-15.333, -2.000)
Non-Hispanic black
T3: -3.167 (-9.833, 3.667)
Hispanic
T3: -9.5 (-19.000, -0.167)
No copollutant models'5
tGeeretal. (2012)
Texas
(n = 1,548,904)
1998-2004
2.3
County average from
monitors
EP: -15.594 (-25.344, -5.844)
Robust to NO2 and O3 in copollutant
models
Fetal and infant mortality
Inverse distance Among preterm deliveries
weighting of monitors EP: 1.16 (1.00, 1.34)
to township or district, M1: 1.22 (1.00, 1.34)
72 monitors M2: 1.22 (1.00, 1.34)
M3: 1.16 (1.00, 1.34)
Among term deliveries
EP: 0.95 (0.82, 1.10)
M1: 1.00 (0.90, 1.16)
M2: 1.00 (0.90, 1.16)
M3: 0.95 (0.86, 1.16)
No copollutant models'5
EP: 1.32 (0.95, 1.84)
T1: 1.23 (1.02, 1.51)
T2: 1.21 (0.89, 1.53)
T3: 1.47 (1.05, 1.69)
No copollutant models'5
2-day lag
1.12 (1.02, 1.24)
Adjusted PM2.5: 1.18 (1.00, 1.40)
Adjusted NO2: 1.15 (1.00, 1.32)
Adjusted CO: 1.05 (0.93, 1.20)
No copollutant models'5
tHwanq et al. Taiwan 5.7
(20111 (n = 9,325 cases)
2001-2007
TFaiz et al. (2012) New Jersey 5.9 Nearest monitor
(n = 994) (within 10 km, 1 of
16 monitors)
1998-2004
TFaiz et al. (2013) New Jersey 5.8 Nearest monitor
(n = 1 277) (within 10 km, 1 of
16 monitors)
1998-2004 '
5-251
-------
Table 5-36 (Continued): Key reproductive and developmental epidemiologic
studies for sulfur dioxide
Study
Location
(Sample Size)
Years
Mean SO2
PPb
Exposure
Assessment
Selected Effect Estimates3
(95% CI)
Woodruff et al.
U.S.
3
Monitors, averaged to
All causes
(2008)
(n = 6,639 cases)
(median)
county
0.93 (0.84, 1.04)
1999-2002
Exposures for 2 mo
after birth
Respiratory
1.09 (0.89, 1.36)
Adjusted PM10, CO, O3:
1.13 (0.79, 1.60)
Adjusted PM2.5, CO, O3:
1.21 (0.79, 1.84)
No copollutant models'5
Developmental
Dales et al. (2006)
Atlanta, GA
(n = 8,586 cases)
1986-2000
4.3 Monitors, averaged to
city
Neonatal hospitalization for
respiratory disease
2-day lag
2.59 (1.05, 4.39)
Adjusted for O3, NO2, CO
1.95 (0.54, 3.68)
Adjusted for O3, NO2, CO, PM10
1.57 (0.25, 3.29)
No copollutant models'5
tClark et al. (2010)
British Columbia
(n = 3,482 cases)
1999-2000
Inverse distance
weighting 3 nearest
monitors (of 14)
within 50 km
Asthma
EP: 1.45 (1.28, 1.84)
1st year of life: 1.45 (1.28, 1.84)
No copollutant models'5
CI = confidence interval; CMAQ = Community Multiscale Air Quality; CO = carbon monoxide; EP = entire pregnancy;
IUGR = intra-uterine growth restriction; max = maximum; M1 = Month 1; M2 = Month 2; M3 = Month 3; n = sample size;
N = population number N02 = nitrogen dioxide; 03 = ozone; PM2.5 = particulate matter with a nominal aerodynamic diameter less
than or equal to 2.5 |jm; PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm; Q1 = 1st
quartile; Q2 = 2nd quartile; Q3 = 3rd quartile; Q4 = 4th quartile; SGA = small for gestational age; S02 = sulfur dioxide; T1 = 1st
trimester; T2 = 2nd trimester; T3 = 3rd trimester; VSGA = very small for gestational age.
aRelative risk per 5-ppb change in S02, unless otherwise noted.
bStudy did not include two-pollutant model results (i.e., S02 and one other air pollutant), but may have included multipollutant
model results (i.e., model that includes 3 or more air pollutants).
fStudies published since the 2008 ISA for Sulfur Oxides.
5-252
-------
Table 5-37 Study specific details from animal toxicological studies of the
reproductive and developmental effects of sulfur dioxide.
Study and Species
Concentration SO2 Exposure
Measured Outcome(s)
Mamatsashvili (1970b)
0.057 or 1.5 ppm for 72 days
Estrus cyclicity duration (F0 and F1),
Rat
litter size, offspring growth (body
weight)
S02 = sulfur dioxide.
Several recent articles have reviewed methodological issues relating to the study of
outdoor air pollution and adverse birth outcomes (Chen et al.. 2010a; Woodruff et al..
2009; Ritz and Wilhelm. 2008; Slama et al.. 2008). Some of the key challenges to
interpretation of birth outcome study results include: (1) the difficulty in assessing
exposure as most studies use existing monitoring networks to estimate individual
exposure to ambient air pollution; (2) the need for detailed exposure data and potential
residential movement of mothers during pregnancy; (3) the inability to control for
potential confounders such as other risk factors that affect birth outcomes (e.g., smoking),
evaluating the exposure window (e.g., trimester) of importance; and (4) the limited
evidence on the physiological modes of action for these effects (Ritz and Wilhelm. 2008;
Slama et al.. 2008). An additional limitation is the failure for many studies of
reproductive and developmental outcomes to adjust for co-occurring air pollutants. As
ozone, PM2 5, and NOx have all been associated with reproductive and developmental
health outcomes, the lack of adjustment makes interpretation of isolated SO2 effects more
difficult.
5.4.2 Fertility, Reproduction, and Pregnancy
Infertility affects approximately 11% of all women ages 15-44 in the U.S. (Chandra et
al.. 2013). and can have negative psychological impacts and affect quality of life;
infertility and subfertility may also potentially signal poorer physiological health. Those
with fertility problems are at higher risk for adverse pregnancy and birth outcomes if they
do become pregnant (Hansen et al.. 2005; Helmerhorst et al.. 2004; Jackson et al.. 2004).
Outcomes studied in this area include fecundity (the ability to conceive frequently,
quantified as length of time to pregnancy) and fertility (the ability to have a live birth).
Studies in this area frequently use populations undergoing assisted reproductive
treatment, as these populations have a large amount of data collected on them during
5-253
-------
treatment and defined menstrual cycles and start points. In cohorts recruited from the
general population, exact timing can be difficult to determine due to reliance on
participant recall, particularly if they are surveyed well after initiation of pregnancy
attempts. Many pregnancies are unplanned, which also adds a level of complication to
quantifying fertility. Researchers may also investigate potential mechanistic links
between pregnancy conditions and biomarkers and later birth outcomes; such as
pregnancy-related hypertension, which is a leading cause of perinatal and maternal
mortality and morbidity (Lee et al.. 2012V
Four recent studies have examined the effects of SO2 on measures of fertility; all use
different populations and outcomes and observed mainly null effects for SO2 exposures.
Recent studies examined semen quality parameters in cohorts of men from Chongqing,
China (Zhou et al.. 2014) and Poland (Radwan et al.. 2015) and observed decreases in
normal morphology with increases in SO2 exposure; however, all other quality metrics
showed null associations. Slama et al. (2013) examined fecundity rate ratios (FRs) with
SO2 exposures before and after the initiation of unprotected intercourse in a Czech
Republic population. Exposures prior to intercourse initiation (long-term, -30 or 60 days)
had slightly reduced FRs; however, SO2 was highly correlated with PM2 5 and NO2 in this
population (i.e., r > 0.7) and stronger reductions in fertility were observed with those
pollutants. Legro et al. (2010) examined odds of live birth in a population undergoing
in vitro fertilization and observed null associations for SO2 with all exposure windows
from medication start to birth (short-term windows during in vitro fertilization, long term
from transfer to pregnancy).
Mixed effect estimates are observed with SO2 exposure across other pregnancy-related
outcomes. Recent studies examined increased blood pressure during pregnancy or
pregnancy-related hypertensive disorders, including pre-eclampsia. Several studies
observed no associations between SO2 exposure during the first trimester and changes in
late pregnancy blood pressure (Lee et al.. 2012) or hypertensive disorders (Michikawa et
al.. 2015); however, a study in Florida observed increased hypertension with higher SO2
exposure during the first trimester (Xu et al.. 2014). Mendola et al. (2016b) observed a
positive association between pre-eclampsia and SO2 exposure among people with asthma,
but not among people without asthmas; the interaction between exposure to SO2 and
asthma was statistically significant for the first trimester exposure window. A small
Iranian study found no association between pre-eclampsia and SO2 above versus below
median concentrations (Nahidi et al.. 2014). Assibev-Mensah et al. (2015) observed no
effect of SO2 on hypertensive disorders in Beijing comparing 2008 Olympic period with
same calendar days in 2009. In fact, there was an inverse relationship between SO2
exposure in the third trimester and hypertensive disorders.
5-254
-------
In other pregnancy-related outcomes, no associations were observed in the Allegheny
County, PA population for short-term near-birth exposures and C-reactive protein, an
inflammatory biomarker linked to increased risk of preterm birth (Lee et al.. 201 lb).
Michikawa et al. (2016) observed positive associations with SO2 exposure and placenta
previa in a Japanese population, although the associations were smaller and less
consistent than those observed for ozone or suspended particulate matter (SPM).
Increases in SO2 exposure during the preconception period and the first trimester were
associated with increased odds of gestational diabetes mellitus (Robledo et al.. 2015).
Assibev-Mensah et al. (2015) examined other fetal-placental conditions, and observed no
associations with SO2 exposure in the first or second trimester, but reported a positive
association with fetal-placental conditions and third trimester SO2 exposures in Beijing
comparing the 2008 Olympic period with the same calendar days in 2009. As part of the
Consortium on Safe Labor, a retrospective cohort study that included 16 counties across
the U.S., Wallace et al. (2016) observed positive associations between premature rupture
of membranes and SO2 exposure averaged over the whole pregnancy, but not for shorter
exposure windows (i.e., days or hours before rupture).
No recent animal studies evaluating fertility and pregnancy were identified. An older
study in laboratory animals exposed to SO2 demonstrated reproductive toxicity in adult
female rodents and their offspring. Adult female albino rats were exposed to either
0.057 ppm or 1.5 ppm SO2 by inhalation for 72 days lYMamatsashvili. 1970b); see
Table 5-371. During the first month of treatment at 1.5 ppm, substantial alterations in
stages of the estrus cycle were seen including significant decreases in duration of diestrus
and metastrus. During the 2nd and 3rd month of exposure, prolongation of estrus
cyclicity was found with exposure to 1.5 ppm SO2, leading to fewer estrus cycles during
the study period. This change was not permanent as by 7 months after exposure ceased,
estrus cyclicity returned to normal. Exposure of adult female rodents to SO2 caused
disruption of estrus cyclicity that was not permanent as it returned to normal after
cessation of SO2 exposure.
While studies of fertility, reproduction, and pregnancy are limited in number, generally,
SO2 exposures appear to have no association with these outcomes. A group of studies
examining hypertensive disorders during pregnancy report inconsistent results, with the
majority observing no association with SO2 exposure. Similarly, studies examining
endpoints related to fertility and other pregnancy conditions are generally inconsistent,
with the majority observing no association, and few studies examining any one specific
outcome. Additionally, these studies do not provide evidence to help reduce uncertainty
related to exposure measurement error, copollutant confounding, or biological
mechanism by which SO2 could cause these effects. These studies are summarized in
Supplemental Table 5S-22 (U.S. EPA. 2017c).
5-255
-------
5.4.3
Birth Outcomes
This section discusses several categories of birth outcomes, including fetal growth
(Section 5.4.3.1). preterm birth (Section 5.4.3.2). birth weight (Section 5.4.3.3). birth
defects (Section 5.4.3.4). fetal mortality (Section 5.4.3.5). and infant mortality
(Section 5.4.3.6).
5.4.3.1 Fetal Growth
Fetal growth can be difficult to quantify; typically, small for gestational age (SGA) or
intra-uterine growth restriction (IUGR) are used. These designations, often used
interchangeably, are defined as infants with a birth weight below the 10th percentile for
gestational age, usually with consideration for sex and race as well. There are a number
of limitations in using SGA/IUGR as a metric of poor fetal growth. One is that a
percentile-based measurement will always quantify a certain percentage of the infant
population as growth restricted whether or not this is truly the case (Wollmann. 1998).
For example, in term infants, it is unlikely that 10% are actually growth restricted.
Whereas in preterm infants, it is likely that more than 10% are growth restricted;
therefore, SGA cases would be overestimated in term infants and underestimated in
preterm infants. In addition, exact definitions shift between studies and some studies use
alternate definitions of SGA/IUGR. For example, some studies use the birth weight
distribution of their study population for defining SGA (Le et al.. 2012). which will
naturally not be identical for every study population, and others use country standards,
likely to be more stable over time (Brauer et al.. 2008; Liu et al.. 2003). An alternate
approach to categorizing growth restriction is to use ultrasound images during gestation
(Woodruff et al.. 2009). This approach has the benefit of examining all fetuses with
ultrasounds, being less subjective to population definition, and distinguishing true growth
restriction from merely small-sized infants. However, not all women receive prenatal care
and ultrasounds, leading to the possibility of selection bias.
Several studies report positive associations between fetal growth and SO2, although
timing of exposure is inconsistent. A recent study conducted in Australia examined
ultrasound measures in midgestation in association with SO2 exposures during early
pregnancy (Hansen et al.. 2008). Hansen et al. (2008) observed decreases in biparietal
diameter and abdominal circumference with increases in SO2 during the first 4 months of
pregnancy [5-ppb SO2 increase in the 1st month: -4.25 mm (-6.81, -1.69) biparietal
diameter; -9.31 mm (-19.31, 0.69) abdominal circumference]. Recent studies using the
traditional definition of SGA/IUGR had mixed results. In Vancouver, increases in ORs
for SGA were observed with entire pregnancy exposures (Brauer et al.. 2008) and with
5-256
-------
1st month and 1st trimester exposures (Liu et al.. 2003). Rich et al. (2009) used an
alternate definition of SGA—having a growth ratio (infant birth weight divided by
median study cohort birth weight) below 0.75 for very SGA (VSGA), and between
0.75-0.85 for SGA—and observed elevated ORs with 1st trimester exposures for SGA,
and 2nd and 3rd trimester exposures for VSGA. Other studies did not observe positive
associations between fetal growth and SO2. In a study conducted in Italy, (Capobussi et
al.. 2016) observed a null association for SGA when SO2 exposure was estimated for the
entire pregnancy, but modest positive associations when exposure was averaged across
the first or second trimester. In a study conducted in Calgary, Edmonton, and Montreal,
Liu et al. (2007) found lowered ORs for IUGR with exposures in Months 1 to 5 of
pregnancy and no associations in Months 6 to 9. Of the two recent studies in the U.S., Le
et al. (2012) observed generally null associations for SGA and 1st and last month
exposures; ORs with trimester exposure windows were null, although ORs became
elevated for the 2nd and 3rd trimesters after adjustment for CO, NO2, and PM10.
In summary, there is inconsistent evidence for increased odds of fetal growth restriction
with exposure to SO2 during pregnancy, and the evidence lacks consistency in fetal
growth definition/metric and in exposure timing. Mean SO2 exposures for these studies
are generally low, although all studies examine average daily SO2 concentrations.
Additionally, these studies do not provide evidence to help reduce uncertainty related to
exposure measurement error, copollutant confounding, or the biological mechanism by
which SO2 could cause these effects. No recent animal studies evaluating fetal growth
were identified. Studies examining the association between SO2 and fetal growth can be
found in Supplemental Table 5S-23 (U.S. EPA. 2017c).
5.4.3.2 Preterm Birth
Preterm birth, delivery that occurs before 37 weeks of completed gestation, is a marker
for fetal underdevelopment and a risk factor for further adverse health outcomes
(e.g., infant mortality, neurodevelopmental problems, growth issues) (Mathews and
MacDorman. 2010; Saigal and Dovle. 2008; IOM. 2007; Gilbert et al.. 2003). PTB is
characterized by multiple etiologies (spontaneous, premature rupture of membranes, or
medically induced), and identifying exact causes of PTB is difficult. It is likely that some
mechanistic pathways are shared among the three groups; however, isolated causes are
also likely to exist. Few, if any, studies distinguish among these three groups in
examining associations between air pollution and PTB.
Given the uncertainty surrounding modes of action leading to PTB, many of the studies
reviewed here consider both short- and long-term exposure periods. For example,
5-257
-------
exposure periods may include all of gestation or a particular trimester for long-term
exposure windows, or weeks or days leading up to birth for short-term exposure
windows. With near-birth exposure periods, development will be at different points for
term and preterm infants (e.g., exposure 2 weeks before birth is at 34 weeks for a
36-week PTB, and 38 weeks for a 40-week term birth), which suggests the possibility of
different modes of action for increases in risk observed with near-birth exposures
compared to exposures in specific periods of fetal development.
There is evidence supporting a relationship between SO2 and preterm birth, primarily
with exposure near-birth and including both older and newer studies. Among a U.S. birth
cohort, Mendola et al. (2016a) examined PTB and exposure to SO2 during different
periods before and during pregnancy, observing generally null results among women with
and without asthma, except for when exposure was limited to weeks near birth
(specifically Weeks 34 and 36) for which positive associations were observed among
women with asthma, but not for women without asthma. Studies in Europe and Asia
reported increased ORs/RRs of PTB with exposures across pregnancy, although not
consistently between studies (Dibben and Clemens. 2015; Zhao etal.. 2011; Leem et al..
2006; Bobak. 2000; Xu et al.. 1995). In a recent time-series analysis, Zhao etal. (2011)
found increased RRs with SO2 exposure lagged 0-3 days from birth, but SO2 was also
highly correlated with PM10 (Pearson correlation coefficient = 0.75) and NO2 (Pearson
correlation coefficient = 0.84) in the study area. Dibben and Clemens (2015) used a
pollution-climate model to assign SO2 concentrations with high spatial resolution as well
as incorporating daily activity data into the exposure and observed null associations with
PTB and modest, positive associations with very preterm birth among births in Scotland.
Oian et al. (2015) observed weak negative or null associations between SO2 exposures
and PTB across a range of different exposure windows among a birth cohort in Wuhan,
China.
In the U.S. and Canada, studies of SO2 and PTB in Pennsylvania (Sagiv et al.. 2005) and
Vancouver (Liu et al.. 2003) found increased ORs with near-birth exposures [6 week
prebirth RR= 1.05 (1.00, 1.10) Sagiv et al. (2005); last month OR= 1.09 (1.01, 1.19) per
5-ppb increase Liu et al. (2003)1. More recently, in a Detroit, MI cohort, Le et al. (2012)
found similar associations for exposures in the last month of pregnancy [OR 4th to 1st
quartile: 1.07 (1.01, 1.14)]. Another Vancouver cohort, examining entire pregnancy
exposure, only observed increases [OR= 1.03 (0.93, 1.15) per 5-ppb SO2 increase] with
PTB <30 weeks (Brauer et al.. 2008). Recent time-series and case-crossover studies in
Atlanta, GA and Brisbane, Australia observed null associations for both 1st month and
near-birth exposures using 1-h max SO2 [exposure during last week of pregnancy RR per
5-ppb increase = 0.99 (0.98, 1.01)] (Darrow et al.. 2009) and SO2 concentrations
24-48 hours preceding the onset of labor (Li et al.. 2016). Finally, a cross-sectional study
5-258
-------
of PTB across the U.S. reported that SO2 showed "nonsignificant" effects with PTB for
exposures during the month of birth (Trasande et al.. 2013). In contrast, a recent study
conducted in Italy observed negative associations between SO2 exposure averaged across
the entire pregnancy as well as each trimester and PTB, suggesting the SO2 exposure was
associated with longer gestation (Capobussi et al.. 2016).
In summary, there is some evidence for an association between exposure to SO2 and
preterm birth particularly with near-birth exposure windows. Studies examining PTB
primarily used average daily SO2. The one study that examined 1-h max SO2 found no
associations for PTB. Recent studies do not provide evidence to help reduce uncertainty
related to exposure measurement error, copollutant confounding, or the biological
mechanism by which SO2 could cause preterm birth. No recent animal studies evaluating
preterm birth were identified. Studies are characterized in Supplemental Table 5S-24
(U.S. EPA. 2017c).
5.4.3.3 Birth Weight
Birth weight is a measure of fetal growth and an important indicator of future infant and
child health. Birth weight is determined by gestational age and intra-uterine growth, as
well as maternal, placental, fetal, and environmental factors. Vulnerability to
environmental insults affecting birth weight may occur throughout pregnancy.
Implantation or formation of the placenta may be disrupted in the earliest weeks of
pregnancy, leading to decreased fetal nutrition throughout pregnancy; or inflammation
might result in constriction of the umbilical cord during the later trimesters resulting in
poor fetal nutrition. As the largest gains in birth weight occur during the last weeks of
gestation, this may be a particularly vulnerable period for birth weight outcomes.
Information on birth weight is routinely collected for vital statistics; given that measures
of birth weight do not suffer the same uncertainties as gestational age or growth
restriction, it is one of the most studied outcomes within air pollution and reproductive
health. Birth weight may be examined as a continuous outcome or a dichotomous
outcome as low birth weight (LBW) (less than 2,500 g or 5 lbs, 8 oz).
Studies examining LBW have found elevated ORs with exposures in the first trimester or
first month (Dugandzic et al.. 2006; Lee et al.. 2003; Liu et al.. 2003; Ha et al.. 2001) and
with entire pregnancy exposures (Capobussi et al.. 2016; Dibben and Clemens. 2015;
Yorifuji et al.. 2015a; Ebisu and Bell. 2012; Kumar. 2012; Morello-Frosch et al.. 2010).
In the two studies that examined distance to monitor, using concentrations from closer
monitors lead to stronger effect estimates (Kumar. 2012; Morello-Frosch et al.. 2010).
5-259
-------
Some studies examining entire pregnancy exposure have also observed null associations
between SO2 and LBW (Braucr ct al.. 2008; Bell et al.. 2007).
Studies examining continuous birth weight (Ag) have inconsistent results. In a northeast
U.S. population, Bell et al. (2007) observed no association with change in birth weight
for entire pregnancy exposure [-2.711 g (-13.253, 7.831) per 5 ppb SO2], including in a
stratified analysis of white and black mothers. Kumar (2012) reported results that shifted
around the null based on distance from monitor in Chicago; some effects were positive,
and some negative but all had wide confidence intervals. And, in a cross-sectional study
across the county, Trasande et al. (2013) reported only "nonsignificant" effects for SO2.
One recent California cohort study reported increases in birth weight with increases in
SO2 exposure in entire pregnancy and first trimester, although effects were reduced with
use of closer monitors (Morello-Frosch et al.. 2010). A recent Texas study observed
decreases in birth weight with county average SO2 exposure for the entire pregnancy
[-15.594 g (-25.344, -5.844)] (Geer et al.. 2012). A study in Beijing during the summer
Olympics of 2008 found increased SO2 in the 8th month of pregnancy associated with
decrements in birth weight; however, SO2 was highly correlated with PM2 5 and CO,
which showed similar patterns of effect (Rich et al.. 2015). Finally, a recent study in
Atlanta found decreases in birth weight with increases in 3rd trimester 1-h max SO2
(Darrow et al.. 2011). This effect was stronger in non-Hispanic white and Hispanic
mothers than non-Hispanic black mothers (Darrow et al.. 2011).
No recent animal studies evaluating birth weight-related outcomes were identified. In
laboratory animals from an older study, exposure to SO2 affected birth outcomes in adult
female rodents and their offspring. Adult female albino rats were exposed to either
0.057 ppm or 1.5 ppm SO2 by inhalation for 72 days (Mamatsashvili. 1970b). At birth,
litter sizes were significantly increased in number from dams that were exposed to SO2
versus control dams (Table 5-37).
In summary, there is some evidence that LBW may be associated with SO2, while
evidence for an association with change in birth weight is inconsistent. Overall, the
results of studies of LBW and birth weight remain inconsistent and these do not provide
evidence to help reduce uncertainty related to exposure measurement error, copollutant
confounding, or the biological mechanism by which SO2 could cause these effects.
Studies for both LBW and change in birth weight can be found in Supplemental
Table 5S-25 (U.S. EPA. 2017c).
5-260
-------
5.4.3.4 Birth Defects
Birth defects are structural and functional abnormalities that can cause physical disability,
intellectual disability, and other health problems. They are a leading cause of infant
mortality and developmental disability in the U.S. (Mai et al.. 2016). Since 2008, there
have been several studies examining birth defects and SO2 during pregnancy, particularly
during Weeks 3-8 of gestation, which is thought to be highly vulnerable to insults
resulting in birth defects. Because birth defects as a whole are rare and specific birth
defects are rarer, these studies often have effect estimates with very wide confidence
intervals. Individual studies often look at different types of birth defects, meaning the
body of work examining any one birth defect may still be limited. Cardiac birth defects
and oral cleft defects are the most commonly studied anomalies. However, results (even
for these defects) are inconsistent across studies. For example, odds of ventricular septal
defects have been found to be increased (Gianicolo et al.. 2014; Stingone et al.. 2014;
Agav-Shav et al.. 2013; Gilboa et al.. 2005). decreased (Hwang et al.. 2015b; Dadvand et
al.. 2011a. b; Rankin et al.. 2009). and null (Strickland et al.. 2009) with increases in SO2
exposure. Odds of cleft lip with or without cleft palate have been found to be increased
(Zhu et al.. 2015). decreased (Hwang and Jaakkola. 2008; Gilboa et al.. 2005). or null
(Polk et al.. 2010; Rankin et al.. 2009) with increases in SO2 exposure. A single study of
limb deformities found increased odds with exposure to SO2 during Weeks 9-12 of
pregnancy (Lin etal.. 2014). Two studies examining repeating chromosomal defects
found no association or correlation between trisomy 21 or any sperm disomy and SO2
(Chung et al.. 2014; Jurewicz et al.. 2014). Studies of any congenital anomaly in Israel
and China have reported inverse associations with increasing SO2 (Farhi et al.. 2014;
Liang et al.. 2014).
In summary, results for birth defects are either inconsistent across studies or limited in
number of studies. No recent animal studies evaluating birth defects were identified.
Studies of birth defects and SO2 are characterized in Supplemental Table 5S-26 (U.S.
EPA. 2017c).
5.4.3.5 Fetal Mortality
Fetal mortality or stillbirth is the intra-uterine death of a fetus. In most areas fetal deaths
are only reported after 20 weeks of completed gestation; this leads to potential bias, as the
population at risk of fetal death is any conception but the actual measured population is
only those fetuses reaching at least 20 weeks gestational age. A single recent case-control
study of spontaneous abortion occurring before 14 weeks of gestation found no
associations with SO2 exposures determined by time-weighted concentrations for
5-261
-------
residence and workplace (Moridi ct al.. 2014). A recent large California cohort found no
associations between stillbirth and increasing SO2 exposure (Green et al.. 2015). In recent
studies of a New Jersey population examining both long-term and short-term exposure
windows, ORs for fetal death were elevated with a 2-day lag [OR per 5-ppb increase in
SO2: 1.12 (1.02, 1.24)] and with exposures across pregnancy and in each trimester,
particularly the 3rd trimester [OR per 5-ppb increase in SO2: 1.47 (1.05, 1.69)] (Faiz et
al.. 2013; Faiz et al.. 2012). Hwang et al. (2011) examined fetal mortality among term
and preterm deliveries in Taiwan, finding elevated associations for exposures during the
1st trimester only among preterm deliveries. Other studies have also found increased
associations between SO2 and fetal mortality, although mean SO2 concentrations were
higher in these studies (Hou et al.. 2014; Pereira et al.. 1998). Pereiraetal. (1998)
observed elevated RRs in a Sao Paulo, Brazil time series with short-term exposure. A
recent study by Enkhmaa et al. (2014) found very strong correlations between seasonal
SO2 and fetal death, and Hou et al. (2014) found elevated ORs with long-term exposures
around the time of conception. However, the Hou et al. (2014) models were unadjusted
for confounding factors, and confidence intervals were very wide. In the study by
Enkhmaa et al. (2014). other pollutants also showed very strong correlations with fetal
death and the pollutants were highly correlated with one another (i.e., r > 0.7).
In summary, although few in number, studies of fetal mortality and SO2 show elevated
associations for both short- and long-term exposures. However, these studies are limited
by the uncertainties associated reproductive and developmental outcomes identified in the
2008 SOx ISA. No recent animal studies evaluating fetal mortality were identified.
Studies are characterized in Supplemental Table 5S-27 (U.S. EPA. 2017c).
5.4.3.6 Infant Mortality
Studies of infant mortality and SO2 are limited in number. In a U.S. study, Woodruff et
al. (2008) observed increased ORs for respiratory-related post-neonatal infant mortality
with long-term (2 months) exposure increases in county-level SO2 concentrations
[OR = 1.09 (0.89, 1.36) per 5-ppb increase]. This association remained after adjusting for
other pollutants. A time-series study in Seoul, South Korea observed increased RRs for
all cause post-neonatal infant mortality with short-term SO2 exposure, although exact
timing of exposure was unclear (Son et al.. 2008). No recent animal studies evaluating
postnatal mortality were identified. Studies are characterized in Supplemental
Table 5S-27 (U.S. EPA. 2017c).
5-262
-------
5.4.4
Developmental Outcomes
5.4.4.1 Respiratory Outcomes
Recent studies examined asthma onset in association with early life exposure to SO2.
Clark et al. (2010). Liu et al. (2016). Deng et al. (2015b). and Deng et al. (2015a)
observed elevated ORs for asthma with SO2 exposure during pregnancy and the first year
of life. Nishimura et al. (2013) observed elevated ORs for asthma with SO2 exposure in
the first 3 years of life, but not the first year of life alone. Asthma onset is covered in
further detail in Section 5.2.1.2.
In a time-series study, Dales et al. (2006) investigated neonatal hospitalizations due to
respiratory causes in Atlanta, GA; they observed elevated ORs with 2-day lagged SO2
exposure. After adjustment for gaseous copollutants, confidence intervals for associations
with gaseous pollutants and PM10 were very large, but effect estimates remained elevated.
Hospitalizations due to respiratory causes are covered in Section 5.2.1.6.
In summary, there is some evidence for an association between gestational and early life
exposure to SO2 and respiratory health effects that extend into early childhood, although
evidence is limited and exposure windows are uncertain. Key studies are summarized in
Table 5-36.
5.4.4.2 Other Developmental Effects
Studies examining other developmental exposures are limited in number. Peel et al.
(2011) examined SO2 exposure with apnea and bradycardia in a subpopulation of infants
in Atlanta and observed no association for either health outcome. Huang et al. (2015a)
observed no associations between prenatal and early life SO2 exposures and atopic
dermatitis among infants in Taiwan. Poursafa et al. (2016) examined the association
between SO2 exposure during pregnancy and markers of endothelial dysfunction
(i.e., ICAM-1, vascular adhesion molecule-1, endothelin-1) in cord blood. They observed
a positive association with endothelin-1, but not for other markers of endothelial
dysfunction. Among a Japanese cohort, prenatal exposure to SO2 was associated with
verbal and fine motor delays assessed at ages 2.5 and 5.5 years (Yorifuji et al.. 2015b). In
an older study from the animal toxicology literature, adult female albino rats were
exposed to either 0.057 ppm or 1.5 ppm SO2 by inhalation, 12 hours/day for 72 days
(Mamatsashvili. 1970b). Changes in offspring postnatal growth or body weight over time
were reported with 1.5-ppm exposure.
5-263
-------
Sulfur dioxide-dependent synaptic injury was measured in adolescent male rats exposed
to 1.24 ppm SO2 for 6 hours/day for 90 days (Yun et al.. 2013). Nonsignificant
morphological changes were seen in the hippocampal synaptic junctions using
transmission electron microscopy. In the hippocampus, the synaptic vesicle membrane
protein synaptophysin (SYP) was significantly downregulated as was ERK1/2
phosphorylation. Phosphorylation is an important contributor to synaptic plasticity. Thus,
SO2 exposure in the adolescent rat contributes to downregulation of synaptic vesicle
protein SYP and decreased ERK1/2 phosphorylation, indicative of disruption at the
hippocampal synapse.
5.4.5 Summary and Causal Determination
Overall the evidence is inadequate to infer a causal relationship between exposure to
SO2 and reproductive and developmental outcomes. This is consistent with the 2008
ISA for Sulfur Oxides, which also concluded the evidence was inadequate to infer a
causal relationship with reproductive and developmental effects. All available evidence,
including more than 50 recent studies, examining the relationship between exposure to
SO2 and reproductive and developmental effects was evaluated using the framework
described in the Preamble to the ISAs (U.S. EPA. 2015b). The key evidence as it relates
to the causal framework is summarized in Table 5-38.
There are several well-designed, well-conducted epidemiologic studies, many described
in papers published since the previous ISA, that indicate an association between SO2 and
reproductive and developmental health outcomes; the bulk of the evidence exists for
adverse birth outcomes. For example, several high quality studies reported positive
associations between SO2 exposures during pregnancy and fetal growth metrics (Le et al..
2012; Rich et al.. 2009; Brauer et al.. 2008; Liu et al.. 2003). preterm birth (Mendola et
al.. 2016a; Le et al.. 2012; Zhao etal.. 2011; Sagiv et al.. 2005; Liu et al.. 2003). birth
weight (Ebisu and Bell. 2012; Darrow et al.. 2011; Morello-Frosch et al.. 2010; Liu et al..
2003). and fetal and infant mortality (Faiz et al.. 2012; Hwang etal.. 2011; Woodruff et
al.. 2008). However, the evidence is not entirely consistent, and has not substantially
reduced any of the uncertainties connected with the associations observed between
exposure to SO2 and birth outcomes that were identified in the previous ISA.
One uncertainty is timing of exposure, wherein associations remain inconsistent among
studies and across outcomes. For example, some studies observe the strongest
associations when exposure is averaged over the entire pregnancy, while others observe
the strongest association when exposure is averaged over either the first, second, or third
trimester. As an exception to this, studies of PTB generally observed positive associations
5-264
-------
between near-birth exposures (e.g., last month of gestation, same, or 3-day lag from birth)
(Mcndola et al.. 2016a; Le et al.. 2012; Zhao etal.. 2011; Sagiv et al.. 2005; Liu et al..
2003).
Table 5-38 Summary of evidence inadequate to infer a causal relationship
between sulfur dioxide exposure and reproductive and
developmental effects.
Rationale for Causal SO2 Concentrations
Determination3 Key Evidence13 Key References'3 Associated with Effects0
Overall reproductive and developmental effects—inadequate to infer a causal relationship
Evidence from multiple
epidemiologic studies of
preterm birth is generally
supportive but key
uncertainties remain
Consistent positive
associations observed with
near-birth exposures to
SO2 and preterm birth after
adjustment for common
potential confounders.
Associations not evaluated
in copollutant models.
Saaiv et al. (2005) Mean: 7.9 ppb
tLe etal. (2012)
Mean: 5.8 ppb
tMendola et al. (2016a)
Mean: 4.0 ppb
Section 5.4.3.2
Associations robust in Liu et al. (2003) Mean: 4.9 ppb
copollutant models
5-265
-------
Table 5-38 (Continued): Summary of evidence inadequate to infer a causal
relationship between sulfur dioxide exposure and
reproductive and developmental effects.
Rationale for Causal
Determination3
Key Evidence13
Key References'3
SO2 Concentrations
Associated with Effects0
Limited and inconsistent
epidemiologic evidence for
other birth outcomes
Several studies show
positive associations with
fetal growth metrics,
although definitions vary
across studies, and timing
of exposure is inconsistent.
Associations not evaluated
in copollutant models
Section 5.4.3.1
Means: 4.9-5.8 ppb
Several high quality studies Section 5.4.3.3 Means: 2.1-13.2 ppb
show associations between
SO2 exposure and low birth
weight but not for change
in birth weight. Timing of
exposure is inconsistent
across studies. Only one
study uses 1-h max for
exposure determination.
Limited and inconsistent Section 5.4.3.4 Reported means: 1.9-6
epidemiologic evidence for
associations with various
birth defects
Limited number of studies Section 5.4.3.6 Mean: 5.7 ppb
of SO2 and fetal death, Mean: 5.8 ppb
positive associations Mean: 5.9 ppb
observed across studies, Mean: 3 ppb
although timing of
exposure and outcome
definitions are inconsistent.
Limited evidence for an
association with SO2 in
respiratory related infant
mortality.
Limited evidence for Section 5.4.4.1 Means: 2-4.3 ppb
positive associations
between prenatal/early life
exposures and childhood
respiratory outcomes
Limited evidence for key Altered menstrual function, Mamatsashvili (1970a) 57 or 1,427 ppb
events in proposed mode fetal growth, and birth
of action weight outcomes with
impaired postnatal growth
for in utero exposed pups
Lack of evidence from
epidemiologic studies to
support an association of
SO2 exposure with
detrimental effects on
fertility or pregnancy
A limited number of studies Section 5.4.4.1
on fertility and pregnancy
outcomes show no
associations with SO2.
Mean 8.4-59 ppb
5-266
-------
Table 5-38 (Continued): Summary of evidence inadequate to infer a causal
relationship between sulfur dioxide exposure and
reproductive and developmental effects.
Rationale for Causal
SO2 Concentrations
Determination3
Key Evidence13
Key References'3 Associated with Effects0
Uncertainty regarding
Limited adjustment for
+(Faiz et al. (2013): Slama
potential confounding by
copollutants, with no clear
etal. (2013): Le et al.
copollutants
directionality or trends for
(2012))
effect estimate shifts after
adjustment
Uncertainty regarding
Fixed-site monitors subject
Chapter 3
exposure measurement
to some degree of
Section 3.4.4.2
error
exposure error.
Uncharacterized spatial
and temporal heterogeneity
may introduce exposure
error in long-term effects
and bias could be toward
or away from the null.
Uncertainty regarding
Associations of exposure to
exposure timing for specific
SO2 at particular windows
outcomes
during pregnancy are
inconsistent among studies
and across outcomes
Max = maximum; S02 = sulfur dioxide.
aBased on aspects considered in judgments of causality and weight of evidence in causal framework in the Preamble to the ISAs
(U.S. EPA. 2015b).
bDescribes the key evidence and references contributing most heavily to causal determination and where applicable to
uncertainties and inconsistencies. References to earlier sections indicate where full body of evidence is described.
°Describes the S02 concentrations with which the evidence is substantiated (for experimental studies, below 2,000 ppb).
fStudies published since the 2008 ISA for Sulfur Oxides.
Another uncertainty centers on spatial and temporal variability in SO2 exposures. SO2 is a
temporally and spatially heterogeneous pollutant; it is difficult to accurately estimate for
"long-term" exposures, and there is the potential for exposure measurement error in
long-term SO2 exposures to bias estimates toward or away from the null
(Section 3.4.4.2V None of the epidemiologic studies made corrections or adjustments for
exposure measurement error or accounted for the potential for bias away from the null,
the potential for which has been demonstrated in simulation studies (see Section 3.4.4.2).
Current epidemiologic methods are not able to disentangle whether associations are due
to extended exposure to moderate concentrations of SO2 or repeated short-term exposure
to peaks in SO2 concentration.
Potential confounding by copollutants may explain some of the observed associations and
cannot be ruled out. SO2 is part of a mix of ambient air pollution; SO2 shares sources with
particulate matter and is chemically linked to sulfate. Few studies evaluate or provide
information that would inform the independent effect of SO2 in the context of the greater
5-267
-------
air pollution mixture, and of those that do, no clear trends for the effects of copollutant
adjustment are apparent (Faiz et al.. 2013; Slama et al.. 2013; Le et al.. 2012).
There is insufficient information on potential modes of action of SO2 on reproductive
outcomes at relevant exposure levels for this ISA (Chapter 4). In a single older study
from Mamatsashvili (1970a'). SO2 inhalation exposure in laboratory rodents demonstrated
reproductive changes in exposed females and their offspring, altered birth outcomes, and
developmental effects. The specific outcomes affected after SO2 exposure included
altered estrus cycle length of F0 and F1 generations, decrements in offspring body weight
gain or growth after in utero exposure, and changes in litter size. The majority of the
remaining animal toxicological evidence for reproductive and developmental effects is
for exposure at 5,000 ppb or greater, doses which are beyond the scope of this document.
Since the 2008 ISA for Sulfur Oxides, researchers have begun evaluating more health
outcomes, including fertility, effects on pregnancy (e.g., pre-eclampsia, gestational
diabetes), and developmental effects. For each of these individual outcomes the literature
base is small, but new studies are quickly accumulating. However, at present there is little
coherence or consistency among epidemiologic and toxicological studies for these
outcomes. In general, it is challenging to synthesize study findings on the wide variety of
health outcomes collected under the reproductive and developmental effects heading.
Given the wide variety of potential mechanisms or adverse outcome pathways that could
affect this breadth of outcomes, coherence is unlikely to be reached given the limited
literature base.
Overall, many uncertainties remain when evaluating the evidence for these health
endpoints; therefore, the evidence is inadequate to infer a causal relationship
between exposure to SO2 and reproductive and developmental outcomes.
5.5 Mortality
5.5.1 Short-Term Exposure
5.5.1.1 Introduction
The 2008 SOx ISA concluded that the collective evidence is "suggestive of a causal
relationship" between short-term SO2 exposure and mortality. Overall, the number of
studies that examined the relationship between short-term SO2 exposure and mortality
was sparse and there was limited data available to inform the potential for copollutant
5-268
-------
confounding. Since the completion of the 2008 SOx ISA (U.S. EPA. 2008d).
epidemiologic literature that has examined the association between short-term SO2
exposure and mortality has expanded. However, similar to the collection of studies
evaluated in the 2008 SOx ISA (U.S. EPA. 2008d). most of the recent studies do not
focus specifically on the SCh-mortality relationship but instead on PM or O3. Of the
studies identified, a limited number have been conducted in the U.S., Canada, and
Europe, with the majority being conducted in Asia due to the increased focus on
examining the effect of air pollution on health in developing countries.
Studies included in the 1982 AQCD that examined the association between short-term
SOx exposure, mainly SO2, and total mortality were limited to historical data on high air
pollution episodes (U.S. EPA. 1982a). These studies were unable to decipher whether the
associations observed were due to particle pollution or SO2. Additional studies evaluated
in the 1986 Second Addendum to the 1982 AQCD (U.S. EPA. 1986) further confirm the
findings of these initial studies, but were still unable to address uncertainties and
limitations related to examining the effect of SO2 exposure on mortality, especially at
lower concentrations.
In the 2008 SOx ISA (U.S. EPA. 2008d). a larger body of literature was available to
assess the relationship between short-term SO2 exposures and mortality; however, these
studies were still limited in that they primarily focused on PM, with SO2 only being
examined in single-pollutant models. These studies found that excess risk estimates for
total mortality due to short-term SO2 exposure from multicity studies and meta-analyses
generally ranged from 0.4 to 2.0% for a 10-ppb increase in 24-h avg SO2 concentrations.
These associations were primarily observed at mean 24-h avg SO2 concentrations
<15 ppb. Studies that examined cause-specific mortality found evidence of risk estimates
larger in magnitude for respiratory and cardiovascular mortality compared to total
mortality with the largest associations for respiratory mortality. The larger
S02-respiratory mortality associations observed in the epidemiologic literature were
coherent with the scientific evidence providing stronger support for SO2 effects on
respiratory morbidity compared to cardiovascular morbidity (U.S. EPA. 2008d).
An examination of potential copollutant confounding of the S02-mortality relationship
was sparse. Studies evaluated in the 2008 SOx ISA found that S02-mortality risk
estimates from copollutant models were robust but imprecise (i.e., wide confidence
intervals). An additional study that examined the potential interaction between
copollutants [i.e., SO2 and BS] did not find evidence of interaction when stratifying days
by high and low concentrations of BS (Katsouvanni et al.. 1997). Of the studies
evaluated, only the Air Pollution and Health: A European Approach (APHEA) study
examined seasonality and potential effect modifiers of the S02-mortality relationship.
5-269
-------
APHEA provided initial evidence that mortality effects are larger during the warm season
and that geographic location may influence city-specific SCh-mortality risk estimates,
respectively (Katsouvanni et al.. 1997). The consistent, positive SCh-mortality
associations observed across studies were supported by an intervention study conducted
in Hong Kong that examined the health impact of converting to fuel oil with low sulfur
content and found evidence suggesting that a reduction in SO2 concentrations leads to a
reduction in mortality (Hedlev et al.. 2002).
As detailed in previous ISAs [e.g., U.S. EPA (2013b)l. this section focuses primarily on
multicity studies because they examine the association between short-term SO2 exposure
and mortality over a large geographic area using a consistent statistical methodology,
which avoids the potential publication bias often associated with single-city studies (U.S.
EPA. 2008d). However, where applicable single-city studies are evaluated that
encompass a long study-duration, provide additional evidence indicating that a specific
population or lifestage is at increased risk of SCh-related mortality, or address a limitation
or uncertainty in the SCh-mortality relationship not represented in multicity studies. The
remaining studies identified are not evaluated in this section due to issues associated with
study design or insufficient sample size and are detailed in Supplemental Table 5S-28
(U.S. EPA. 2017c).
The organization of the material on short-term SO2 exposure and mortality is as follows.
Section 5.5.1.2 evaluates studies that examined the association between short-term SO2
exposure and mortality, with the remaining sections addressing key limitations and
uncertainties in the S02-mortality relationship that were evident at the completion of the
2008 SOx ISA (U.S. EPA. 2008d). Subsequent sections evaluate whether there is
evidence of confounding (i.e., copollutants and seasonal/temporal) (Section 5.5.1.3).
effect modification (i.e., sources of heterogeneity in risk estimates across cities or within
a population) (Section 5.5.1.4). modification of the S02-mortality association including
seasonal heterogeneity (Section 5.5.1.4). and the S02-mortality C-R relationship and
related issues, such as the lag structure of associations (Section 5.5.1.5).
5.5.1.2 Associations between Short-Term Sulfur Dioxide Exposure and Mortality in
All-Year Analyses
Multicity studies and meta-analyses evaluated in the 2008 SOx ISA reported consistent,
positive associations between short-term SO2 exposure and total mortality in all-year
analyses (U.S. EPA. 2008d). Although only a small number of multicity studies have
been conducted since the completion of the 2008 SOx ISA. These studies, as well as a
meta-analysis of studies conducted in Asia (Atkinson et al.. 2012). build upon and
provide additional evidence for an association between short-term SO2 exposure and total
5-270
-------
mortality (Figure 5-17). Air quality characteristics and study specific details for the
studies evaluated in this section are provided in Table 5-39.
Table 5-39 Air quality characteristics of multicity studies and meta-analyses
evaluated in the 2008 SOx ISA and recently published multicity
studies and meta-analyses.
Study
Location
Years
Mortality
Outcome(s)
Averaging
Time
Mean
Concentration
PPb
Upper
Percentile
Concent-
rations ppb
North America
Dominici et al.
(2003)
72 U.S. cities
(NMMAPSf
1987-1994
Total
24-h avg
0.4-14.2
...
Burnett et al.
(2004)
12 Canadian cities
1981-1999
Total
cardiovascular
respiratory
24-h avg
0.9-9.6
—
tMoolqavkar
et al. (2013)
85 U.S. cities
(NMMAPS)b
1987-2000
Total
24-h avg
...
...
Europe
Katsouvanni et
al. (1997)
12 European cities
(APHEA-1)
1980-1992
Total
24-h avg
5.0-28.2°
90th:
17.2-111.8
Biqaeri et al.
(2005)
Eight Italian cities
(MISA-1)
1990-1999
Total
cardiovascular
respiratory
24-h avg
2.5-15.6
95th:
6.0-50.1
Max:
7.1-111.0
Hoek (2003)
Netherlands
1986-1994
Total
cardiovascular
respiratory
24-h avg
3.5-5.6
—
tBeralind et al.
(2009)
Five European
citiesd
1992-2002
Total
24-h avg
CO
I
o
...
tBellini et al.
(2007)
15 Italian cities
(MISA-2)
1996-2002
Total
cardiovascular
respiratory
24-h avg
—
—
5-271
-------
Table 5-39 (Continued: Air quality characteristics of multicity studies and meta
analyses evaluated in the 2008 SOx ISA and recently
published multicity studies and meta analyses.
Study
Location
Years
Mortality
Outcome(s)
Averaging
Time
Mean
Concentration
PPb
Upper
Percentile
Concent-
rations ppb
Asia
tKan et al.
(2010): Wona
et al. (2008b):
Wona et al.
(2010)
Four Asian cities
(PAPA)
1996-2004e
Total
cardiovascular
respiratory
24-h avg
5.0-17.1
75th:
6.0-21.5
Max:
23.4-71.7
tChen et al.
(2012b)
17 Chinese cities
(CAPES)
1996-2010'
Total
cardiovascular
respiratory
24-h avg
6.1-38.2
75th:
6.5-56.1
Max:
25.2-298.5
tChen et al.
(2013)
Eight Chinese
cities
1996-2008'
Stroke
24-h avg
6.1-32.1
—
tMenq et al.
(2013)
Four Chinese
cities
1996-20089
COPD
24-h avg
6.8-19.1
...
Meta-analyses
Stieb et al.
(2003)
Meta-analysis
1958-1999b
Total
24-h avg
0.7-75.2
...
HEI (2004)
Meta-analysis
(South Korea,
China, Taiwan,
India, Singapore,
Thailand, Japan)
1980-2003h
Total
24-h avg
~10->200
tAtkinson et
al. (2012)
Meta-analysis
(Asia)
1980-2007'
Total
cardiovascular
respiratory
COPD
24-h avg
tShah et al.
(2015)
Meta-analysis
1948-Jan
2014
Stroke
NR
6.2J
Max: 30.2
5-272
-------
Table 5-39 (Continued: Air quality characteristics of multicity studies and meta
analyses evaluated in the 2008 SOx ISA and recently
published multicity studies and meta analyses.
Study
Location Years
Mortality
Outcome(s)
Averaging
Time
Mean
Concentration
PPb
Upper
Percentile
Concent-
rations ppb
tYanq et al.
Meta-analysis 1996-2013
Stroke
24-h avg
Asia: 11.4°
75th: Asia:
(2014b)
(Asia, Europe, and
Europe: 5.2C
18.6
North America)
North America:
Europe: 2.3
4.2b
North
America:
7.6
APHEA = Air Pollution and Health: A European Approach study; avg = average; CAPES = China Air Pollution and Health Effects
Study; COPD = chronic obstructive pulmonary disease; ISA = Integrated Science Assessment; max = maximum;
MISA = meta-analysis of the Italian studies on short-term effects of air pollution; NMMAPS = The National Morbidity Mortality Air
Pollution Study; NR = not reported; PAPA = Public Health and Air Pollution in Asia; S02 = sulfur dioxide; SOx = sulfur oxides.
aOf the 90 cities included in the NMMAPS analysis only 72 had S02 data.
bOf the 108 cities included in the analyses using NMMAPS data, only 85 had S02 data.
°Median concentration(s).
dS02 data was not available for Barcelona; therefore, the S02 results only encompass four cities.
eThe study period varied for each city, Bangkok: 1999-2003, Hong Kong: 1996-2002, and Shanghai and Wuhan: 2001-2004.
'Study period varied for each city and encompassed 2 to 7 yr, Hong Kong was the only city that had air quality data prior to 2000.
9Study period varied from 2 to 7 yr. Hong Kong was the only city that had air quality data prior to 2001.
hStudies included within this meta-analysis were published during this time period.
'ear defined represents the year in which studies were published that were included in the meta-analysis.
'The mortality time series of studies included in the meta-analysis spanned these years.
f = Studies published since the 2008 ISA for Sulfur Oxides.
5-273
-------
Study
Location
Age
Lag
Dominici et al. (2003)
72 U.S. cities (NMMAPS)
All
1
Burnett etal. (2004)
12 Canadian cities
All
0-2
Katsouyanniet al. (1997)
12 European cities (APHEA 1)
All
Variable (0-3 days)
Biggeri et al. (2005)
8 Italian cities (MISA-1)
All
0-1
Hoek et al. (2003)
Netherlands
All
0-6
Stieb et al. (2002)
Meta-analysis (Worldwide)
All
Variable
HEI (2004)a
Meta-analysis (Asia)
All
Variable
|Moolgavkar et al. (2013)
85 U.S. cities (NMMAPS)
All
1
|Berglind et al. (2009)b
5 European cities
35-74
0-1
|Bellini et al. (2007)
15 Italian cities (MISA-2)
All
0-1
tChen et al. (2012)
17 Chinese cities (CAPES)
All
0-1
|Kan et al. (2010)c
4 Asian cities (PAPA)
All
0-1
t Atkinson et al. (2012)
Meta-analysis (Asia)
All
Variable
-5.0
-m—
0.0 5.0
% Increase (95% Confidence Interval)
10.0
APHEA = Air Pollution and Health: A European Approach study; CAPES = China Air Pollution and Health Effects Study; ISA = Integrated Science Assessment; MISA = meta-analysis
of the Italian studies on short-term effects of air pollution; NMMAPS = The National Morbidity Mortality Air Pollution Study; PAPA = Public Health and Air Pollution in Asia.
3Meta-analysis of Asian cities: South Korea, China, Hong Kong, Taipei, India, Singapore, Thailand, Japan (HEI. 2004).
bStudy was of myocardial infarction survivors therefore only included individuals 35+ (Berqlind et al.. 2009).
cKan et al. (2010) reported results that were also found in (Wong et al.. 2010; Wong et al. (2008b)).
Note: f and red text/circles = recent studies published since the 2008 ISA for Sulfur Oxides; black text/circles = U.S. and Canadian studies evaluated in the 2008 ISA for Sulfur Oxides.
Corresponding quantitative results are reported in Supplemental Table 5S-29 (U.S. EPA. 2017c).
Figure 5-17 Percent increase in total mortality from multicity studies and meta-analyses evaluated in the 2008
ISA for Sulfur Oxides (black circles) and recently published multicity studies (red circles) for a
10-ppb increase in 24-hour avg sulfur dioxide concentrations.
5-274
-------
When focusing on specific causes of mortality, some studies evaluated in the 2008 SOx
ISA reported similar risk estimates across mortality outcomes [e.g., (Zmirou et al. (1998);
Katsouvanni et al. (1997))!. while others indicated larger risk estimates for respiratory
mortality (Figure 5-18). However, a study conducted in the Netherlands by Hoek (2003)
suggested that specific cardiovascular mortality outcomes have larger risk estimates
compared to all cardiovascular, total, and respiratory-related mortality outcomes. Recent
multicity mortality studies provide additional support indicating larger risk estimates for
respiratory mortality compared to total and cardiovascular mortality. Additionally, the
results from the studies depicted in Figure 5-18 lend additional support to the body of
evidence indicating SC>2-induced respiratory effects presented in the 2008 SOx ISA, as
well as Section 5.2 of this ISA. Unlike the results reported in Hoek (2003). recent studies
do not provide evidence indicating associations larger in magnitude for SCh-related
cardiovascular mortality compared to other mortality outcomes.
5-275
-------
S tilth
Katsouyanniet al. (1997)
Zmirou etal. (1998)a
Katsouyanniet al. (1997)
Zmirou etal. (1998)
Biggeri et al. (2005)
Hoek et al. (2003)
•("Bellini etal. (2007)
"I"Atkins on et al. (2012)
tChen et al. (2012)
tChen et al. (2013)
tMeng et al. (2013)
tKan et al. (2010)d
Location
7 W. European Cities (APHEA1)
5 W. European Cities (APHEA1)
5 C. European Cities (APHEA1)
5 C. European Cities (APHEA1)
8 Italian cities (MISA-1)
Netherlands
15 Italian cities (MISA-2)
Meta-analysis (Asia)
17 Chinese cities (CAPES)
17 Chinese cities (CAPES)
8 Chines e cities (CAPES)b
17 Chinese cities (CAPES)
4 Chinese cities(c)
4 As ian cities (PAPA)
Mortality
Total
Cardiovas cular
Respiratory
Total
Cardiovas cular
Respiratory
Total
Cardiovas cular
Respiratory
Total
Cardiovascular
Heart Failure
Thrombo s is -related
COPD
Pneumonia
Total
Cardiovas cular
Respiratory
Total
Cardiovas cular
Respiratory
COPD
Total
Cardiovas cular
Stroke
Respiratory
COPD
Total
Cardiovas cular
Respiratory
Lag
Variable (0-3 days)
Variable (0-3 days)
0-1
0-6
0-1
Variable
0-1
0-1
-6 -4 -2 0 2 4 6 8 10 12 14 16 18 20
% Increase (95% Confidence Iiiterv.il)
APHEA = Air Pollution and Health: A European Approach study; CAPES = China Air Pollution and Health Effects Study; COPD = chronic obstructive pulmonary disease;
ISA = Integrated Science Assessment; MISA = meta-analysis of the Italian studies on short-term effects of air pollution; PAPA = Public Health and Air Pollution in Asia.
Total mortality = circle; cardiovascular-related mortality = triangle; and respiratory-related mortality = diamond.
a(Zmirou et al.. 1998) reported on only five of the seven cities included in (Katsouvanni et al.. 1997). which had cause-specific mortality data and were included in the analysis.
b(Chen et al.. 2013) examined stroke only in the China Air Pollution and Health Effects Study cities that had stroke data.
c(Meng et al.. 2013) was not part of CAPES, but the four cities included had data for the same years as CAPES.
d(Kan et al.. 2010) reported results which were also presented in (Wong et al.. 2008b) and (Wong et al.. 2010).
Note: f and red text/circles = recent studies published since the 2008 ISA for Sulfur Oxides; black text/circles = U.S. and Canadian studies evaluated in the 2008 ISA for Sulfur Oxides.
Corresponding quantitative results are reported in Supplemental Table 5S-30 (U.S. EPA. 2017c).
Figure 5-18 Percent increase in total, cardiovascular, and respiratory mortality from multicity studies
evaluated in the 2008 ISA for Sulfur Oxides (black) and recently published multicity studies (red)
for a 10-ppb increase in 24-h avg sulfur dioxide concentrations.
5-276
-------
5.5.1.3
Potential Confounding of the Sulfur Dioxide-Mortality Relationship
A limitation of the studies evaluated in the 2008 SOx ISA, was the relatively sparse
analyses of the potential confounding effects of copollutants on the SCh-mortality
relationship (U.S. EPA. 2008d). The 2008 SOx ISA specifically stated that the "potential
confounding and lack of understanding regarding the interaction of SO2 with
copollutants" was one of the major limitations of the scientific literature that contributed
to the conclusion that the evidence is "suggestive of a causal relationship" between
short-term SO2 exposures and mortality. Copollutant analyses conducted in recent studies
further attempt to identify whether SO2 has an independent effect on mortality. In
addition to examining potential copollutant confounding, some studies have also
examined whether the covariates included in statistical models employed to examine
short-term SO2 exposures and mortality adequately control for the potential confounding
effects of season/temporal trends and weather.
Examination of Potential Copollutant Confounding
In the 2008 SOx ISA (U.S. EPA. 2008d). the analysis of potential copollutant
confounding was limited to studies conducted by Dominici et al. (2003) within the U.S.
as part of the National Morbidity Mortality Air Pollution Study (NMMAPS),
Katsouvanni et al. (1997) in Europe as part of the APHEA-1 study, Hoek (2003) in the
Netherlands, and Burnett et al. (2004) in 12 Canadian cities. Copollutant models in these
studies focused on the effect of PM10, BS, or NO2 on the S02-mortality relationship. The
S02-mortality risk estimate was found to either increase (Hoek. 2003) or to be slightly
attenuated (Dominici et al.. 2003; Katsouvanni et al.. 1997) in models with BS or PM10;
while risk estimates were reduced, but still remained positive in models with NO2
(Burnett et al.. 2004). Additionally, there was limited evidence from Burnett et al. (2000)
of attenuation of the SO2 association when PM2 5 was included in the model. Recent
multicity studies conducted in the U.S. and Asia have also examined whether there is
evidence of copollutant confounding; however, similar to the literature base considered in
the 2008 SOx ISA (U.S. EPA. 2008d). the evaluation of copollutant confounding on the
S02-mortality relationship has remained limited.
In a study of 108 U.S. cities using data from the NMMAPS for 1987-2000 (of which 85
had SO2 data), Moolgavkar et al. (2013) used a subsampling approach where 4 randomly
selected cities were removed from the 108 cities over 5,000 bootstrap cycles to examine
associations between short-term air pollution concentrations and total mortality. This
approach was used instead of the two-stage Bayesian hierarchical approach employed in
5-277
-------
the original NMMAPS analysis, which assumes that city-specific risk estimates are
normally distributed around a national mean (Dominici et al.. 2003). In a single-pollutant
model using 100 df (~7 df/year, which is consistent with NMMAPS) to control for
temporal trends, Moolgavkar et al. (2013) found a 1.5% (95% CI: 1.1, 1.7) increase in
total (nonaccidental) mortality at lag 1 for a 10-ppb increase in 24-h avg SO2
concentrations. In a copollutant analysis, the S02-mortality risk estimate remained robust
and was similar in magnitude to the single pollutant result upon the inclusion of PM10
[1.3% (95% CI: 0.4, 2.0)]. An analysis of the influence of NO2 on SCh-mortality risk
estimates was not conducted. The results of Moolgavkar et al. (2013) provide additional
support for an SCh-mortality association, as observed in Dominici et al. (2003). through
an analysis that included more cities and used a different statistical approach than
previously employed in multicity studies.
Additional multicity studies in Asia conducted more extensive analyses of potential
copollutant confounding by examining the effect of gaseous pollutants, in addition to
PM10, on the S02-mortality relationship. In a study of 17 Chinese cities as part of
CAPES, Chen et al. (2012b) examined associations between short-term SO2 exposures
and multiple mortality outcomes. The potential confounding effects of other pollutants on
the SCh-mortality relationship was assessed in copollutant models with PM10 and NO2.
Within the cities examined, SO2 was found to be moderately correlated with PM10
(r = 0.49) and NO2 (r = 0.65). The results from copollutant models (Table 5-40) indicate
that although SO2 risk estimates remained positive, they were attenuated by
approximately 39-54% in models with PM10 and 65-79% in models with NO2. These
results are consistent with those observed in Chen et al. (2013). which focused on stroke
mortality in a subset of the CAPES cities (i.e., eight cities) and also reported a similar
reduction in SO2 risk estimates in models with PM10 and NO2.
5-278
-------
Table 5-40 Percent increase in total, cardiovascular, and respiratory mortality
for a 10-ppb increase in 24-h avg sulfur dioxide concentrations at lag
0-1 in single and copollutant models.
Total Mortality Cardiovascular Mortality Respiratory Mortality
Copollutant % Increase (95% CI) % Increase (95% CI) % Increase (95% CI)
SO2 — 1.98 (1.24,2.69) 2.19 (1.24,3.15) 3.31 (2.05,4.59)
+PM10 1.10 (0.45,1.76) 1.00 (0.08,1.92) 2.03 (0.89,3.17)
+NO2 0.42 (-1.56, 1.00) 0.47 (-0.47, 1.42) 1.16 (-0.03, 2.37)
CI = confidence interval; N02 = nitrogen dioxide; PM10 = particulate matter with a nominal aerodynamic diameter less than or equal
to 10 |jm; S02 = sulfur dioxide.
Source: Adapted from Chen et al. (2012^.
Kan et al. (2010) examined the association between short-term SO2 exposures and
mortality within four Asian cities as part of the PAPA study. Although the authors did not
examine copollutant models in a combined four-city analysis, they did on a city-to-city
basis. Similar to Chen et al. (2012b). in single pollutant models across cities and
mortality outcomes, there was evidence of a consistent positive association (Figure 5-19).
Of note is the highly imprecise estimate (i.e., wide confidence interval) for Bangkok, but
it is speculated that the variability in risk estimates for Bangkok could be attributed to the
lack of variability in SO2 concentrations in this city compared to the Chinese cities
(standard deviation in SO2 concentrations of 1.8 ppb; Chinese cities: 4.6-9.7 ppb) (Kan et
al.. 2010). Across mortality outcomes and cities, SCh-mortality risk estimates were
attenuated, and in many cases null in copollutant models with NO2. However, only in
Shanghai and Wuhan were SO2 correlations with NO2 greater than 0.60 (r = 0.64 and
0.76, respectively). Similarly, SO2 was also found to be moderately correlated with PM10
in Shanghai (r = 0.67) and Wuhan (r = 0.65), but SO2 mortality risk estimates, although
attenuated, remained positive across cities. In copollutant models with O3, SO2 mortality
risk estimates were almost unchanged compared to single-pollutant results.
Recent multicity studies add to the limited number of studies that have examined the
potential confounding effects of copollutants on the S02-mortality relationship. Within
the only recent U.S. study, Moolgavkar et al. (2013) reported that SCh-mortality risk
estimates remained robust in copollutant models with PM10, which is consistent with
Dominici et al. (2003). but these studies did not evaluate potential confounding by
gaseous pollutants. Studies that examined gaseous pollutants, including Chen et al.
(2012b) and Kan et al. (2010) along with Burnett et al. (2004). found that in models with
NO2, SO2 risk estimates were reduced to a large extent, but remained positive. However,
5-279
-------
the overall assessment of copollutant confounding remains limited, and it is unclear how
the results observed in Asia translate to other locations, specifically due to the unique air
pollution mixture and higher concentrations observed in Asian cities.
Modeling Approaches to Control for Weather and Temporal Confounding
Mortality risk estimates may be sensitive to model specification, which includes the
selection of weather covariates to include in statistical models to account for the potential
confounding effects of weather in short-term exposure studies. As such, some recent
studies have conducted sensitivity analyses to examine the influence of alternative
approaches to control for the potential confounding effects of weather on mortality risk
estimates.
5-280
-------
Total Mortality
4
3 -
2
1
0
-1
BK
HK
SH
WH
I T TI IT11
SC1C2C3 SC1C2C3 SC1C2C3 SC1C2C3
Cardiovascular Mortality
6 -
4 .
2 -
0 ....
-2 -
-4
BK
HK
SH
WH
I;:i ; :I'
"I 1 1 1-
—I 1 1 1-
—I 1 1 I-
—I 1 1 T"
S C1C2C3 S C1C2C3 S C1C2C3 S C1C2C3
Respiratory Mortality
BK
HK
SH
WH
IilLlJil
S C1C2C3 S C1C2C3 SC1C2C3 S C1C2C3
BK = Bangkok; C1 = sulfur dioxide + nitrogen dioxide; C2 = sulfur dioxide + PM10; C3 = sulfur dioxide + ozone; HK = Hong Kong; S = single-pollutant model; SH = Shanghai;
WH = Wuhan.
Source: Figure adapted from Kan et al. (20101. Reprinted with permission of Elsevier.
Figure 5-19 Percent increase in total, cardiovascular, and respiratory mortality associated with a 10 |jg/m3
(3.62 ppb) increase in 24-h avg sulfur dioxide concentrations, lag 0-1, in single and copollutants
models in Public Health and Air Pollution in Asia cities.
5-281
-------
As part of CAPES, Chen et al. (2012b) examined the influence of alternative lag
structures for controlling the potential confounding effects of temperature on the
SCh-mortality relationship by varying the lag structure of the temperature variable
(i.e., lag 0, lag 0-3, or lag 0-7). The authors found that although the SCh-mortality
associations remained positive and statistically significant across alternative lag
structures, risk estimates were attenuated as the number of lag days specified increased.
The attenuation observed when using a temperature variable lagged from 0-3 to 0-7 days
could be due to Chen et al. (2012b) only including one temperature term in the statistical
model. This approach differs from that used in some of the seminal multicity studies
(e.g., NMMAPS, APHEA) that include a temperature term averaged over multiple days
(e.g., average of lag 1-3 days). A second temperature term is often included in models, in
addition to a same-day temperature term, to account for (1) the potential delayed effects
of temperature on mortality and (2) potential residual confounding due to temperature.
Temporal
In addition to examining the influence of model specification on mortality risk estimates
through the use of alternative weather covariates, recent studies have also examined
whether air pollution-mortality risk estimates are sensitive to the df per year employed to
control for temporal trends.
Within CAPES, Chen et al. (2012b) examined the influence of increasing the number of
degrees of freedom per year (i.e., 4, 6, 8, and 10 df per year) to control for temporal
confounding on SCh-mortality risk estimates. The authors found that as the number of df
per year increased, the percent increase in both total and cause-specific mortality
attributed to SO2 was slightly attenuated but remained positive across the range of df
examined (Figure 5-20.)
The results of Chen et al. (2012b) are consistent with those reported by Kan et al. (2010)
in an analysis of each individual city within the PAPA study. In models using 4, 6, 8, 10,
or 12 df per year, the authors reported relatively similar SCh-mortality risk estimates
across cities. However, as depicted in Figure 5-20. and in some cities in Figure 5-21.
using 4 df per year likely leads to inadequate control for temporal trends based on the
higher risk estimate observed compared to increasing the degrees of freedom.
5-282
-------
2.0
c
'ii
o
k_
03
o.
1.5
CO
¦E
O
£
•S 1.0
0)
u)
ra
0)
b
c
0.5 -
0.0
-0.5
10
10
10
Total mortality
Cardiovascular mortality
Respiratory mortality
Source: fChen et al.. 2012b1. Reprinted with permission of Elsevier.
Figure 5-20 Percent increase in daily mortality associated with a 10 |jg/m3
(3.62 ppb) increase in 24-h avg sulfur dioxide concentrations at
lag 0-1 days using various degrees of freedom per year for time
trend, China Air Pollution and Health Effects Study cities,
1996-2008.
5-283
-------
All natural causes, all ages, S02
MiM
M}{
DF per year
BK = Bangkok; CI = confidence interval; DF = degrees of freedom; HK = Hong Kong; SH = Shanghai; S02 = sulfur dioxide;
WH = Wuhan.
Source: fKan et al.. 20101. Reprinted with permission of Elsevier.
Figure 5-21 Percent increase in total mortality associated with a 10 |jg/m3
(3.62 ppb) increase in 24- h avg sulfur dioxide concentrations at
lag 0-1 in Public Health and Air Pollution in Asia cities, using
different degrees of freedom per year for time trend.
Unlike Chen et al. (2012b) and Kan et al. (2010). which conducted systematic analyses of
the influence of increasing the df per year to control for temporal trends on the
SCh-mortality relationship, Moolgavkar et al. (2013) only compared models that used
50 df (-3.5 df per year) or 100 df (~7 df per year). Similar to both Chen et al. (2012b) and
Kan et al. (2010). the authors reported relatively similar SC^-mortality risk estimates in
both models [1.6% (95% CI: 0.9, 1.9) for a 10-ppb increase in 24-h avg SO2
concentrations at lag 1 in the 50-df model and 1.5% (95% CI: 1.1, 1.7) in the 100 df
model].
Overall, the studies that examined the effect of alternative approaches to control for the
potentially confounding effects of weather and temporal trends report relatively
consistent SC^-mortality risk estimates across models. The results of these studies are
further supported by an analysis conducted by Sacks et al. (2012). which examined
5-284
-------
whether the different modeling approaches (to control for both weather and temporal
trends) used in a number of multicity studies (e.g., NMMAPS, APHEA) resulted in
similar risk estimates when using the same data set. In all-year analyses focusing on
cardiovascular mortality, SCh-mortality risk estimates remained relatively stable across
models using different weather covariates and a varying number of df per year (ranging
from 4 to 8 df per year across models) to control for temporal trends. Although the results
of Sacks et al. (2012) are consistent with Chen et al. (2012b). Kan et al. (2010). and
Moolgavkar et al. (2013) in all-year analyses, seasonal analyses indicate that differences
in model specification may be more important when examining effects by season for
some pollutants, such as SO2.
5.5.1.4 Modification of the Sulfur Dioxide-Mortality Relationship
Individual- and Population-Level Factors
To date, a limited number of studies have examined potential factors that may increase
the risk of SCh-related mortality. In the 2008 SOx ISA (U.S. EPA. 2008d). only
Katsouvanni et al. (1997) examined potential effect measure modifiers and within the
APHEA-2 study reported that geographic location may influence city-specific
SCh-mortality risk estimates. Similar to the 2008 SOx ISA, only few recent multicity
studies [i.e., (Chen et al. (2012b); Berglind et al. (2009); Wong et al. (2008b))1 conducted
extensive analyses of potential effect measure modifiers of the SCh-mortality relationship
as detailed in Chapter 6. These studies along with some single-city studies focusing on
SO2 and mortality provide limited evidence for potential differences in the risk of
SCh-related mortality by lifestage, sex, and socioeconomic status (SES).
Season and Weather
A limited number of studies have examined whether there is evidence of seasonal
differences or that certain weather patterns modify the SCh-mortality relationship. In the
2008 SOx ISA, only Zmirou et al. (1998) examined whether there are seasonal
differences in S02-mortality risk associations in a subset of the APHEA-1 cities. The
authors found some indication of larger associations in the summer months compared to
the winter months.
Since the completion of the 2008 SOx ISA, only a few recent studies have examined
whether there are seasonal differences in S02-mortality associations, and these studies
reported results consistent with Zmirou et al. (1998V In a study of 15 Italian cities,
meta-analysis of the Italian studies on short-term effects of air pollution (MISA-2),
5-285
-------
Bellini et al. (2007) is the only multicity study that examined whether there were seasonal
differences in SCh-mortality risk estimates. The authors found a similar pattern of
associations across mortality outcomes with SCh-mortality risk estimates being larger in
the summer compared to the winter (total mortality: summer 3.2% vs. winter 1.4%;
respiratory mortality: summer 12.0% vs. winter 4.1%; cardiovascular mortality: summer
9.4% vs. winter 1.6%). These results are consistent, with the only U.S.-based study that
examined seasonal patterns in SCh-mortality associations. In a study conducted in New
York City focusing on cardiovascular mortality, Ito etal. (2011) reported larger risk
estimates in the warm season [2.9% (95% CI: -1.2, 7.1)] compared to the cold season
[0.0% (95% CI: -1.7, 1.8)] for a 10-ppb increase in 24-h avg SO2 concentrations.
Instead of examining whether only specific seasons modify the S02-mortality
association, Vanos et al. (2013) focused on weather patterns, referred to as synoptic
weather types, in a study of 10 Canadian cities. Distinct weather types were identified by
combining a number of variables including temperature, dew point temperature, sea level
pressure, cloud cover, and wind velocity. Across the nine different synoptic weather
types examined for SO2, Vanos et al. (2013) reported that mortality risk estimates in all
age analyses tended to be larger in magnitude for dry versus moist weather types,
particularly in warmer seasons.
Overall, the limited number of studies that conducted seasonal analyses reported initial
evidence indicating larger S02-mortality associations during the summer season.
Additionally, there is preliminary evidence that specific weather patterns in combination
with certain seasons may modify the SCh-mortality association.
5.5.1.5 Sulfur Dioxide-Mortality Concentration-Response Relationship and Related
Issues
Lag Structure of Associations
Of the studies evaluated in the 2008 SOx ISA, the majority selected lag days a priori and
did not extensively examine the lag structure of associations for short-term SO2
exposures and mortality. These studies primarily focused on single- or multiday lags
within the range of 0-3 days. However, in a study in the Netherlands, Hoek (2003)
conducted more extensive analyses to examine whether there was evidence of immediate
or delayed S02-mortality effects. The authors provided preliminary evidence of larger
S02-mortality risk estimates at a multiday lag of 0-6 days compared to a single-day lag
(i.e., lag 1 day). Recent multicity studies have conducted additional analyses further
examining the lag structure of associations for short-term SO2 exposures and mortality.
5-286
-------
Chen et al. (2012b). within CAPES, examined individual lag days (lag day 0 to 7) and a
multiday lag of 0-1 days. As depicted in Figure 5-22. the authors found evidence of
immediate SO2 effects on mortality that slowly declined over time with the multiday lag
of 0-1 days exhibiting the largest risk estimate across mortality outcomes.
a
E
2.0
1.5
1.0 -
a)
b
= 0.5
C
•1.
'"-J
k_
a.
^ 0.0
-0.5
"ilUIi
01234567 01
Total mortality
01234567 01
Cardiovascular mortality
01234567 01
Respiratory mortality
Source: fChen et al.. 2012b1. Reprinted with permission of Elsevier.
Figure 5-22
Percent increase in daily mortality associated with a 10 |jg/m3
(3.62 ppb) increase in 24-h avg sulfur dioxide concentrations,
using various lag structures for sulfur dioxide in the China Air
Pollution and Health Effects Study cities, 1996-2008.
Kan et al. (2010) also examined the lag structure of associations for the SC^-mortality
relationship within the PAPA study, but did not examine an extensive number of
alternative lags, instead focusing on lag 0 and moving averages of 0-1 and 0-4 days
(Figure 5-23). Unlike Chen et al. (2012b). which focused on the combined risk estimate
across all cities, Kan et al. (2010) examined the lag structure of associations both within
individual cities and in a combined analysis across all PAPA cities. The results of both
the individual city and combined analysis are consistent with those observed by Chen et
al. (2012b) in CAPES (i.e., the effect largest in magnitude across the lag days examined
occurred primarily at lag 0-1 days) (Figure 5-22).
5-287
-------
BK
HK
SH
WH
0
Fixed-
effect
Combined
I
Random-
effect
Combined
II ;1
—I 1 1
0 0-1 0-4
-i 1 1—
0 0-1 0-4
-i 1 1—
0 0-1 0-4
-i 1 1—
0 0-1 0-4
n 1 1—
0 0-1 0-4
n 1 1—
0 0-1 0-4
BK = Bangkok; HK = Hong Kong; SH = Shanghai; WH = Wuhan.
Source: Kan et al. (20101. Reprinted with permission of Elsevier.
Figure 5-23 Percent increase in total mortality associated with a 10 |jg/m3
(3.62 ppb) increase in 24-h avg sulfur dioxide concentrations for
different lag structures in individual Public Health and Air
Pollution in Asia cities and in combined four city analyses.
Bellini et al. (2007) took a slightly different approach to examining the lag structure of
associations in a study of 15 Italian cities (MISA-2) by focusing on whether there was
evidence of mortality displacement. The authors reported larger SC^-mortality effects at
lag 0-15 days (3.8% for a 10-ppb increase in 24-h avg SO2 concentrations) compared to a
lag of 0-1 days (1.6%), which supports no evidence of mortality displacement.
Additional information on the lag structure can be observed by examining the percent
increase in mortality associated with short-term SO2 exposures at each individual lag day
of the lag 0-15-day model. The individual lag day results remained positive up to
approximately lag day 10, which is consistent with the results from Chen et al. (2012b)
(Figure 5-22). However, examining associations at single-day lags over a week, such as
10 days, may be uninformative due to potential inadequate control for weather variables
at these longer durations. Additionally, these longer lags may not be biologically
plausible due to controlled human exposure and animal toxicological studies
5-288
-------
demonstrating that effects attributed to SO2 exposure are rather immediate
(Section 5.2.1.2V
Overall, the limited analyses that have examined the lag structure of associations for
short-term SO2 exposures and mortality suggest that the greatest effects occur soon after
exposure (lag 0-1). However, the studies evaluated indicate that positive associations
may persist longer although the magnitude of those effects diminishes over time.
Concentration-Response Relationship
The studies evaluated in the 2008 SOx ISA (U.S. EPA. 2008d). as well as prior
assessments, have not conducted formal analyses of the SCh-mortality C-R relationship.
Although limited in number, a few recent studies published since the completion of the
2008 SOx ISA have conducted analyses to examine the shape of the S02-mortality C-R
relationship and whether a threshold exists in the combined C-R relationship across
multiple cities, or in an evaluation of single-city C-R relationships in the context of a
multicity study. However, these studies have not conducted extensive analyses examining
alternatives to linearity in the shape of the S02-mortality C-R relationship.
Using a subsampling approach, Moolgavkar et al. (2013) examined the shape of the C-R
relationship between short-term air pollution exposures and mortality in the NMMAPS
data set by applying a nonlinear function (i.e., natural splines with 6 df) to each pollutant.
As demonstrated in Figure 5-24. the analysis conducted by Moolgavkar et al. (2013)
provides support for a linear, no threshold relationship between short-term SO2 exposures
and total mortality.
5-289
-------
0.06
0.04
cc
cc
0.02
-0.02
0
10
20
30
40
50
60
Lag-1 S02
RR = relative risk; S02 = sulfur dioxide.
Note: Pointwise means and 95% confidence intervals adjusted for size of the bootstrap sample (d = 4).
Source: Reprinted from Environmental Health Perspectives; Moolaavkar et al. (2013).
Figure 5-24 Flexible ambient concentration-response relationship between
short-term sulfur dioxide (ppb) exposure (24-h avg
concentrations) and total mortality at lag 1.
In the four-city PAPA study, Kan et al. (2010) also examined the SC^-mortality C-R
relationship, but only focused on the shape of the C-R curve in each individual city. The
C-R curve for the SCh-mortality relationship was assessed by applying a natural spline
smoother with 3 df to SO2 concentrations. To examine whether the SCh-mortality
relationship deviates from linearity, the deviance between the smoothed (nonlinear)
pollutant model and the unsmoothed (linear) pollutant model was examined. When
examining the deviance, the authors only reported evidence for potential nonlinearity in
Hong Kong. However, across the cities, there is evidence of a linear, no threshold,
relationship within the range of SO2 concentrations where the data density is the highest,
specifically within the IQR (Figure 5-25). The linear relationship is most pronounced in
Shanghai and Wuhan, with evidence of an inverted U-shape for Bangkok and Hong
Kong. It should be noted, there is an overall lack of confidence in the shape of the C-R
curve at the high end of the distribution of SO2 concentrations in Bangkok and Shanghai
due to the lower data density within this range of concentrations observed in both cities.
A difficulty apparent in comparing the results across cities within Kan et al. (2010) is the
5-290
-------
drastically different range of SO2 concentrations in Bangkok and Hong Kong compared
with Shanghai and Wuhan. However, the cities with similar distributions of SO2
concentrations also have similar shapes to their respective SCh-mortality C-R curves.
Bangkok
10 20 30 40 50
S02 concentration (^g/ni3)
Shanghai
50 100 150
S02 concentration (^g/m3)
0.3
0.2
0.1
0.0
-0.1
-0.1
Hong Kong
LLli LL
0 20 40 60 SO 100
S02 concentration (f-tg/m3)
Wuhan
50 100 150
S02 concentration (^g/m3)
S02 = sulfur dioxide.
Note: x-axis is the average of lag 0-1 24-h avg S02 concentrations (|jg/m3). Solid lines indicate the estimated mean percent change
in daily mortality, and the dotted lines represent twice the standard error. Thin vertical lines represent the interquartile range of S02
concentrations within each city, while the thin vertical bar represents the World Health Organization guideline of 20 pg/m3 for a 24-h
avg time of S02.
Source: Reprinted from Environmental Health Perspectives; (Wong et al.. 2008b1).
Figure 5-25 Concentration-response curves for total mortality (degrees of
freedom = 3) for sulfur dioxide in each of the four Public Health
and Air Pollution in Asia cities.
5-291
-------
Both Moolgavkar et al. (2013) and Kan et al. (2010) examined the shape of the
SCh-mortality C-R relationship by focusing on all-cause (total) mortality. Additional
information on the shape of the C-R curve can be assessed in studies that focused on
cause-specific mortality as discussed in Section 5.2.1.8 (respiratory mortality) and
Section 5.3.1.9 (cardiovascular mortality). In studies of multiple Chinese cities, Meng et
al. (2013) and Chen et al. (2013) examined the shape of the C-R relationship for mortality
and short-term air pollution exposures on COPD and stroke mortality, respectively. In
both studies the authors conducted similar analyses of linearity by examining the
deviance between linear and spline models. Meng et al. (2013) and Chen et al. (2013)
both found no evidence of a deviation in linearity in the SO2-COPD mortality and
SCh-stroke mortality relationship, respectively (Figures 5-11 and 5-16).
To date, studies have conducted a rather limited exploration of potential alternatives to
linearity when examining the shape of the C-R relationship, which in combination with
the potential measurement error due to the spatial and temporal variability in SO2
concentrations (Sections 3.4.2.2 and 3.4.2.3). complicates the interpretation of the
SCh-mortality C-R relationship. With these limitations in mind, studies that examined the
C-R relationship provide initial evidence that indicates a log-linear, no threshold
relationship between short-term SO2 concentrations and mortality, specifically within the
range of SO2 concentrations where the data density is highest. Some differences in the
shape of the curve were observed on a city-to-city basis, which is consistent with the
mortality C-R results that have been reported for other criteria air pollutants.
5.5.1.6 Summary and Causal Determination
Recent multicity studies evaluated since the completion of the 2008 SOx ISA
continue to provide consistent evidence of positive associations between short-term
SO2 exposures and total mortality. Although the body of evidence is larger, key
uncertainties and data gaps still remain, which contribute to the conclusion that the
evidence for short-term SO2 exposures and total mortality is suggestive of, but not
sufficient to infer, a causal relationship. This conclusion is consistent with that reached
in the 2008 SOx ISA (U.S. EPA. 2008d). Recent multicity studies evaluated have further
informed key uncertainties and data gaps in the SCh-mortality relationship identified in
the 2008 SOx ISA including confounding, modification of the S02-mortality relationship,
potential seasonal differences in S02-mortality associations, and the shape of the
S02-mortality C-R relationship. However, questions remain regarding whether SO2 has
an independent effect on mortality, which can be attributed to: (1) the limited number of
studies that examined potential copollutant confounding, (2) the relative lack of
copollutant analyses with PM2 5, (3) and the evidence indicating attenuation of
5-292
-------
SCh-mortality associations in copollutant models with NO2 and PM10. Additionally, all of
the studies evaluated averaged SO2 concentrations over multiple monitors and used a
24-h avg exposure metric when assigning exposure, which may not adequately capture
the spatial and temporal variability in SO2 concentrations (Sections 3.4.2.2. and 3.4.2.3).
While correlations between 24-h avg and 1-h max SO2 concentrations are high (r > 0.75)
at most monitors, lower correlations may occur at some monitors and in individual
studies, which can add uncertainty to the ability of 24-h avg metrics to capture peak SO2
concentrations. This section describes the evaluation of evidence for total mortality, with
respect to the causal determination for short-term exposures to SO2 using the framework
described in Table II of the Preamble to the ISAs (U.S. EPA. 2015^. The key evidence,
as it relates to the causal framework, is summarized in Table 5-41.
Table 5-41 Summary of evidence, which is suggestive of, but not sufficient to
infer, a causal relationship between short-term sulfur dioxide
exposure and total mortality.
Rationale for Causal
Determination3
Key Evidence13
Key References'3
SO2
Concentrations
Associated with
Effects0
Consistent epidemiologic
evidence from multiple,
high quality studies at
relevant SO2
concentrations
Increases in mortality in multicity studies
conducted in the U.S., Canada, Europe, and
Asia
Section 5.5.1.2
Fiaure 5-15
Mean 24-h avg:
U.S., Canada,
South America,
Europe:
0.4-28.2d ppb
Asia:
0.7->200 ppb
Table 5-39
The magnitude of SO2 associations remained Sections 5.5.1.3,
positive, but were reduced in copollutant 3.4.3
models with PM10 and NO2. No studies
examined copollutant models with PM2.5. SO2
generally exhibits low to moderate
correlations with other NAAQS pollutants at
collocated monitors, and attenuation of SO2-
mortality association may be a reflection of
spatial variability among the pollutants.
U.S. studies that examine the association Section 3.4.2.2
between short-term SO2 exposures and
mortality rely on single or the average of
multiple monitors in an area and SO2
generally has low to moderate spatial
correlations across urban geographical
scales
Uncertainty regarding
potential confounding by
copollutants
Uncertainty regarding
exposure measurement
error
5-293
-------
Table 5-41 (Continued): Summary of evidence, which is suggestive of, but not
sufficient to infer, a causal relationship between short
term sulfur dioxide exposure and total mortality.
so2
Concentrations
Rationale for Causal Associated with
Determination3 Key Evidence13 Key References'3 Effects0
Uncertainty due to limited
coherence and biological
plausibility with
cardiovascular and
respiratory morbidity
evidence
Generally supportive, but not entirely Section 5.3.1.11
consistent epidemiologic evidence for Table 5-31
ischemic events such as triggering a
myocardial infarction. Inconclusive
epidemiologic and experimental evidence for
other cardiovascular endpoints. Uncertainties
with respect to the independent effect of SO2
on cardiovascular effects contributing to
limited coherence and biological plausibility
for S02-related cardiovascular mortality,
which comprises -35% of total mortality®
Consistent evidence of asthma exacerbations Section 5.2.1.9
from controlled human exposure studies Table 5-21
demonstrating respiratory effects
(i.e., respiratory symptoms and decreased
lung function) in response to typically
5-10-min exposures, with generally
supportive evidence from short-term SO2
exposure epidemiologic studies
demonstrating asthma-related morbidity,
specifically hospital admissions and ED
visits. Uncertainty as to the biological
mechanism that explains the continuum of
effects leading to SCh-related respiratory
mortality, which comprises -8% of total
mortality®
Avg = average; ED = emergency department; NAAQS = National Ambient Air Quality Standards; N02 = nitrogen dioxide;
PM10 = particulate matter with a nominal aerodynamic diameter less than or equal to 10 |jm; PM2.5 = in general terms, particulate
matter with a nominal aerodynamic diameter less than or equal to 2.5 |jm; S02 = sulfur dioxide.
aBased on aspects considered in judgments of causality and weight of evidence in causal framework in Tables I and II of the
Preamble to the ISAs (U.S. EPA. 2015b).
bDescribes the key evidence and references, supporting or contradicting, contributing most heavily to causal determination and,
where applicable, to uncertainties or inconsistencies. References to earlier sections indicate where full body of evidence is
described.
°Describes the S02 concentrations with which the evidence is substantiated.
dThe value of 28.2 represents the median concentration from Katsouvanni et al. (1997).
statistics taken from American Heart Association (2011).
5-294
-------
Collectively, the evidence from recent multicity studies of short-term SO2 exposures and
mortality consistently demonstrate positive SCh-mortality associations in single-pollutant
models. In the limited number of studies that conducted copollutant analysis, correlations
between SO2 and other pollutants were low (r < 0.4) to moderate (r = 0.4-0.7). Although
SCh-mortality associations remain positive in copollutant models with PM10 and NO2,
they were often attenuated to a large degree, questioning the independent effect of SO2 on
mortality. However, SO2 is more spatially variable than other pollutants as reflected in
the generally low to moderate spatial correlations across urban geographical scales
(Section 3.4.2.2); therefore, the attenuation in SO2 associations in copollutant models
may be a reflection of the different degree of exposure error across pollutants
(Section 3.4.3). It is important to note, the majority of recent studies that examined
potential copollutant confounding have been conducted in Asian countries where
correlations between pollutants may be higher, possibly limiting the generalizability of
results to other study areas where SO2 concentrations along with the concentrations of
other air pollutants are much lower. This is reflected in the results of Moolgavkar et al.
(2013) in a U.S. multicity study where there was very little evidence of attenuation of the
SCh-mortality association in copollutant models with PM10, whereas the multicity studies
conducted in Asian cities showed a rather pronounced reduction in SO2 associations. In
addition to copollutant analyses, recent studies examined the influence of the extent of
temporal adjustment and the lag structure for weather covariates on the S02-mortality
association. When examining, the extent of temporal adjustment, multiple studies
reported similar S02-mortality associations across a range of degrees of freedom per year.
Only Chen et al. (2012b') examined the lag structure for weather covariates, specifically
temperature, and found evidence of a difference in S02-mortality associations as the
number of lag days increased, but this could be attributed to the analysis being based on
only one covariate for temperature.
An examination of factors that may contribute to increased risk of S02-related mortality,
as discussed in Chapter 6. found evidence indicating that older adults (>65 years of age)
may be at increased risk with very limited evidence of potential differences by sex and
socioeconomic status. In the 2008 SOx ISA, initial evidence suggested potential seasonal
differences in S02-mortality associations, particularly in the summer months. A recent
multicity study conducted in Italy along with single-city studies conducted in the U.S.
add to this initial body of evidence suggesting larger associations during the summer or
warm months. Preliminary evidence indicates that not only season, but season in
combination with specific weather patterns, may modify the SCh-mortality association.
Additionally, an examination of different modeling approaches provides evidence that the
magnitude of the seasonal association may depend on the modeling approach employed
to control for the potential confounding effects of weather (Sacks et al.. 2012).
5-295
-------
Those studies that examined the lag structure of associations for the SCh-mortality
relationship generally observed that there is evidence of an immediate effect (i.e., lag 0 to
1 days) of short-term SO2 exposures on mortality. Multicity studies conducted in the U.S.
and Asia have examined the shape of the C-R relationship and whether a threshold exists
in both a multi- and single-city setting. These studies have used different statistical
approaches and consistently demonstrated a linear relationship with no evidence of a
threshold within the range of SO2 concentrations where the data density is highest. The
evidence of linearity in the SCh-mortality C-R relationship is further supported by studies
of cause-specific mortality as detailed in Section 5.3.1.9 (respiratory mortality) and
Section 5.2.1.9 (cardiovascular mortality). However, to date, studies have not conducted
extensive analyses exploring alternatives to linearity when examining the shape of the
SCh-mortality C-R relationship.
Overall, recent epidemiologic studies build upon and support the conclusions of the 2008
SOx ISA for total mortality. However, the biological mechanism that could lead to
mortality as a result of short-term SO2 exposures has not been clearly characterized. This
is evident when evaluating the underlying health effects (i.e., cardiovascular effects in
Section 5.3 and respiratory effects in Section 5.2) that could lead to cardiovascular
(-35% of total mortality) and respiratory (-9% of total mortality) mortality, the
components of total mortality most thoroughly evaluated (Hoyert and Xu. 2012). For
cardiovascular effects, the evidence is "inadequate to infer a causal relationship" with
exposure to short-term SO2 concentrations. An evaluation of epidemiologic studies that
examined the relationship between short-term SO2 exposure and cardiovascular effects
found a number positive associations but the evidence was not entirely consistent. Within
the collective body of evidence for cardiovascular effects, important uncertainties remain
especially regarding disentangling whether there is an independent effect of SO2 on
cardiovascular effects, which is the same uncertainty in total mortality studies. Overall,
this evidence complicates the interpretation of the relationship between SO2 and
cardiovascular mortality." For respiratory effects the evidence indicates a causal
relationship for short-term SO2 exposures. The strongest evidence for respiratory effects
is from studies examining S02-related asthma exacerbations, specifically controlled
human exposure studies demonstrating respiratory effects (i.e., respiratory symptoms and
decreased lung function) (Section 5.2.1.2) in people with asthma in response to
short-term, generally 5-10-minutes, SO2 exposures. The results from controlled human
exposure studies are generally supported by epidemiologic studies reporting
respiratory-related morbidity including hospital admissions and ED visits, specifically for
asthma. However, the biological mechanism that explains the continuum of effects that
could lead to respiratory-related mortality remains unclear. Additionally, it is important
to note epidemiologic studies that examine the association between short-term SO2
exposures and mortality rely on single or the average of multiple monitors over an area to
5-296
-------
assign exposure. Therefore, the exposure assessment approach used in the mortality
studies may contribute to exposure measurement error and underestimate associations
observed due to the spatially heterogeneous distribution of SO2 concentrations over a
wide area (Section 3.4.2.2). In conclusion, the consistent positive associations observed
across various multicity studies is limited by the uncertainty due to whether SO2 is
independently associated with total mortality, the representativeness of monitors and the
24-h avg SO2 exposure metric in capturing the spatial and temporal variability in
exposure to SO2 (Sections 3.4.2.2 and 3.4.2.3). and the uncertainty in the biological
mechanism that could lead to SC>2-induced mortality (Section 4.3). Collectively, this
body of evidence is suggestive, but not sufficient to conclude there is a causal
relationship between short-term SO2 exposure and total mortality.
5.5.2 Long-Term Exposure
In past reviews, a limited number of epidemiologic studies assessed the relationship
between long-term exposure to SO2 and mortality in adults. The 2008 SOx ISA
concluded that the scarce amount of evidence was "inadequate to infer a causal
relationship" (U.S. EPA. 2008d'). The 2008 SOx ISA identified concerns as to whether
the observed associations were due to SO2 alone, or if sulfate or other particulate SOx,
such as H2SO4, or PM indices could have contributed to these associations. The
possibility that the observed effects may not be due to SO2, but other constituents that
come from the same source as SO2, or that PM may be more toxic in the presence of SO2
or other components associated with SO2, could not be ruled out. Overall, a lack of
consistency across studies, inability to distinguish potential confounding by copollutants,
and uncertainties regarding the geographic scale of analysis limited the interpretation of
the causal relationship between long-term exposure to SO2 and mortality. Consistent with
the conclusion of the 2008 SOx ISA, the collective evidence informing the association
between long-term SO2 exposure and mortality continues to be limited. Despite the
improved consistency of the associations between long-term exposure to SO2 and both
respiratory and total mortality with the inclusion of recent cohort studies, these studies do
not address uncertainties identified in the 2008 SOx ISA.
This section includes a review of the evidence for an association between long-term
exposure to SO2 and mortality, integrating evidence presented in previous NAAQS
reviews with evidence that is newly available to this review. The evidence in this section
will focus on epidemiologic studies because experimental studies of long-term exposure
and mortality are generally not conducted. However, this section will draw from the
morbidity evidence presented for different health endpoints across the scientific
disciplines (i.e., animal toxicological, controlled human exposure studies, and
5-297
-------
epidemiology) to support the association observed for cause-specific mortality. Studies
are discussed by geographic region, with U.S. studies discussed in Section 5.5.2.1.
European studies in Section 5.5.2.2. and Asian studies in Section 5.5.2.3. Section 5.5.2.4
describes studies that evaluated the SCh-mortality relationship over small geographic
scales. A brief summary of the studies included in these sections can be found in
Table 5-42.
Table 5-42 Summary of studies of long-term exposure and mortality.
Correlation Selected Effect
Location Mean SO2 with Other Estimates
Study Years ppb Exposure Assessment Pollutants (95% Cl)a
tHartetal. (2011)
Cohort: N = 53,814
Deaths (all-causes):
4,806
Deaths (respiratory):
317
Deaths (COPD): 209
Deaths (lung cancer):
800
U.S.
(SO2:
1985-2000;
follow-up:
1985-2000)
4.8 Annual average
exposures based on
residential address from
model using spatial
smoothing and GIS-based
covariates; current
calendar year and
long-term average from
1985-2000
All causes:
1.09 (1.03, 1.15)
Respiratory:
1.10 (0.89, 1.35)
COPD:
0.93 (0.71, 1.22)
Lung cancer:
1.11 (0.98, 1.27)
Krewski et al. (2000) U.S.
HSC Cohort:
N = 8,111
Deaths (all-causes,
HSC): 1,239
ACS Cohort:
N = 559,049)
Deaths (all-causes,
ACS): 43,361
HSC:
(SO2:
1977-1985;
follow-up:
1974-1991)
ACS:
(SO2: 1980;
follow-up:
1982-1989)
HSC: HSC: mean levels from
1.6-24.0 fixed-site monitors
ACS: 9.3 ACS: City-specific annual
mean
HSC:
PM2.5: 0.85
SO4: 0.85
NO2: 0.84
All causes:
HSC:
1.05 (1.02, 1.09)
ACS:
1.06 (1.05, 1.07)
Lung cancer: HSC:
1.03 (0.91, 1.16)
Pope et al. (2002)
Cohort: N = 539,000
Deaths: NR
U.S.
(SO2:
1982-1998;
follow-up:
1982-1998)
6.7-9.7 Average across
monitoring stations in
each metropolitan area for
each study year using
daily average (i.e., 24-h
avg) concentrations,
averaged over 1 yr (1980)
and the entire study
period (1982-1998)
All causes:
1.03 (1.02, 1.05)
tLipfert et al. (2009)
Cohort: N = 67,938
Deaths (all-causes):
44,653
U.S.
(SO2: 1999;
follow-up:
1976-2001)
4.3 County-level estimates
from AER plume-in-grid
air quality model; based
on 1999 emissions
inventory from point and
area sources for
36 x 36-km grid squares
Subject-
weighted:
EC: 0.68
NOx: 0.65
SO42": 0.79
All causes:
1.02 (1.01, 1.03)
5-298
-------
Table 5-42 (Continued): Summary of studies of long term exposure and mortality.
Study
Location
Years
Mean SO2
PPb
Exposure Assessment
Correlation
with Other
Pollutants
Selected Effect
Estimates
(95% Cl)a
tKrewski et al. (2009) U.S.
Cohort: N = 513,450 (SO2: 1980;
Deaths: NR follow-up:
1982-2000)
9.6
City-specific annual mean
All causes:
1.02 (1.02, 1.03)
Lung cancer:
1.00 (0.98, 1.02)
Lipfert et al. (2006a)
Cohort: N = 70,000
Deaths: NR
U.S.
(SO2:
1999-2001;
follow-up:
1997-2001)
16.3 County-level "peak"
concentrations
Subject-
weighted:
PM2.5: 0.71
NO2: 0.41
Peak O3: 0.21
Peak CO:
0.41
SO42": 0.77
OC: 0.34
EC: -0.13
All causes:
0.99 (0.97, 1.01)
Abbey et al. (1999)
Cohort: N = 6,338
Deaths (all-causes):
1,575
Deaths (lung cancer):
30
U.S.
(SO2:
1966-1992;
follow-up:
1977-1992)
5.6 ZIP code-level mo
IQR. 3 7 averages cumulated and
averaged overtime
Mean
concentration:
PM10: 0.31
Os: 0.09
SO4: 0.68
When
exceeding
100 ppb (O3)
or 100 |jg/m3
(PM10)
PM10: -0.05
Os: 0.13
All causes:
Men:
1.07 (0.92, 1.25)
Women:
1.00 (0.88, 1.14)
Lung cancer:
Men:
2.52 (1.34, 4.77)
Women:
4.40 (2.34, 8.33)
Beelen et al. (2008b)
Cohort: N = 120,852
Deaths (all-causes):
17,610
Deaths (respiratory):
1,016
Deaths (lung cancer):
1,888
Netherlands
(SO2:
1976-1985,
1987-1996;
follow-up:
1987-1996)
5.2 IDW to regional
3Q- 1 g background monitors at
baseline residential
address
All causes:
0.94 (0.80, 1.10)
Respiratory:
0.92 (0.64, 1.31)
Lung cancer:
0.99 (0.73, 1.35)
Nafstad et al. (2004)
Cohort: N = 16,209
Deaths (all causes):
4,227
Deaths (respiratory):
200
Deaths (lung cancer):
382
Norway
(SO2:
1974-1995;
follow-up:
1972-1998)
3.6 Model results (per square
kilometer) for some
year/urban locations,
supplemented with
background monitoring
data
All causes:
0.97 (0.95, 1.01)
Respiratory:
1.04 (0.91, 1.19)
Lung cancer:
1.00 (0.91, 1.11)
Filleul et al. (2005)
Cohort: N = 14,284
Deaths (all causes):
2,396
France
(SO2:
1974-1976;
follow-up:
1974-2000)
3.0-8.2 3-yr mean concentrations BS: 0.29
for 24 areas in seven TSP: 0.17
different cities NO-0.01
NO2 -0.10
All causes:
1.01 (0.99, 1.04)
Lung cancer:
0.99 (0.90, 1.09)
5-299
-------
Table 5-42 (Continued): Summary of studies of long term exposure and mortality.
Study
Location
Years
Mean SO2
PPb
Exposure Assessment
Correlation
with Other
Pollutants
Selected Effect
Estimates
(95% Cl)a
tBentaveb et al.
(2015)
Cohort: N = 20,327
Deaths (all-causes):
1,967
Deaths (respiratory):
284
Deaths (CVD): 165
France 2.3
(SO2:
1989-2008;
follow-up:
1989-2013)
Annual concentrations
from CHIMERE
chemical-transport model,
2-km resolution
Os: -0.13
PM2.5: 0.58
PM10: 0.57
PM10-2.5: 0.30
NO2: 0.56
All causes: 1.23
(0.98, 1.52)
Respiratory: 0.76
(0.43, 1.33)
CVD: 0.85 (0.44,
1.67)
tHansell et al. (2016)
Cohort: N = 367,658
Deaths (all causes):
47,775
Deaths (respiratory):
5,300
Deaths (COPD):
2,413
Deaths (lung cancer):
3,154
Deaths (CVD):
23,923
England
(SO2: 1971,
1981, 1991;
follow-up:
1971-2009)
1971
1981
1991
32.4
16.4
11.2
LUR models for annual
concentrations in 1971,
1981, and 1991
1991
All causes: 1.09
(1.05, 1.15)
Respiratory: 1.20
(1.09, 1.33)
COPD: 1.43 (1.23,
1.66)
Lung cancer: 1.29
(1.12, 1.47)
CVD: 1.05 (0.99,
1.13)
tCarev et al. (2013)
Cohort: N = 823,442
Deaths (all causes):
81,636
Deaths (respiratory):
10,408
Deaths (lung cancer):
5,192
England
(SO2: 2002;
follow-up:
2003-2007)
1.5
SD: 0.8
IQR: 0.8
Annual mean for 1-km PM10: 0.45
grid cells from air NO2: 0.37
dispersion models (poor O3: -0.41
validation results for SO2)
All causes:
1.26 (1.19, 1.34)
Respiratory:
1.67 (1.42, 1.97)
Lung cancer:
1.34 (1.06, 1.58)
5-300
-------
Table 5-42 (Continued): Summary of studies of long term exposure and mortality.
Study
Location
Years
Mean SO2
PPb
Exposure Assessment
Correlation
with Other
Pollutants
Selected Effect
Estimates
(95% Cl)a
tAncona et al. (2015) Rome, Italy 2.5 |jg/m3 Lagrangian particle
Cohort: N = 85,559
Deaths (all causes):
5,878
Deaths (CVD): 2,240
Deaths (respiratory):
384
(SOx:
2001-2010;
follow-up:
2001-2010)
SOx
SD: 0.9
dispersion model (SPRAY
Ver. 5) used SOx as
exposure marker for
petrochemical refinery
emissions
PM10: 0.81
H2S: 0.78
All causes:
Men:
1.04 (0.92, 1.18)
Women:
0.93 (0.81, 1.07)
CVD:
Men:
1.08 (0.89, 1.31)
Women:
1.00 (0.81, 1.25)
IHD:
Men:
1.05 (0.79, 1.41)
Women:
1.25 (0.89, 1.75)
Respiratory:
Men:
1.31 (0.88, 1.95)
Women:
0.64 (0.32, 1.28)
tCao et al. (2011)
Cohort: N = 70,947
Deaths (all causes):
8,319
China
(SO2:
1991-2000;
follow-up:
1991-2000)
27.7 Annual average by linking
fixed-site monitoring data
with residential ZIP code
All causes:
1.02 (1.02, 1.03)
CVD:
1.02 (1.00, 1.03)
Respiratory:
1.04 (1.02, 1.06)
Lung cancer:
1.06 (1.03, 1.08)
tChen et al. (2016)
Cohort: N = 39,054
Deaths (all causes):
1,353
Deaths (lung cancer):
140
China
(SO2:
1998-2009;
follow-up:
1998-2009)
25.5 1-yr avg and time-varying
exposure from monitoring
stations calculated from
24-h avg
Lung cancer:
1.02 (1.01, 1.03)
tDonq et al. (2012)
Cohort: N = 9,941
Deaths (all causes):
505
Deaths (respiratory):
72
China
(SO2:
1998-2009;
follow-up:
1998-2009)
23.9
SD: 5.7
1-yr avg from five
monitors
Respiratory:
1.05 (0.96, 1.16)
tZhanq et al. (2011)
Shenyang, 23.9
1-yr avg and yearly
All causes:
Cohort: N = 9,941
China
deviations in each of five
0.93 (0.90, 0.99)
Deaths (all causes):
256
(SO2:
1998-2009;
monitoring stations
calculated from 24-h avg
follow-up:
1998-2009)
5-301
-------
Table 5-42 (Continued): Summary of studies of long term exposure and mortality.
Study
Location
Years
Mean SO2
PPb
Exposure Assessment
Correlation
with Other
Pollutants
Selected Effect
Estimates
(95% Cl)a
tKatanoda et al.
(2011)
Cohort: N = 63,520
Deaths (all causes):
6,687
Deaths (respiratory):
690
Deaths (lung cancer):
518
Japan
(SO2:
1974-1983;
follow-up:
1983-1995)
2.4-19.0 Annual mean
concentrations from
monitoring station near
each of eight study areas
Pearson:
SPM: 0.47
Respiratory:
1.20 (1.15, 1.24)
COPD:
1.15 (0.94, 1.41)
Pneumonia:
1.20 (1.16, 1.25)
Lung cancer:
1.12 (1.03, 1.22)
Elliott et al. (2007)
Cohort: N = 662,343
Deaths (all causes):
52,792
Deaths (respiratory):
8,471
Deaths (lung cancer):
3,473
Great Britain
(SO2:
1966-1970,
1990-1994;
follow-up:
1982-1986,
1994-1998)
12.2-41.4 4-yr exposure windows
from annual average
concentrations from
monitoring sites located in
residential areas
All causes:
1.02 (1.02, 1.02)
Respiratory:
1.06 (1.06, 1.07)
Lung cancer:
1.00 (0.99, 1.01)
tBennett et al. (2014)
Warwickshire,
NR
Single recorded level for
Heart failure:
Ecologic study at
U.K.
each ward from 2010
1.11 (0.988, 1.22)
Ward level
(SO2: 2010;
mortality
data:
2007-2012)
tWana et al. (2009)
Brisbane,
5.4
1-h max from
Cardiopulmonary:
No individual data
Australia
13 monitoring stations
1.26 (1.03, 1.54)
(SO2:
aggregated to annual
1996-2004;
means used with IDW
follow-up:
1996-2004)
tWanq et al. (2014a)
China
46.31
Annual average across
Life expectancy:
Hypothetical cohort
(SO2:
monitoring stations in
10-|jg/m3 increase in
2004-2010;
85 city regions
SO2 correlated with
life table:
0.28-0.47 yr
2010)
decrease in life
expectancy
ACS = American Cancer Society; AER = Atmospheric and Environmental Research; avg = average; BS = black smoke;
CHIMERE = regional chemistry transport model; CI = confidence interval; CO = carbon monoxide; COPD = chronic obstructive
pulmonary disease; CVD = cardiovascular disease; EC = elemental carbon; GIS = geographic information systems;
H2S = hydrogen sulfide; HSC = Harvard Six Cities; IDW = inverse distance weighting; IHD = ischemic heart disease;
IQR = interquartile range; LUR = land use regression; max = maximum; ; n = sample size; N = population number; NO = nitric
oxide; N02 = nitrogen dioxide; NOx = the sum of NO and N02; NR = not reported; 03 = ozone; OC = organic carbon;
PM2.5 = particulate matter with a nominal aerodynamic diameter less than or equal to 2.5 |jm; PM10 = particulate matter with a
nominal aerodynamic diameter less than or equal to 10 |jm; PM10-2.5 = particulate matter with a nominal aerodynamic diameter
less than or equal to 10 |jm and greater than a nominal 2.5 |jm; SD = standard deviation; S02 = sulfur dioxide; S04 = sulfate;
S042" = sulfate ion; SOx = sulfur oxide; SPM = suspended particulate matter; TSP = total suspended solids.
aEffect estimates are standardized per 5-ppb increase in S02 concentrations.
fStudies published since the 2008 ISA for Sulfur Oxides.
5-302
-------
5.5.2.1
U.S. Cohort Studies
A number of longitudinal cohort studies have been conducted in the U.S. and have found
small, statistically significant positive associations between long-term exposure to SO2
and total mortality (Hart ct al.. 2011; Lipfert et al.. 2009; Pope et al.. 2002; Krewski et
al.. 2000). The body of evidence is smaller and less consistent when these studies
examine cause-specific mortality, although Hartetal. (2011) observed positive, yet
imprecise associations (i.e., wide 95% confidence intervals) with respiratory, lung cancer,
and cardiovascular mortality. In the Trucking Industry Particle Study, Hartetal. (2011)
used the work records for over 50,000 men employed in four U.S. trucking companies to
identify all-cause and cause-specific mortality. Occupational exposures were assigned
based on job title, while exposures to ambient air pollution (i.e., PM10, SO2, and NO2
averaged over the study period) were determined using spatial smoothing and geographic
information system (GlS)-based covariates using residential address. All three pollutants
were independently associated with all-cause mortality, with central estimates the highest
for the association with NO2 and lowest for the association with PM10. Both NO2 and SO2
were positively associated with lung cancer, cardiovascular disease, and respiratory
disease mortality, and negatively associated with COPD mortality. Correlation
coefficients between SO2 and other measured air pollutants were not reported, making it
difficult to evaluate for the potential of copollutant confounding on the associations
attributed to SO2. There was no evidence of confounding by occupational exposures
(based on job-title).
The Harvard Six Cities study is a prospective cohort study of the effects of air pollution
with the main focus on PM components in six U.S. cities and provides limited evidence
for an association between mortality and exposure to SO2. Cox proportional hazards
regression was conducted with data from a 14- to 16-year follow-up study of 8,111 adults
in the six cities. Dockery et al. (1993) reported that lung cancer and cardiopulmonary
mortality were more strongly associated with the concentrations of inhalable and fine PM
and sulfate particles than with the levels of TSP, SO2, NO2, or acidity of the aerosol.
Krewski etal. (2000) conducted a sensitivity analysis of the Harvard Six Cities study and
examined associations between gaseous pollutants (i.e., O3, NO2, SO2, and CO) and
mortality, observing positive associations between SO2 and total mortality and
cardiopulmonary deaths. In this data set SO2 was highly correlated with PM2 5 (r = 0.85),
sulfate (r = 0.85), and NO2 (r = 0.84), making it difficult to attribute the observed
associations to an independent effect of SO2.
Pope et al. (1995) investigated associations between long-term exposure to PM and the
mortality outcomes in the American Cancer Society (ACS) cohort and provides limited
5-303
-------
evidence for an association between exposure to SO2 and mortality. Ambient air pollution
data from 151 U.S. metropolitan areas in 1981 were linked with individual risk factors in
552,138 adults who resided in these areas when enrolled in the prospective study in 1982.
Death outcomes were ascertained through 1989. Gaseous pollutants were not analyzed in
the original analysis. Extensive reanalysis of the ACS data, augmented with additional
gaseous pollutants data, showed positive associations between mortality and SO2, but not
for the other gaseous pollutants (Jcrrctt et al.. 2003; Krewski et al.. 2000). Pope et al.
(2002) extended analysis of the ACS cohort with double the follow-up time (to 1998) and
triple the number of deaths compared to the original study (Pope et al.. 1995). Both PM2 5
and SO2 were associated with all the mortality outcomes, although only SO2 was
associated with the deaths attributable to "all other causes." The association of SO2 with
mortality for "all other causes" makes it difficult to interpret the effect estimates due to a
lack of biological plausibility for this association. More recently, Krewski et al. (2009)
conducted an extended reanalysis of the study conducted by Pope et al. (2002). including
examination of ecologic covariates (e.g., education attainment, housing characteristics,
income) and evaluation of exposure windows. The inclusion of ecologic covariates
generally resulted in increased risk estimates, with the greatest effect on mortality from
IHD. The authors also evaluated individual time-dependent exposure profiles to examine
whether there is a critical exposure time window most strongly associated with mortality
from ambient air pollution. The time window immediately preceding death (1-5 years)
produced the strongest effects for mortality associated with exposure to SO2, while later
time windows (6-10 years and 11-15 years) generally showed null associations between
SO2 and mortality.
Lipfert et al. (2000) conducted an analysis of a national cohort of -70,000 male U.S.
military veterans who were diagnosed as hypertensive in the mid-1970s and were
followed up for about 21 years (up to 1996) and provides scant evidence for an
association between exposure to SO2 and mortality. This cohort was 35% black and 57%
of the cohort were current smokers (81% of the cohort had been smokers at one time).
PM2 5, PM10, particulate matter with a nominal aerodynamic diameter less than or equal
to 10 (im and greater than a nominal 2.5 ^m (PM10-2.5), TSP, sulfate, CO, O3, NO2, SO2,
and lead (Pb) were examined in these analyses. The county of residence at the time of
entry to the study was used to estimate exposures. Four exposure periods (from 1960 to
1996) were defined, and deaths during each of the three most recent exposure periods
were considered. The results for SO2 as part of their preliminary screening were generally
null. Lipfert et al. (2000) noted that Pb and SO2 were not found to be associated with
mortality, and thus, were not considered further. They also noted that the pollution effect
estimates were sensitive to the regression model specification, exposure periods, and the
inclusion of ecological and individual variables. The authors reported that indications of
concurrent mortality risks were found for NO2 and peak O3. In a subsequent analysis,
5-304
-------
Lipfert et al. (2006b) examined associations between traffic density and mortality in the
same cohort, extending the follow-up period to 2001. As in their previous study (Lipfert
et al.. 2000). four exposure periods were considered but included more recent years, and
reported that traffic density was a better predictor of mortality than ambient air pollution
variables with the possible exception of O3. The log-transformed traffic density variable
was only weakly correlated with SO2 (r = 0.32) and PM2 5 (r = 0.50) in this data set.
Lipfert et al. (2006a) further extended analysis of the veterans' cohort data to include the
U.S. EPA's Speciation Trends Network (STN) data, which collected chemical
components of PM2 5. They analyzed the STN data for 2002, again using county-level
averages. PM2 5 and gaseous pollutant data for 1999 through 2001 were also analyzed. As
in the previous study (Lipfert et al.. 2006b). traffic density was the most important
predictor of mortality, but associations were also observed for elemental carbon,
vanadium, nickel, and nitrate. Ozone, NO2, and PM10 also showed positive but weaker
associations. Once again, no associations were observed between long-term exposure to
SO2 and mortality. Lipfert et al. (2009) re-examined these associations, this time
averaging the exposure variables over the entire follow-up period (1976-2001). For this
exposure period, they observed positive associations between SO2 and mortality. When
the data set was stratified by county-level traffic density, the SO2 association with
mortality was stronger in the counties with high density traffic, and attenuated to near
null in the counties with lower traffic density. The fact that the association between
long-term exposure to SO2 and mortality is only observed in areas where traffic density
has been characterized as high, along with the moderate to strong correlations between
SO2 and other traffic-related pollutants (e.g., PM2 5, N02,NOx, EC) in these analyses,
makes it difficult to discern whether these associations are truly attributable to SO2, or
could be due to some other traffic-related pollutant or mixture of pollutants.
Abbey et al. (1999) investigated associations of long-term ambient concentrations of
PM10, sulfate, SO2, O3, and NO2 with mortality in a cohort of 6,338 nonsmoking
California Seventh-Day Adventists. Monthly indices of ambient air pollutant
concentrations at 348 monitoring stations throughout California were interpolated to ZIP
codes according to home or work location of study participants, cumulated, and then
averaged over time. They reported associations between PM10 and total mortality for
males and nonmalignant respiratory mortality for both sexes. SO2 was positively
associated with total mortality for males but not for females. Generally, null associations
were observed for cardiopulmonary deaths and respiratory mortality for both males and
females.
Overall, the majority of the limited evidence informing the association between long-term
exposure to SO2 and mortality from U.S. cohort studies was included in the 2008 SOx
ISA. A recent cohort study of male truck drivers (Hartetal.. 2011) provided some
5-305
-------
additional evidence for an association between long-term exposure to SO2 and both
respiratory mortality and total mortality, while updates to the ACS (Krcwski et al.. 2009)
and Veterans (Lipfcrt et al.. 2009) cohort studies provides some limited evidence for an
association with total mortality, although none of these recent studies help to resolve the
uncertainties identified in the 2008 SOx ISA related to copollutant confounding or the
geographic scale of the analysis.
5.5.2.2 European Cohort Studies
A number of European cohort studies examined the association between SO2
concentrations and both total mortality and cause-specific mortality and found generally
inconsistent results. Beelen et al. (2008b) analyzed data from the Netherlands Cohort
Study on Diet and Cancer with 120,852 subjects. Traffic-related pollutants (BS, NO2,
SO2, PM2 5), and four types of traffic-exposure estimates were analyzed. While the local
traffic component was estimated for BS, NO2, and PM2 5, no such attempt was made for
SO2. Traffic intensity on the nearest road was associated with all-cause mortality and a
larger RR was observed for respiratory mortality. Results were similar for BS, NO2, and
PM2 5, but no associations were observed for SO2.
Several studies noting declining SO2 concentrations during the follow-up period (from
the mid-1970s through the mid-1990s) did not observe positive associations with
mortality. Nafstad et al. (2004) linked data from 16,209 males (aged 0 to 49 years) living
in Oslo, Norway with data from the Norwegian Death Register and with estimates of the
average annual air pollution levels at the participants' home addresses. PM was not
considered in this study because measurement methods changed during the study period.
Exposure estimates for NOx and SO2 were constructed using models based on subject
addresses, emission data for industry, heating, and traffic, and measured concentrations.
While NOx was associated with total, respiratory, lung cancer, and ischemic heart disease
deaths, SO2 did not show any associations with mortality. In this study, SO2 levels were
reduced by a factor of 7 during the study period (from 5.6 ppb in 1974 to 0.8 ppb in
1995), whereas NOx did not show any clear downward trend. Filleul et al. (2005) linked
daily measurements of SO2, TSP, BS, NO2, and nitric oxide (NO) with data on mortality
for 14,284 adults who resided in 24 areas from seven French cities enrolled in the Air
Pollution and Chronic Respiratory Diseases survey in 1974. Models were run before and
after exclusion of six area monitors influenced by local traffic as determined by an
NO:N02 ratio of >3. Before exclusion of the six areas, none of the air pollutants was
associated with mortality outcomes. After exclusion of these areas, analyses showed
associations between total mortality and TSP, BS, NO2, and NO but not SO2 or
acidimetric measurements. In this study, SO2 levels declined by a factor of two to three
5-306
-------
(depending on the city) between the 1974 through 1976 period and the 1990 through
1997 period. The changes in air pollution levels over the study period complicate
interpretation of reported effect estimates.
Carey et al. (2013) examined the associations between long-term exposure to ambient air
pollutants and total and cause-specific mortality in a national English cohort
(n = 835,607). The authors used air dispersion models to estimate annual mean air
pollution concentrations for 1-km grid cells for a single year prior to the follow-up
period. Model validation using national air quality monitors and networks demonstrated
good agreement for NO2 and O3, moderate agreement for PM10 and PM2 5, but relatively
poor agreement for SO2 (R2 = 0-0.39). The authors observed positive associations with
total mortality for all of the air pollutants, and these associations were stronger for PM2 5,
NO2, and SO2 with respiratory and lung cancer mortality. Associations were generally not
observed between cardiovascular mortality and any of the pollutants. Although the
authors observed positive associations between SO2 and mortality (especially respiratory
mortality), these associations are difficult to interpret due to the poor validation of the
dispersion model for SO2. Ancona et al. (2015) used a Lagrangian particle dispersion
model (see Section 3.3.2.4 for details) to estimate annual means of SOx (as an exposure
marker for emissions from a petrochemical refinery) in Rome, Italy and associations with
all-cause and cause-specific mortality among men and women. The authors did not
present any validation results for their dispersion model. Predicted concentrations of SOx
were highly correlated with predicted concentrations of PM10 (r = 0.81), and because SOx
was used as an exposure marker for petrochemical refinery emissions, it would likely be
correlated with other stack or fugitive refinery emissions, including PM2 5 and VOCs. The
authors observed associations for all-cause mortality and CVD mortality that were near
the null value for both men and women. When restricted to IHD mortality, the association
remained near the null value for men, but was elevated among women. Conversely,
slightly increased risks were observed for respiratory mortality and mortality due to
digestive diseases among men, while the risks for these were attenuated among women.
Due to the unknown validity of the dispersion model and the high correlations with
additional copollutants, it is difficult to interpret these associations.
Overall, the results of the European cohort studies provide very little evidence for an
association between long-term exposure to SO2 and mortality. The majority of these
studies were included in the 2008 SOx ISA (Beelen et al.. 2008b; Filleul et al.. 2005;
Nafstad et al.. 2004). Only the study by Carey et al. (2013) provided new evidence for
this review. None of the studies used copollutant models or accounted for potential
confounding or effect measure modification by other ambient air pollutants, including
sulfate. The study by Carey et al. (2013) had the potential to inform uncertainties related
to the geographic scale of the exposure assessment; however, the poor validation results
5-307
-------
of the dispersion model used to estimate the SO2 concentrations for 1-km grid cells
makes it difficult to interpret these results.
5.5.2.3 Asian Cohort Studies
Four recent cohort studies have been conducted in China to examine the association
between long-term exposure to SO2 and mortality (Chen et al.. 2016; Dong et al.. 2012;
Cao etal.. 2011; Zhang et al.. 2011) and observed inconsistent results. Each of these
studies used annual area-wide average concentrations from fixed-site monitoring stations
to assign exposure. Notably, the mean SO2 concentrations in these study areas were much
higher than concentrations observed in other locations (see Table 5-42). Cao etal. (2011)
observed generally modest positive associations with all-cause, respiratory and lung
cancer mortality. Chen et al. (2016) observed a positive association with lung cancer
mortality, although the correlation between SO2 and PM10 was high (r > 0.94), and it is
possible that copollutant confounding could at least partially explain this relationship.
Dong et al. (2012) observed a modest, positive association with respiratory mortality,
while Zhang etal. (2011) observed modest negative associations with all-cause mortality.
Katanoda et al. (2011) conducted a cohort study in Japan investigating the association
between long-term exposure to PM2 5, NO2, and SO2 and lung cancer and respiratory
mortality. The authors used annual mean concentrations from fixed-site monitoring
stations near each of eight study areas. The authors observed positive associations of
long-term exposure to PM2 5, NO2, and SO2 with lung cancer and respiratory mortality,
with the strongest effect observed for the SO2 associations.
Overall, these recent Asian cohort studies provide some new evidence of an association
between long-term exposure to SO2 and mortality; however, they generally report similar
associations for other ambient air pollutants, and do not evaluate for potential bias due to
copollutant confounding (using copollutants models, reporting correlation coefficients
between SO2 and other measured pollutants, or other methods). Generally, these recent
studies do not help to resolve the uncertainties identified in the 2008 SOx ISA related to
copollutant confounding or the geographic scale of the analysis.
5.5.2.4 Cross-Sectional Analysis Using Small Geographic Scale
Elliott et al. (2007) examined associations of BS and SO2 with mortality in Great Britain
using a cross-sectional analysis. However, unlike the earlier ecological cross-sectional
mortality analyses in the U.S. in which mortality rates and air pollution levels were
compared using large geographic boundaries (i.e., metropolitan statistical areas or
5-308
-------
counties), Elliott et al. (2007) compared the mortality rates and air pollution
concentrations using a much smaller geographic unit, the electoral ward, with a mean
area of 7.4 km2 and a mean population of 5,301 per electoral ward. Of note, SO2 levels
declined from 41.4 ppb in the 1966 to 1970 period to 12.2 ppb in 1990 to 1994. This type
of analysis does not allow adjustments for individual risk factors, but the study did adjust
for socioeconomic status data available for each ward from the 1991 census. Social
deprivation and air pollution were more highly correlated in the earlier exposure
windows. They observed positive associations for both BS and SO2 with mortality
outcomes. The estimated effects were stronger for respiratory illness than other causes of
mortality for the most recent exposure period and most recent mortality period (when
pollution levels were lower). The adjustment for social deprivation reduced the effect
estimates for both pollutants. Simultaneous inclusion of BS and SO2 reduced effect
estimates for BS but not SO2. Elliott et al. (2007) noted that the results were consistent
with those reported in the Krewski et al. (2000) reanalysis of the ACS study. Similarly,
Bennett et al. (2014) observed a positive association between ward-level SO2
concentrations measured in 2010 and ward-level data on heart failure mortality from
2007-2012 in Warwickshire, U.K. Stronger associations were observed for estimated
benzene exposure in this population, while estimated PM exposure was inversely
associated with heart failure mortality. These analyses are ecological, but the exposure
estimates in the smaller area compared to that in the U.S. cohort studies may have
resulted in less exposure measurement error, and the large underlying population appears
to be reflected in the narrow confidence bands of effect estimates.
In a recent cross-sectional analysis, Wang et al. (2009) examined the long-term exposure
to gaseous air pollutants (i.e., NO2, O3, and SO2) and cardio-respiratory mortality in
Brisbane, Australia. Pollutant concentrations were estimated for small geographic units,
statistical local areas, using inverse distance weighting. The authors observed a positive
association between cardio-respiratory mortality and SO2, but generally null associations
for NO2 and O3.
The results of these cross-sectional studies are inconsistent, with much higher mortality
effects attributed to SO2 in Brisbane, Australia (Wang et al.. 2009) and Warwickshire,
U.K. (Bennett et al.. 2014) than throughout Great Britain (Elliott et al.. 2007). While each
of these studies took a geospatial approach to their analyses, the cross-sectional nature of
the study designs and the lack of control for potential bias due to copollutant confounding
limit the interpretation of their results.
5-309
-------
5.5.2.5
Summary and Causal Determination
Overall, the majority of the limited evidence informing the association between long-term
exposure to SO2 and mortality was included in the 2008 SOx ISA. The 2008 SOx ISA
identified concerns regarding (1) the consistency of the observed associations;
(2) whether the observed associations were due to SO2 alone, or if sulfate or other
particulate SOx or PM indices could have contributed to these associations; and (3) the
geographic scale of the exposure assessment. Specifically, the 2008 SOx ISA noted the
possibility that the observed effects may not be due to SO2, but other co-occurring
pollutants that come from the same source as SO2, or that PM may be more toxic in the
presence of SO2 or other components associated with SO2, could not be ruled out. None
of the epidemiologic studies made corrections or adjustments for exposure measurement
error or accounted for the potential for bias away from the null, the potential for which
has been demonstrated in simulation studies (see Section 3.4.4.2). A lack of consistency
across studies, inability to distinguish potential confounding by copollutants, and
uncertainties regarding the geographic scale of analysis limited the interpretation of the
causal relationship between long-term exposure to SO2 and mortality.
Figure 5-26 presents total mortality effect estimates associated with long-term exposure
to SO2. The overall range of effects spans 0.93 to 1.26 per 5-ppb increase in the annual
(or longer period) average SO2 concentration. The analyses of the Harvard Six Cities and
the ACS cohort data, which likely provide effect estimates that are most useful for
evaluating possible health effects in the U.S., observed effect estimates of 1.02 to 1.06,
while the effect estimate from the recent cohort study of truck drivers was 1.09. Note that
each of the U.S. cohort studies has its own advantages and limitations. The Harvard Six
Cities data have a small number of exposure estimates, but the study cities were carefully
chosen to represent a range of air pollutant exposures. The ACS cohort had far more
subjects, but the population was more highly educated than the representative U.S.
population. Because educational status appeared to be an important effect modifier of air
pollution effects in both studies, the overall effect estimate for the ACS cohort may not
be generalizable to the broader population. The evidence from the cohort studies
conducted in Europe and Asia is generally similar to that observed from the U.S. cohort
studies. That is, the magnitude of the effect estimates is generally similar, although there
is greater inconsistency in the direction of the association. Also, the effect estimate
observed by Carey et al. (2013) is much higher than that observed in any of the other
studies. Generally, these results are consistent with a recent study (Wang et al.. 2014a')
that evaluated the correlation between life expectancy and SO2 concentrations in
85 major city regions in China. After accounting for a surrogate for socioeconomic status,
they observed that city regions with higher SO2 concentrations were correlated with
lower life expectancies.
5-310
-------
Figure 5-27 presents the cause-specific mortality effect estimates associated with
long-term exposure to SO2. The overall range of effects spans 0.93 to 4.40 per 5-ppb
increase in the annual (or longer period) average SO2 concentration. Generally, there was
a trend toward more positive associations for respiratory and lung cancer mortality
compared to cardiopulmonary, cardiovascular, and other causes of death. Specifically,
recent studies examining respiratory mortality provide some evidence that this cause of
death may be more consistently associated with long-term exposure to SO2 than other
causes of death. This is consistent with both the short- and long-term exposure to SO2
that are associated with respiratory effects.
The recent evidence is generally consistent with the evidence in the 2008 SOx ISA. The
biggest notable difference is in the improved consistency in the association between
long-term exposure to SO2 and both respiratory and total mortality that comes from the
inclusion of recent cohort studies. However, none of these recent studies help to resolve
the uncertainties identified in the 2008 SOx ISA related to copollutant confounding or the
geographic scale of the analysis. All available evidence for mortality due to long-term
exposure to SO2 was evaluated using the framework described in Table II of the
Preamble to the IS As (U.S. EPA. 2015bV The key evidence as it relates to the causal
framework is summarized in Table 5-43. The overall evidence is inadequate to infer a
causal relationship between long-term exposure to SO2 and total mortality among adults.
5-311
-------
Study
Location Mean Notes
(PPb)
tHartetal. 2011
USA
4.8
Men
Krewski et al. 2000
USA
1.6-24.0 HSC
Krewski et al. 2000
USA
9.3
ACS
Pope et al. 2002
USA
6.7-9.7
ACS
+Krewski et al. 2009
USA
9.6
ACS
Lipfert et al. 2006
USA
16.3
Men
+Lipfert et al. 2009
USA
4.3
Men
Abbey et al. 1999
USA
5.6
Men
Abbey et al. 1999
USA
5.6
Women
Nafstad et al. 2004
Norway
3.6
Men
Beelen et al. 2008
The Netherlands
5.2
Case-cohort -
Filleul et al. 2005
France
3.0-8.2
+Bentayeb et al. 2015
France
2.3
+Carey et al. 2013
England
1.5
+Hansell et al. 2016
England
32.4
1971
+Hansell et al. 2016
England
16.4
1981
+Hansell et al. 2016
England
11.2
1991
Elliott et al. 2007
Great Britain
12.2-41.4
tCao et al. 2011
Eastern China
27.7
+Zhang et al. 2011
Shenyang, China
23.9
O.S
0.9 1 1.1 1.2 1.3 1.4
Relative Risk (95% Confidence Interval)
ACS = American Cancer Society Study; HSC = Harvard Six Cities Study.
Note: studies in red are recent studies. Studies in black were included in the 2008 ISA for Sulfur Oxides. Relative risks are
standardized to a 5-ppb increase in sulfur dioxide concentrations. Corresponding quantitative results are reported in Supplemental
Table 5S-31 (U.S. EPA. 2017c).
Figure 5-26 Relative risks (95% confidence interval) of sulfur
dioxide-associated total mortality.
5-312
-------
Study
tHartetal. 2011
Nafstad et al. 2004
Beelen et al. 2008
Elliott et al. 2007
tBentayeb et al. 2015
tHansell et al. 2016
tHansell et al. 2016
tHansell et al. 2016
+Cao et al. 2011
tCarey et al. 2013
+Dong et al. 2012
tKatanoda et al. 2011
tHartetal. 2011
tKatanoda et al. 2011
tKatanoda et al. 2011
tHartetal. 2011
Krewski et al. 2000
tKrewski et al. 2009
Abbey et al. 1999
Abbey et al. 1999
Nafstad et al. 2004
Beelen et al. 2008
Filleul et al. 2005
tCarey et al. 2013
tHansell et al. 2016
tHansell et al. 2016
tHansell et al. 2016
Elliott et al. 2007
tCao et al. 2011
tKatanoda et al. 2011
Krewski et al. 2000
tKrewski et al. 2009
Abbey et al. 1999
Abbey et al. 1999
Filleul et al. 2005
Elliott et al. 2007
tWang et al. 2009
tHartetal. 2011
Beelen et al. 2008
Elliott et al. 2007
tBentayeb et al. 2015
tHansell et al. 2016
tHansell et al. 2016
tHansell et al. 2016
tCao et al. 2011
tZhang et al. 2011
tHartetal. 2011
tKrewski et al. 2009
Nafstad et al. 2004
tCarey et al. 2013
Nafstad et al. 2004
tZhang et al. 2011
Beelen et al. 2008
Elliott et al. 2007
tKrewski et al. 2009
Location
USA
Norway
The Netherlands
Great Britain
France
England
England
England
Eastern China
England
China
Japan
USA
Japan
Japan
USA
USA
USA
USA
USA
Norway
The Netherlands
France
England
England
England
England
Great Britain
Eastern China
Japan
USA
USA
USA
USA
France
Great Britain
Brisbane, Australia
USA
The Netherlands
Great Britain
France
England
England
England
Eastern China
Shenyang, China
USA
USA
Norway
England
Norway
Shenyang, China
The Netherlands
Great Britain
USA
Mean
Notes
(PPb)
4.8
Men
3.6
Men
5.2
12.2-41.4
2.3
32.4
1971
16.4
1981
11.2
1991
27.7
1.5
23.9
2.4-19.0
4.8
COPD
2.4-19.0
2.4-19.0
4.8
9.6
5.6
5.6
3.6
5.2
I.5
32.4
16.4
II.2
12.2-41.4
27.7
2.4-19.0
COPD
Pneumonia
Men
HSC
ACS
Men
Women
Men
1971
1981
1991
9.6
5.6
5.6
12.2-41.4
5.4
HSC
ACS
Men
Women
4.8
5.2
12.2
2.3
32.4
16.4
11.2
27.7
23.9
4.8
9.6
3.6
1.5
3.6
23.9
Men
41.4
1971
1981
1991
IHD - Men
IHD
IHD -Men
Circulatory
Cerebrovascular-Men —
Cerebrovascular *
5.2
12.2-41.4
9.6
Respiratory
Lung Cancer
2.52 (1.34,4.77)
4.40 (2.34,8.33)
Cardiopulmonary
Cardiovascular
Other
0.6
1.2
1.4
1.6
Relative Risk (95% Confidence Interval)
ACS = American Cancer Society Study; COPD = chronic obstructive pulmonary disease; HSC = Harvard Six Cities Study;
IHD = ischemic heart disease.
Note: Studies in red are recent studies. Studies in black were included in the 2008 ISA for Sulfur Oxides. Relative risks are
standardized to a 5-ppb increase in sulfur dioxide concentrations. Corresponding quantitative results are reported in Supplemental
Table 5S-32 (U.S. EPA. 2017c).
Figure 5-27 Relative risks (95% confidence interval) of sulfur
dioxide-associated cause-specific mortality.
5-313
-------
Table 5-43 Summary of evidence, which is inadequate to infer a causal
relationship between long-term sulfur dioxide exposure and total
mortality.
Rationale for
Causal
Determination3
Key Evidence13
Key References'3
SO2 Concentrations
Associated with Effects0
Some Small, positive associations
epidemiologic between long-term exposure to
studies report SO2 and mortality in the HSC
positive cohort, the ACS cohort, and the
associations Veterans cohort, even after
but results are adjustment for common
not entirely potential confounders
consistent
Krewski et al. (2000)
tKrewski et al. (2009)
Jerrett et al. (2003)
Krewski et al. (2000)
tLipfert et al. (2009)
Mean: 1.6-24.0 ppb
City-specific annual mean:
9.3-9.6 ppb
County-level mean from air
quality model: 4.3 ppb
Recent cohort studies in the
U.S. observe increases in total
mortality and mortality due to
lung cancer and cardiovascular
and respiratory disease, but
exposure assessment and
statistical methods were not
adequate for study of SO2
tHart et al. (2011)
Annual average at residential
address from model: 4.8 ppb
Some
epidemiologic
studies report
no associations
No association observed in
European cohort studies for
total, respiratory, or
cardiovascular mortality
Beelen et al. (2008b)
IDWto regional monitors:
5.2 ppb
Nafstad et al. (2004)
Model/monitor hybrid: 3.6 ppb
Filleul et al. (2005)
3-yr mean: 3.0-8.2 ppb
Uncertainty due
to potential
confounding
from correlated
pollutants
When reported, correlations
with copollutants were
generally moderate (0.4-0.7) to
high (>0.7). Confounding of
observed associations by other
pollutants or pollutant mixtures
cannot be ruled out
Table 5-42
Uncertainty
regarding
exposure
measurement
error
SO2 has low (<0.4) to moderate Section 3.4.2
(0.4-0.7) spatial correlations
across urban geographical
scales. The geographical scale
for estimating exposure used in
these studies may be too large
for a highly spatially
heterogeneous pollutant such
as SO2.
Exposure measurement error in Section 3.4.4.2
long-term SO2 exposure can
lead to bias toward or away
from the null
5-314
-------
Table 5-43 (Continued): Summary of evidence, which is inadequate to infer a
causal relationship between long term sulfur dioxide
exposure and total mortality.
Rationale for
Causal
Determination3
Key Evidence13
Key References'3
SO2 Concentrations
Associated with Effects0
No coherence
with evidence
for respiratory
and
cardiovascular
morbidity
No evidence for long-term
exposure and respiratory health
effects in adults to support the
observed associations with
respiratory mortality
Section 5.2.2.6
No evidence for long-term
exposure and cardiovascular
health effects in adults to
support the observed
associations with
cardiovascular mortality
Section 5.3.2.6
ACS = American Cancer Society; HSC = Harvard Six Cities; IDW = inverse distance weighting; S02 = sulfur dioxide.
aBased on aspects considered in judgments of causality and weight of evidence in causal framework in Tables I and II of the
Preamble to the ISAs (U.S. EPA. 2015b).
bDescribes the key evidence and references contributing most heavily to causal determination and, where applicable, to
uncertainties or inconsistencies. References to earlier sections indicate where full body of evidence is described.
°Describes the S02 concentrations with which the evidence is substantiated (for experimental studies, below 5,000 ppb).
fStudies published since the 2008 ISA for Sulfur Oxides.
5.6 Cancer
5.6.1 Introduction
The 2008 SOx ISA summarized the literature on SO2 concentrations and lung cancer as
"inadequate to infer a causal relationship" (U.S. EPA. 2008d'). Multiple studies across the
U.S. and Europe investigated the relationship of SO2 concentrations with lung cancer
incidence and mortality. Many studies reported generally null associations, including
some that were limited by a small number of cancer cases, but some studies demonstrated
positive associations. The body of literature characterizing the carcinogenic, genotoxic,
and mutagenic effects of exposure to SO2 has grown since the 2008 SOx ISA. The animal
toxicology literature of SO2 exposure is dominated by studies of SO2 acting as a
cocarcinogen or tumor promoter, with one study of SO2 inhalation associated with an
increased rate of lung tumor formation in lung tumor-susceptible female rodents.
Genotoxicity and mutagenicity studies show mixed results with null studies in a
Drosophila model and positive micronuclei (MN) findings in a mouse inhalation model
of SO2 exposure. These recent studies have not informed the uncertainties identified in
5-315
-------
the previous review, including uncertainties due to exposure measurement error, potential
copollutant confounding, and limited mechanistic evidence or biological plausibility.
The cancer section of the ISA characterizes epidemiologic associations of SO2 exposure
with cancer incidence or cancer mortality, as well as the animal toxicology
carcinogenicity studies (Section 5.6.2). Subsections discuss the evidence relating to lung
cancer (Section 5.6.2.1). bladder cancer (Section 5.6.2.2). and other cancers
(Section 5.6.2.3V Laboratory studies of mutagenicity or genotoxicity are discussed in
Section 5.6.3. Supplemental Tables provide detailed summaries of the respective new
epidemiologic [Table 5S-33 (U.S. EPA. 2017c)! and genotoxic/mutagenic [Table 5S-34
(U.S. EPA. 2017c) literature.
5.6.2 Cancer Incidence and Mortality
5.6.2.1 Lung Cancer Incidence and Mortality
International studies exploring the associations between SO2 concentrations and lung
cancer incidence have provided inconsistent results. No recent studies on SO2
concentration and lung cancer incidence in the U.S. have been published. Large studies
conducted using the Netherlands Cohort Study on Diet and Cancer examined the
association between SO2 concentration and lung cancer incidence (Brunekreef et al..
2009; Beelen et al.. 2008a). Null associations were reported in both analyses of the full
cohort and a case-cohort design. None of the analyses adjusted for copollutants. An
ecological study in Israel examining lung cancer incidence among men also reported null
results for the association with SO2 concentrations (Eitan et al.. 2010). Results were
relatively unchanged when adjusting for PM10. No association was observed between SO2
concentrations and lung cancer hospitalizations among men or women in southern France
in an ecological study that did not control for copollutants (Pascal et al.. 2013). However,
an ecological analysis performed among women in Taiwan demonstrated a positive
association between SO2 concentration and lung cancer incidence (Tseng et al.. 2012).
This positive association remained in a regression model adjusted for other pollutants
(CO, NO2, NO, O3, and PM10; none of these air pollutants exhibited an association with
lung cancer incidence). The association was present in analyses for both types of lung
cancer examined, adenocarcinomas and squamous cell carcinomas. Thus, overall,
multiple ecologic studies have been performed examining SO2 concentrations and lung
cancer incidence with inconsistent findings, and analyses using a large cohort study
reported no association between SO2 concentrations and lung cancer incidence but had no
control of copollutant confounders. Each of these studies used SO2 concentrations
5-316
-------
measured at fixed-site monitors to assign exposure. Beelen et al. (2008a) and Brunekreef
et al. (2009) used inverse distance weighting between the fixed-site monitor location and
residential address, and combined this with the output of land use regression (LUR)
models for urban contributions. Eitan et al. (2010) generated spatially interpolated
surfaces for a 7-year period, while the other ecological studies relied on annual averages
from the fixed-site monitors. None of the studies corrected for exposure measurement
error.
Studies in the U.S. have reported inconsistent findings for the association between SO2
concentrations and lung cancer mortality (see Section 5.5.2 and Figure 5-27). No
association between SO2 concentrations and lung cancer mortality was present in a report
by the Health Effects Institute (Krewski et al.. 2009). Estimates stratified by high school
education (less than high school education, high school education, or greater) were also
examined, and no association was present in either subgroup. In addition to the entire
time period of the study, the researchers also examined 5-year increments, none of which
demonstrated an association. However, a recent study of men in the trucking industry
found a slight positive association between SO2 concentrations and lung cancer mortality
(Hartetal.. 2011). With the inclusion of PM10 and NO2 in the model, the 95% CI
included the null but the point estimate was in the positive direction and only slightly
attenuated.
Recent studies have also been performed in Asia and Europe examining the relationship
between SO2 and lung cancer mortality. In China, a positive association was observed
between SO2 and lung cancer mortality (Chen et al.. 2016; Cao etal.. 2011). In the study
by Cao et al. (2011). this association was relatively unchanged with adjustment for either
TSP or NOx. A study in Japan also reported a positive association between SO2 and lung
cancer mortality (Katanoda et al.. 2011). However, the estimate was reduced when
additional potential confounders (smoking of parents during subjects' childhood,
consumption of nonyellow or nongreen vegetables, occupation, and health insurance)
were controlled for and no copollutant assessment was performed. Positive associations
were also observed for suspended PM, PM2 5, and NO2 concentrations. When examining
subgroups, the association was highest among male smokers. The point estimate was
similar to the overall estimate for male former smokers, but the 95% confidence interval
was wide due to the small size of the study population. The estimate was lowest among
female never smokers. The number of male never smokers and female smokers were too
small to assess individually. A study in the U.K. also demonstrated a positive association
between SO2 concentration and lung cancer mortality (Carey et al.. 2013). The
association was slightly attenuated when education was included in the model instead of
income. However, a large study using the Netherlands Cohort Study on Diet and Cancer
reported no association between SO2 concentration and lung cancer mortality (Brunekreef
5-317
-------
ct al.. 2009). This study was mentioned above and also did not demonstrate an
association between SO2 concentration and lung cancer incidence. No copollutant models
were examined. In summary, consistent with studies conducted in the U.S. examining
SO2 concentrations and cancer mortality, recent studies performed in Asia and Europe
also had inconsistent findings. Many of these studies used SO2 concentrations measured
at fixed-site monitors to assign exposure, and none of the studies corrected for exposure
measurement error. Brunekreef et al. (2009) used inverse distance weighting between the
fixed-site monitor location and residential address, and combined this with the output of
LUR models for urban contributions. Hartetal. (2011) used spatial smoothing, and Carey
et al. (2013) used a dispersion model constructed with emissions data to assign exposure.
A study in Italy used a Lagrangian dispersion model to estimate SOx concentrations as a
marker for refinery plant emissions [exposure assessment technique summarized in
Section 3.3.2.4 (Ancona et al.. 2015)1. The relationship between these estimates and
cancer mortality and hospitalizations were investigated. No association was observed for
lung cancer among men or women; however, these results are difficult to interpret. The
estimated SOx concentrations were highly correlated with estimates of PM10, which is
expected as SOx was being treated as a marker for petrochemical refinery emissions. This
makes interpretation difficult as copollutant models were not shown for lung cancer and
additionally the validity of the model is unknown.
A recent meta-analysis (Chen et al.. 2015a) combined the results of five studies of SO2
and lung cancer and found an overall OR of 1.03 (95% CI: 1.02, 1.05), although one of
the five studies lYCao etal.. 2011); characterized above] accounted for nearly 80% of the
weight contributing to the overall OR and was the only study of the five to observe a
positive and statistically significant association between SO2 exposure and lung cancer.
Three of the remaining studies included in the meta-analysis observed null associations
between SO2 and lung cancer.
Sulfur Dioxide Lung Carcinogenesis, Cocarcinogenic Potential, and Tumor
Promotion in Laboratory Animal Models
The toxicological evidence for effects of sulfur dioxide in carcinogenicity, mutagenicity,
or genotoxicity is characterized below.
Direct evidence of carcinogenicity was studied evaluating incidence of lung tumors in a
lung adenoma-susceptible mouse strain, (the LX mouse), with chronic exposure to sulfur
dioxide at 500 ppm, 5 minutes/day, 5 days/week for 2 years (Peacock and Spence. 1967).
S02-exposed female mice had an increase in the number of lung tumors subgrouped as
(1) adenomas and (2) primary carcinomas versus controls. Males had a smaller increase
5-318
-------
in adenomas versus controls and similar levels of primary carcinomas compared to
controls.
Evidence exists for SO2 to be a cocarcinogen (Pauluhn et al.. 1985): SO2 and
benzo[a]pyrene (B[a]P), coexposure increased the incidence of lung tumor formation in
rodents versus B[a]P exposure alone. Chronic coexposure to SO2 and B[a]P resulted in
increased incidence of upper respiratory tract neoplasia in rats (Laskin et al.. 1976) and
hamsters (Pauluhn et al.. 1985) over B[a]P exposure alone. SO2 exposure shortened the
induction period for spontaneous squamous cell lung tumor formation after B[a]P
exposure (Laskin et al.. 1976); rats were exposed 5 days a week, 6 hours/day for their
lifetime to 10 ppm SO2 alone via inhalation or 4 ppm SO2 + 10 mg/m3 B[a]P (1 hour
B[a]P/day). SO2 exposure also shortened the induction time for
methylcholanthrene-induced carcinogenesis.
Multiple studies explored SO2 as a cocarcinogen or promoter after particulate-induced
tumorigenesis. In a study of SPM—induced tumorigenesis (proliferative lesions of
pulmonary endocrine cells) in the rat, SO2 did not exacerbate SPM-dependent hyperplasia
when rats were exposed to the mixture of SPM and SO2 (Ito et al.. 1997). Adult male rats
were exposed to SO2 for 11 months, 16 hours/day ± SPM for 4 weeks, once/week by
intratracheal injection. SO2 did not act as a tumor promoter or cocarcinogen in this model.
In a separate study of diesel exhaust particle- (DEP-) dependent lung tumorigenesis, SO2
was able to promote DEP-dependent tumorigenesis (Ohvama et al.. 1999). Adult male
rats were intratracheally instilled with diesel exhaust particle extract-coated carbon black
particles (DEcCBP) and exposed to 4 ppm SO2 for 10 months. Eighteen months after
starting the experiment, the animals were examined for respiratory tract tumors and DNA
adducts were measured in lung tissue. Lung tumors and DNA adducts were seen in
animals with coexposure to SO2 and DEcCBP but not in animals only exposed to
DEcCBP. SO2 acted as a tumor promoter in animals exposed to DEcCBP. In a separate
investigation, hamsters were exposed to diesel engine exhaust (separately with and
without particles) and a mixture of SO2 and NO2 with or without exposure to the
carcinogen diethyl-nitrosamine to investigate the potential cocarcinogenic effect of
exposure to the dioxides mixture and diesel engine exhaust in the respiratory tract
(Heinrich et al.. 1989). These adult male hamsters were exposed for 19 hours/day,
5 days/week for 6, 10.5, 15, or 18 months to diesel exhaust, filtered diesel exhaust
(without particles), a dioxide mixture of NO2 (5 ppm) and SO2 (10 ppm), or clean air.
Two exposure groups from each of the aforementioned test groups were also given a
single subcutaneous injection of diethylnitrosamine (DEN) (3 mg or 6 mg/kg body
weight). Exposure to the dioxide mixture by itself did not elevate tumor rate (tumor
induction), nor did it exacerbate DEN-dependent effects (tumor promotion) in the
hamster. In summary, a comparison of multiple studies of SO2 coexposure with particles
5-319
-------
reported mixed results in various models of carcinogenicity, cocarcinogenic potential, or
tumor promotion.
Oncogene and tumor suppressor genes also appear to be affected by SO2 exposure,
especially with coexposure to B[a]P. Synergistic expression of c-fos and c-jun with SO2
and B[a]P coexposure was observed in rodent lungs (Oin and Meng. 2006). SO2 and
B[a]P coexposure in male Wistar rats (26.5 ppm SO2 inhalation, 6 hours/day for 7 days;
3 mg B[a]P instilled) statistically significantly downregulated expression of tumor
suppressor genes pi 6 and myc. and increased expression of oncogenes c-myc, H-ras, and
p53. Others have reported that SO2 exposure alone could induce p53 expression in rats
(Bai and Meng. 2005).
5.6.2.2 Bladder Cancer Incidence and Mortality
Several studies on the relationship between SO2 concentrations and bladder cancer
incidence and mortality have been published since the 2008 SOx ISA (U.S. EPA. 2008d).
Positive associations were observed in studies of bladder cancer mortality but not bladder
cancer incidence. An ecological study in southern France reported on the relationship
between SO2 concentrations and hospitalizations for bladder cancer without examination
of copollutant models (Pascal et al.. 2013). Null associations were observed for men and
women. Another ecological study in Israel examining bladder cancer incidence also
reported sex-stratified results (Eitan et al.. 2010). Neither sex demonstrated an association
between SO2 concentrations and bladder cancer in models with and without adjustment
for PM10. However, an association was observed in a study examining the relationship
between SO2 and bladder cancer mortality (Liu et al.. 2009a'). Liu et al. ^OOga)
investigated the association between SO2 and bladder cancer mortality using controls with
mortality due to causes unrelated to neoplasm or genitourinary-related disease and
matched by sex, year of birth, and year of death. A positive association was observed
between SO2 concentration in the second and third tertiles of exposure and bladder cancer
mortality. For further investigations, the authors created a three-level exposure variable
combining NO2 and SO2 concentrations: the lowest tertile of SO2 and NO2 concentrations
(<4.32 ppb and <20.99 ppb, respectively), the highest tertile of SO2 and NO2
concentrations (>6.49 ppb and >27.33 ppb, respectively), and other
categorizations/combinations. The ORs were 1.98 (95% CI 1.36, 2.88) for the highest
level ofN02 and SO2 and 1.37 (95% CI 1.03, 1.82) for the middle level categorizations.
Although the point estimates are higher than those observed for SO2 alone [see
Supplemental Table 5S-33, (U.S. EPA. 2017c)!. the 95% confidence intervals overlap,
and therefore, conclusions that NO2 and SO2 combined contribute to higher odds of
mortality than either alone cannot be drawn. Finally, a study using SOx concentration
5-320
-------
estimated using a Lagrangian dispersion model reported no association between SOx
concentration and bladder cancer mortality or hospitalizations among men or women
(Ancona et al.. 2015). However, results of this study are difficult to interpret because of
unknown validity of the model (see Section 3.3.2.4) and high correlation with PMio and
hydrogen sulfide (H2S).
5.6.2.3 Incidence of Other Cancers
Recent studies of SO2 concentrations and other cancer types have been published since
the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d). but provide limited information on
associations with SO2. An ecological study in southern France investigated the
relationships between SO2 and hospitalizations for breast cancer, acute leukemia,
myeloma, and non-Hodgkin lymphoma (Pascal et al.. 2013). Null associations were
observed in sex-stratified analyses among men and women, with the exception of a
positive association between SO2 and acute leukemia among men. However, the authors
urge caution when interpreting the results due to a small number of male acute leukemia
cases. This study did not examine copollutant confounding. Another ecologic study used
Surveillance, Epidemiology, and End Results data to examine the correlation between
SO2 concentrations and breast cancer incidence (Wei et al.. 2012). A positive relationship
was detected, but there was no control for potential confounders of other air pollutants (of
which CO, NOx, and VOCs, but not PM10, also demonstrated a positive correlation with
breast cancer incidence). Both of these studies are limited by their ecologic nature and the
lack of individual-level data. A cross-sectional study was conducted in South Korea that
looked at the association between symptom scores for prostate cancer and emissions data
for SOx (measured in kg/year/person) and a number of other air pollutants (Shim et al..
2015). In logistic regression models adjusted for age, the authors observed positive
associations between men living in areas with greater emissions of SOx and symptom
scores for prostate cancer. Similar results were observed for NOx, CO, PM10, VOCs, and
NH3. The lack of control for potential confounding by other air pollutants or risk factors
(e.g., smoking, SES) limit the interpretation of these results.
A cohort study examined the relationship between SOx concentrations, estimated using a
Lagrangian dispersion model, and hospitalizations and mortality for various cancer types
(Ancona etal.. 2015). No associations were found between SOx concentrations and either
hospitalizations or mortality due to cancers of the stomach, colon/rectum, liver, kidney,
brain, or breast. Positive associations were observed for SOx concentration and mortality
due to pancreatic and larynx cancers among women but not men. The 95% confidence
interval showed a large degree of imprecision in the estimates for cancer of the larynx.
The association with pancreatic cancer was not robust to adjustment with H2S or PM10.
5-321
-------
When examining the association between estimated SOx concentration and
hospitalizations, a positive, but imprecise, association was observed for cancer of the
larynx among women and an inverse association was noted for cancers of lymphatic and
hematopoietic tissue.
5.6.2.4 Summary of Cancer Incidence and Mortality
Similar to studies of SO2 concentrations and lung cancer in the previous ISA (U.S. EPA.
2008d), recent studies of SO2 concentrations and lung cancer have provided inconsistent
results (Carey et al.. 2013; Pascal et al.. 2013; Tseng et al.. 2012; Cao etal.. 2011; Hart et
al.. 2011; Katanoda et al.. 2011; Eitan et al.. 2010; Brunekreef et al.. 2009; Beelen et al..
2008a'). with two of the largest studies reporting null results (Brunekreef et al.. 2009;
Krewski et al.. 2009). Studies of bladder cancer appear to find no association between
SO2 concentrations and bladder cancer incidence (Pascal et al.. 2013; Eitan et al.. 2010).
but a study of SO2 concentration and bladder cancer mortality reported a positive
association (Liu et al.. 2009a'). Limited information is available regarding other cancers.
Animal toxicology models of SO2 inhalation exposure show SO2 acting as a promoter or
cocarcinogen, with one study showing increased lung tumor formation in a lung
tumor-prone animal model.
5.6.3 Genotoxicity and Mutagenicity
Multiple studies of genotoxicity or mutagenesis with SO2 in vivo or SO2 in vitro exposure
have been reported in the literature and are detailed below in Supplemental Table 5S-34
(U.S. EPA. 2017c).
After inhalation exposure to SO2, mouse bone marrow micronuclei formation was
significantly elevated in both males and females after exposure to SO2 (5.4, 10.7, 21.4, or
32.1 ppm SO2, 4 hours/day for 7 days) (Meng et al.. 2002). The polychromatophilic
erythroblasts of the bone marrow (MNPCE) were formed in significantly increased
numbers with SO2 exposure. Another study recapitulated these findings; subacute
exposure to SO2 (10.7 ppm SO2, 6 hours/day for 5 days) induced a significant increase in
MNPCE with this effect attenuated by exogenous antioxidant SSO pretreatment (Ruan et
al.. 2003).
The rate of DNA single strand breaks induced by B[a]P exposure in fetal hamster lung
cells (50 ppm for 2 weeks) (Pool et al.. 1988b) and rat liver cells (2.5, 5, 9.9, or 19.9 ppm,
4 hours/day for 7 days) (Pool et al.. 1988a) was significantly attenuated by concomitant
exposure to SO2 (50 ppm for 2 weeks).
5-322
-------
Genotoxicity testing of Drosophila sperm for sex-linked recessive lethals after feeding
larvae 0.04 M or 0.08 M sodium sulfite in a 1% glucose solution was performed, and no
increase was found above background. One caveat is that sulfite can interact with
glucose, making the exposure assessment more complicated.
Multiple studies of genotoxicity or mutagenesis with SO2 in vivo or in vitro exposure
have been reported in the literature and are summarized in Supplemental Table 5S-34
(U.S. EPA. 2017c). Mixed results of genotoxicity or mutagenicity have been reported
after SO2 exposure including positive associations with SO2 inhalation exposure in the
mouse MN assay.
5.6.4 Summary and Causal Determination
The overall evidence for long-term SO2 exposure and cancer is inadequate to infer a
causal relationship. This conclusion is based on the inconsistent evidence from
epidemiologic studies, as well as mixed evidence within the animal toxicology and mode
of action framework for mutagenesis and genotoxicity. In past reviews, a limited number
of epidemiologic studies had assessed the relationship between long-term SO2
concentrations and cancer incidence and mortality. The 2008 ISA for Sulfur Oxides
concluded that the evidence was "inconclusive" (U.S. EPA. 2008d'). Recent studies
include evidence on lung cancer as well as new types of cancer, evaluating both
incidence and mortality. However, the additional recent evidence has not informed any of
the uncertainties identified in the previous review, including uncertainties due to
exposure measurement error, potential copollutant confounding, and limited mechanistic
evidence or biological plausibility. All available evidence for cancer due to long-term
SO2 concentrations was evaluated using the framework described in Table II of the
Preamble to the IS As (U.S. EPA. 2015b). The key evidence as it relates to the causal
framework is summarized in Table 5-44.
5-323
-------
Table 5-44 Summary of evidence, which is inadequate to infer a causal
relationship between long-term sulfur dioxide exposure and cancer.
Rationale for Causal SO2 Concentrations
Determination3 Key Evidence13 Key References'3 Associated with Effects0
Among a small body
Generally null associations
Section 5.6.2
Means varied across studies
of evidence, evidence
from studies of cancer
including areas estimating mean
from epidemiologic
incidence, with some
concentrations of SO2 as low as
studies is inconsistent
observed increases in lung
cancer and bladder cancer
mortality in studies conducted
in the U.S., Europe, and Asia
1.49 ppb to as high as 27.87 ppb.
Associations observed with
bladder cancer mortality at levels
as low as 4.39-6.09 ppb.
Uncertainty due to
Fixed-site monitors used in
Section 3.4.2.2
exposure
cancer studies may not
measurement error
capture spatial variability of
SO2 concentrations.
Exposure measure Section 3.4.4.2
measurement error in
long-term SO2 exposure
assessment can bias toward
or away from the null
Uncertainty due to Correlations of SO2 with other Section 3.4.3
confounding by pollutants vary by study or
correlated copollutants are not examined. Some
pollutants are moderately to
highly correlated with SO2 but
are not always taken into
account as potential
confounders
Uncertainty due to Studies in a Peacock and Spence 500,000 ppb
limited coherence with tumor-susceptible mouse (1967)
toxicological evidence model, females had
increased numbers of lung
adenomas and carcinomas.
Studies of facilitation of
metastasis and coexposures
with known carcinogens show
mixed SO2 related effects.
Laskin et al. (1976)
10,000 ppb
Pauluhn et al. (1985)
172,000 ppb
Ohvama et al. (1999)
4,000 ppb
Heinrich et al. (1989)
5,000 or 10,000 ppb
Ito et al. (1997)
4,000 ppb
Section 5.6.2.1
5-324
-------
Table 5-44 (Continued): Summary of evidence, which is inadequate to infer a
causal relationship between long term sulfur dioxide
exposure and cancer.
Rationale for Causal
Determination3
Key Evidence13
Key References'3
SO2 Concentrations
Associated with Effects0
Some evidence
identifies key events
within the MOA from
mutagenesis and
genotoxicity
Mixed evidence of
mutagenicity and genotoxicity
formation in animal cells
exposed to SO2
Mena et al. (2002).
Ruan et al. (2003).
Pool et al. (1988b)
Section 5.6.3
5,000, 10,700, 21,400,
32,100 ppb
MOA = mode of action; S02 = sulfur dioxide.
aBased on aspects considered in judgments of causality and weight of evidence in causal framework in Tables I and II of the
Preamble to the ISAs (U.S. EPA. 2015b).
bDescribes the key evidence and references contributing most heavily to causal determination and, where applicable, to
uncertainties or inconsistencies. References to earlier sections indicate where full body of evidence is described.
°Describes the S02 concentrations with which the evidence is substantiated (for experimental studies, below 5,000 ppb).
Overall, there is inconsistent evidence for an association between long-term SO2
exposure and cancer from epidemiologic and toxicological studies. Some of the
epidemiologic studies observed positive associations while others did not. Some of these
studies with positive associations were relatively unchanged with the inclusion of various
cofounders and copollutants, although many did not evaluate the potential for copollutant
confounding. Cohort studies have reported null associations between SO2 concentrations
and lung cancer incidence. Similarly, some ecological studies also reported no
associations; although, an ecological study in Taiwan among women did report an
association between SO2 concentrations and lung cancer incidence that was relatively
unchanged when including other pollutants. Positive associations were also observed in a
study of SO2 concentrations and bladder cancer mortality but not in ecological studies of
bladder cancer incidence. The study of bladder cancer mortality examined the
relationship between bladder cancer mortality and joint exposure to high levels ofNCh
and SO2, but no copollutant assessment was performed controlling for NO2 or other air
pollutants. None of the epidemiologic studies made corrections or adjustments for
exposure measurement error, or accounted for the potential for bias away from the null,
the potential for which has been demonstrated in simulation studies (see Section 3.4.4.2).
Animal toxicological studies employing SO2 exposure with other known carcinogens
provide some evidence, showing that inhaled SO2 can increase tumor load in laboratory
rodents. Toxicological data provided by a study in LX mice, lung adenoma susceptible
animals, showed evidence of the direct carcinogenic potential of SO2. Other studies in
animal models show SO2 as a cocarcinogen with B[a]P or as a tumor promoter with
particulate-induced tumorigenesis. Nonetheless, toxicological data provide no clear
evidence of SO2 acting as a complete carcinogen and not all epidemiologic studies report
positive associations.
5-325
-------
Collectively, the inconsistent evidence from several toxicological and epidemiologic
studies is inadequate to infer a causal relationship between long-term exposure to
SO2 and cancer incidence and mortality.
5-326
-------
Annex for Chapter 5: Evaluation of Studies on Health Effects of
Sulfur Oxides
Table A-1 Scientific considerations for evaluating the strength of inference
from studies on the health effects of sulfur oxides.
Evaluation Factors
Study Design
Controlled Human Exposure:
Studies should clearly describe the primary and any secondary objectives of the study, or specific hypotheses being
tested. Study subjects should be randomly exposed without knowledge of the exposure condition. Preference is
given to balanced crossover (repeated measures) or parallel design studies that include control exposures (e.g., to
clean filtered air). In crossover studies, a sufficient and specified time between exposure days should be employed
to avoid carry over effects from prior exposure days. In parallel design studies, all arms should be matched for
individual characteristics such as age, sex, race, anthropometric properties, and health status. In studies evaluating
effects of disease, appropriately matched healthy controls are desired for interpretative purposes.
Animal Toxicology:
Studies should clearly describe the primary and any secondary objectives of the study, or specific hypotheses being
tested. Studies should include appropriately matched control exposures (e.g., to clean filtered air, time matched).
Studies should use methods to limit differences in baseline characteristics of control and exposure groups. Studies
should randomize assignment to exposure groups and where possible conceal allocation from research personnel.
Groups should be subjected to identical experimental procedures and conditions; animal care including housing,
husbandry, etc. should be identical between groups. Blinding of research personnel to study group may not be
possible due to animal welfare and experimental considerations; however, differences in the monitoring or handling
of animals in all groups by research personnel should be minimized.
Epidemiology:
Inference is stronger for studies that clearly describe the primary and any secondary aims of the study, or specific
hypotheses being tested.
For short-term exposure, time-series, case crossover, and panel studies are emphasized over cross-sectional
studies because they examine temporal correlations and are less prone to confounding by factors that differ
between individuals (e.g., SES, age). Studies with large sample sizes and conducted over multiple years are
considered to produce more reliable results. If other quality parameters are equal, multicity studies carry more
weight than single-city studies because they tend to have larger sample sizes and lower potential for publication
bias.
For long-term exposure, inference is considered to be stronger for prospective cohort studies and case-control
studies nested within a cohort (e.g., for rare diseases) than cross-sectional, other case-control, or ecologic studies.
Cohort studies can better inform the temporality of exposure and effect. Other designs can have uncertainty related
to the appropriateness of the control group or validity of inference about individuals from group-level data. Study
design limitations can bias health effect associations in either direction.
5-327
-------
Study Population/Test Model
Controlled Human Exposure:
In general, the subjects recruited into study groups should be similarly matched for age, sex, race, anthropometric
properties, and health status. In studies evaluating effects of specific subject characteristics (e.g., disease, genetic
polymorphism, etc.), appropriately matched healthy controls are preferred. Relevant characteristics and health
status should be reported for each experimental group. Criteria for including and excluding subjects should be
clearly indicated. For the examination of populations with an underlying health condition (e.g., asthma),
independent, clinical assessment of the health condition is ideal, but self reporting of physician diagnosis generally
is considered to be reliable for respiratory and cardiovascular disease outcomes.3 The loss or withdrawal of
recruited subjects during the course of a study should be reported. Specific rationale for excluding subject(s) from
any portion of a protocol should be explained.
Animal Toxicology:
Ideally, studies should report species, strain, substrain, genetic background, age, sex, and weight. Unless data
indicate otherwise, all animal species and strains are considered appropriate for evaluating effects of SO2 exposure.
It is preferred that the authors test for effects in both sexes and multiple lifestages, and report the result for each
group separately. All animals used in a study should be accounted for, and rationale for exclusion of animals or data
should be specified.
Epidemiology:
Confidence in results is greater in studies that recruit the study population from the target population and examine a
study population that is representative of the target population. Studies with high participation and low drop-out over
time that is not dependent on exposure or health status are considered to have low potential for selection bias.
Clear indication of criteria for including and excluding subjects can facilitate assessment of selection bias. For
populations with an underlying health condition, independent, clinical assessment of the health condition is valuable,
but self report of physician diagnosis generally is considered to be reliable for respiratory and cardiovascular
outcomes. Comparisons of groups with and without an underlying health condition are more informative if groups
are from the same source population. Selection bias can influence results in either direction or may not affect the
validity of results but rather reduce the generalizability of findings to the target population.
Pollutant
Controlled Human Exposure:
The focus is on studies testing SO2 exposure.
Animal Toxicology:
The focus is on studies testing SO2 exposure.
Epidemiology:
The focus is on studies testing SO2 exposure.
Exposure Assessment or Assignment
Controlled Human Exposure:
For this assessment, the focus will be on studies that use SO2 concentrations less than or equal to 2 ppm
(Section 1.2). Studies that use higher exposure concentrations may provide information relevant to mode of action,
dosimetry, inter-species variation, or at-risk human populations. Controlled human exposure studies considering
short-term, (e.g., generally exposures from 5-10 min, to 0.2-0.6 ppm SO2, were emphasized) (Section 1.2).
5-328
-------
Animal Toxicology:
For this assessment, the focus will be on studies that use SO2 concentrations less than or equal to 2,000 ppb
(Section 1.2). Studies that use higher exposure concentrations may provide information relevant to mode of action,
dosimetry, inter-species variation, or at-risk human populations. Studies should characterize pollutant concentration,
temperature, and relative humidity and/or have measures in place to adequately control the exposure conditions.
The focus is on inhalation exposure. Noninhalation exposure experiments may provide information relevant to mode
of action. In vitro studies may be included if they provide mechanistic insight or examine similar effects as in vivo,
but are generally not included. All studies should include exposure control groups (e.g., clean filtered air).
Epidemiology:
Of primary relevance are relationships of health effects with the ambient component of exposure to SO2. However,
information about ambient exposure rarely is available for individual subjects; most often, inference is based on
ambient concentrations. Studies that compare exposure assessment methods are considered to be particularly
informative. Inference is stronger when the duration or lag of the exposure metric corresponds with the time course
for physiological changes in the outcome (e.g., up to a few days for symptoms) or latency of disease (e.g., several
years for cancer).
Given the spatial heterogeneity in ambient SO2 and potentially variable relationships between personal exposures
and ambient concentrations (Sections 3.4.2.2 and 3.4.1). validated methods that capture the extent of variability for
the particular study design (temporal vs. spatial contrasts) and location carry greater weight. Fixed-site
measurements, whether averaged across multiple monitors or assigned from the nearest or single available
monitor, have well-recognized limitations in capturing spatial variation in air pollutants. Monitors impacted by large
SO2 sources are particularly subject to concentration fluctuations due to changes in emission rates and
meteorological conditions and may not fully represent population exposure. Results based on fixed-site
measurements can be informative if correlated with personal exposures, closely located to study subjects, highly
correlated across monitors within a location, used in locations with well-distributed sources, or combined with
time-activity information.
In studies of short-term exposure, temporal variability of the exposure metric is of primary interest. Metrics that may
capture variation in ambient sulfur oxides and strengthen inference include concentrations in subjects'
microenvironments and individual-level outdoor concentrations combined with time-activity data. Atmospheric
models may be used for exposure assessment in place of or to supplement SO2 measurements in epidemiologic
analyses. Dispersion models (e.g., AERMOD) can provide valuable information on fine-scale temporal and spatial
variations (within tens of km) of SO2 concentrations, which is particularly important for assessing exposure near
large stationary sources. Alternatively, grid-scale models (e.g., CMAQ) that represent SO2 exposure over relatively
large spatial scales (e.g., typically greater than 4 * 4-km grid size) often do not provide enough spatial resolution to
capture acute SO2 peaks that influence short-term health outcomes. Uncertainty in exposure predictions from these
models is largely influenced by model formulations and the quality of model input data pertaining to emissions or
meteorology, which tends to vary on a study-by-study basis.
For long-term exposures, models that capture within-community spatial variation in individual exposure may be
given more weight for spatially variable ambient SO2.
Exposure measurement error often attenuates health effect estimates or decreases the precision of the association
(i.e., wider 95% CIs), particularly associations based on temporal variation in short-term exposure (Section 3.4.4.1).
However, exposure measurement error can bias estimates away from the null, particularly for long-term exposures
(Section 3.4.4.2).
Outcome Assessment/Evaluation
Controlled Human Exposure:
Endpoints should be assessed in the same manner for control and exposure groups (e.g., time after exposure,
methods, endpoint evaluator) using valid, reliable methods. Blinding of endpoint evaluators is ideal, especially for
qualitative endpoints (e.g., histopathology). For each experiment and each experimental group, including controls,
precise details of all procedures carried out should be provided including how, when, and where. Time of the
endpoint evaluations is a key consideration that will vary depending on the endpoint evaluated. Endpoints should be
assessed at time points that are appropriate for the research questions.
5-329
-------
Animal Toxicology:
Endpoints should be assessed in the same manner for control and exposure groups (e.g., time after exposure,
methods, endpoint evaluator) using valid, reliable methods. Blinding of endpoint evaluators is ideal, especially for
qualitative endpoints (e.g., histopathology). For each experiment and each experimental group, including controls,
precise details of all procedures carried out should be provided including how, when, and where. Time of the
endpoint evaluations is a key consideration that will vary depending on the endpoint evaluated. Endpoints should be
assessed at time points that are appropriate for the research questions.
Epidemiology:
Inference is stronger when outcomes are assessed or reported without knowledge of exposure status. Knowledge
of exposure status could produce artifactual associations. Confidence is greater when outcomes assessed by
interview, self-reporting, clinical examination, or analysis of biological indicators are defined by consistent criteria
and collected by validated, reliable methods. Independent, clinical assessment is valuable for outcomes such as
lung function or incidence of disease, but report of physician diagnosis has shown good reliability.3 Outcomes
assessed at time intervals that correspond with the time course for physiological changes (e.g., up to a few days for
symptoms) are emphasized. When health effects of long-term exposure are assessed by acute events such as
symptoms or hospital admissions, inference is strengthened when results are adjusted for short-term exposure.
Validated questionnaires for subjective outcomes such as symptoms are regarded to be reliable,15 particularly when
collected frequently and not subject to long recall. For biological samples, the stability of the compound of interest
and the sensitivity and precision of the analytical method is considered.
If not based on knowledge of exposure status, errors in outcome assessment tend to bias results toward the null.
Potential Copollutant Confounding
Controlled Human Exposure:
Exposure should be well characterized to evaluate independent effects of SO2.
Animal Toxicology:
Exposure should be well characterized to evaluate independent effects of SO2.
Epidemiology:
Not accounting for copollutant confounding can produce artifactual associations; thus, studies that examine
copollutant confounding carry greater weight. The predominant method is copollutant modeling, which is especially
informative when measurement error is comparable for copollutants and correlations are not high. Interaction and
joint effect models are examined to a lesser extent. Evaluating correlations between SO2 and copollutants and
comparing health associations between SO2 and copollutants in single-pollutant models can add to the analysis of
potential copollutant confounding, particularly when exposure measurement error is comparable among pollutants.
Studies that examine SO2 only in single-pollutant models provide minimal information on the potential for copollutant
confounding. Copollutant confounding is evaluated based on the extent of observed correlations and relationships
with health effects. Highly variable correlations have been observed between SO2 and other criteria pollutants at
collocated monitors (Section 3.4.3). ranging from negative to strong correlations, making evaluation of copollutant
confounding necessary on a study-specific, rather than a general, basis.
Other Potential Confounding Factors
Controlled Human Exposure:
Preference is given to studies using experimental and control groups that are matched for individual level
characteristics (e.g., body weight, smoking history, age) and time-varying factors (e.g., seasonal and diurnal
patterns).
5-330
-------
Animal Toxicology:
Preference is given to studies using experimental and control groups that are matched for individual level
characteristics (e.g., body weight, litter size, food and water consumption) and time-varying factors (e.g., seasonal
and diurnal patterns).
Epidemiology:
Factors are considered to be potential confounders if demonstrated in the scientific literature to be related to health
effects and correlated with SO2. Not accounting for confounders can produce artifactual associations; thus, studies
that statistically adjust for multiple factors or control for them in the study design are emphasized. Less weight is
placed on studies that adjust for factors that mediate the relationship between SO2 and health effects, which can
bias results toward the null. In the absence of information linking health risk factors to SO2, a factor may be
evaluated as a potential effect measure modifier, but uncertainty is noted as to its role as a confounder.
Confounders vary according to study design, exposure duration, and health effect and may include, but are not
limited to, the following:
For time-series and panel studies of short-term exposure:
• Respiratory effects—meteorology, day of week, season, medication use, allergen exposure (potential
effect modifier)
• Cardiovascular effects—meteorology, day of week, season, medication use
• Total mortality—meteorology, day of week, season, long-term temporal trends
For studies of long-term exposure:
• Respiratory effects—socioeconomic status, race, age, medication use, smoking, stress
• Cardiovascular, reproductive, and development effects—socioeconomic status, race, age, medication use,
smoking, stress, noise
• Total mortality—socioeconomic status, race, age, medication use, smoking, comorbid health conditions
• Cancer—socioeconomic status, race, age, occupational exposure
Statistical Methodology
Controlled Human Exposure:
Statistical methods should be clearly described and appropriate for the study design and research question
(e.g., correction for multiple comparisons). Generally, statistical significance is used to evaluate the findings of
controlled human exposure studies. Detection of statistical significance is influenced by a variety of factors
including, but not limited to, the size of the study, exposure and outcome measurement error, and statistical model
specifications. Sample size is not a criterion for exclusion; ideally, the sample size should provide adequate power
to detect hypothesized effects (e.g., sample sizes less than three are considered less informative). Because
statistical tests have limitations, consideration is given to both trends in data and reproducibility of results.
Animal Toxicology:
Statistical methods should be clearly described and appropriate for the study design and research question
(e.g., correction for multiple comparisons). Generally, statistical significance is used to evaluate the findings of
animal toxicology studies. Detection of statistical significance is influenced by a variety of factors including, but not
limited to, the size of the study, exposure and outcome measurement error, and statistical model specifications.
Sample size is not a criterion for exclusion; ideally, the sample size should provide adequate power to detect
hypothesized effects (e.g., sample sizes less than three are considered less informative). Because statistical tests
have limitations, consideration is given to both trends in data and reproducibility of results.
5-331
-------
Epidemiology:
Multivariable regression models that include potential confounding factors are emphasized. However, multipollutant
models (more than two pollutants) are considered to produce too much uncertainty due to copollutant collinearity to
be informative. Models with interaction terms aid in the evaluation of potential confounding as well as effect
modification. Sensitivity analyses with alternate specifications for potential confounding inform the stability of
findings and aid in judgments of the strength of inference of results. In the case of multiple comparisons,
consistency in the pattern of association can increase confidence that associations were not found by chance alone.
Statistical methods that are appropriate for the power of the study carry greater weight. For example, categorical
analyses with small sample sizes can be prone to bias results toward or away from the null. Statistical tests such as
f-tests and Chi-squared tests are not considered sensitive enough for adequate inferences regarding
pollutant-health effect associations. For all methods, the effect estimate and precision of the estimate (i.e., width of
95% CI) are important considerations rather than statistical significance.
AERMOD = American Meteorological Society/U.S. EPA Regulatory Model; CI = confidence interval; CMAQ = Community
Multiscale Air Quality; SES = socioeconomic status; S02 = sulfur dioxide.
aToren et al. (1993); (Muraia et al. (2014); Weakley et al. (2013); Yang et al. (2011); Heckbert et al. (2004); Barr et al. (2002);
Muhaiarine et al. (1997)).
"Burnev et al. (1989).
5-332
-------
CHAPTER 6 POPULATIONS AND LIFESTAGES
POTENTIALLY AT INCREASED RISK
FOR HEALTH EFFECTS RELATED TO
SULFUR DIOXIDE EXPOSURE
6.1 Introduction
Interindividual variation in human responses to ambient air pollution exposure can result
in some groups or lifestages being at increased risk for health effects. The NAAQS are
intended to protect public health with an adequate margin of safety. Protection is
provided for both the population as a whole and those potentially at increased risk for
health effects in response to exposure to a criteria air pollutant (e.g., SO2) [see Preamble
to the ISAs (U.S. EPA. 2015b)!. The scientific literature has used a variety of terms to
identify factors and subsequently populations or lifestages that may be at increased risk
of an air pollutant-related health effect, including susceptible, vulnerable, sensitive, and
at risk, with recent literature introducing the term response-modifying factor (Vinikoor-
Imler et al.. 2014) [see Preamble to the ISAs (U.S. EPA. 2015bVI. Acknowledging the
inconsistency in definitions for these terms across the scientific literature and the lack of
a consensus on terminology in the scientific community, this chapter takes an inclusive
and all-encompassing approach and focuses on identifying those populations or lifestages
potentially "at risk" of an S02-related health effect.
As discussed in the Preamble to the ISAs (U.S. EPA. 2015b). risk of health effects from
exposure to SO2 may be modified as a result of intrinsic (e.g., pre-existing disease,
genetic factors) or extrinsic factors (e.g., sociodemographic or behavioral factors),
differences in internal dose (e.g., due to variability in ventilation rates or exercise
behaviors), or differences in exposure to air pollutant concentrations (e.g., more time
spent in areas with higher ambient concentrations). Some factors may lead to a reduction
in risk and are recognized during the evaluation process, but for the purposes of
identifying those populations or lifestages at greatest risk to inform decisions on the
NAAQS, the focus of this chapter is on characterizing those factors that may increase
risk. While the emphasis of this chapter is for individual factors that may increase the risk
of an SCh-related health effect, it is recognized that in many cases, portions of the
population are at increased risk of an SCh-related health effect due to a combination of
factors [e.g., residential location and socioeconomic status (SES)], but information on the
interaction among factors remains limited. Thus, the following sections identify, evaluate,
and characterize the overall confidence that individual factors potentially result in
6-1
-------
increased risk for SCh-related health effects [see Preamble to the ISAs (U.S. EPA.
2015b)].
The objective of this chapter is to identify, evaluate, and characterize the overall
confidence that various factors may increase the risk of an SCh-related health effect in a
population or lifestage, building on the conclusions drawn herein with respect to SO2
exposure and health effects. The broad categories of factors evaluated in this chapter
include pre-existing disease (Section 6.3). genetic factors (Section 6.4). and
sociodemographic and behavioral factors (Section 6.5).
6.2 Approach to Evaluating and Characterizing the Evidence
for At-Risk Factors
This chapter takes a systematic approach to identifying and evaluating factors that may
increase the risk of a population or specific lifestage to an ambient SCh-related health
effect. This chapter is complementary to the characterization of health evidence presented
in Chapter 5. and the systematic approach is described in detail in the Preamble to the
ISAs (U.S. EPA. 2015b). Briefly, in contrast to the overall evaluation of SO2 exposures
and health effects presented in Chapter 5. this chapter specifically aims to identify and
characterize the populations and lifestages at increased risk of an SCh-related health
effect. While Chapter 5 includes discussion of some populations and lifestages in order to
explicitly characterize the causal nature between SO2 exposure and health effects based
on the body of evidence (e.g., children, individuals with asthma), this chapter applies a
systematic approach to evaluating evidence that can inform the identification of such
populations and lifestages and applies a formal framework to transparently characterize
the strength of this evidence [see Preamble to the ISAs (U.S. EPA. 2015b)l. This chapter
informs the NAAQS review with regard to identification of populations at risk.
The evidence evaluated in this chapter includes relevant studies discussed in Chapter 5 of
this ISA and builds on the evidence presented in the 2008 ISA for Sulfur Oxides (U.S.
EPA. 2008d'). Based on the approach developed in previous ISAs (U.S. EPA. 2016d.
2013a. b). evidence is integrated across scientific disciplines and health effects, and
where available, with information on exposure and dosimetry (Chapter 3 and Chapter 4).
Greater emphasis is placed on those health outcomes for which a "causal" or "likely to be
causal" relationship is concluded in Chapter 5 of this ISA, while information from studies
of health outcomes for which the causal determination is "suggestive" is used as
supporting evidence where appropriate. Studies examining health outcomes for which an
"inadequate" relationship was concluded are not included in this chapter due primarily to
the uncertainty in the independent association between exposure to SO2 and the health
6-2
-------
outcome; as a result, these studies are unable to provide information on whether certain
populations are at increased risk of SCh-related health effects. Conclusions are drawn
based on the overall confidence that a specific factor may result in a population or
lifestage being at increased risk of an SCh-related health effect.
As discussed in the Preamble to the ISAs (U.S. EPA. 2015b). this evaluation includes
evidence from epidemiologic, controlled human exposure, and toxicological studies in
addition to considering relevant exposure-related information. With regard to
epidemiologic studies, the evaluation focuses on those studies that include stratified
analyses to compare populations or lifestages exposed to similar air pollutant
concentrations within the same study design along with consideration of the strengths and
limitations of each study. Other epidemiologic studies that do not stratify results but
instead examine a specific population or lifestage can provide supporting evidence for the
pattern of associations observed in studies that formally examine effect modification.
Similar to the characterization of evidence in Chapter 5. the greatest emphasis is placed
on patterns or trends in results across studies. Experimental studies in human subjects or
animal models that focus on factors, such as genetic background or health status, are
evaluated because they provide coherence and biological plausibility of effects observed
yin epidemiologic studies. Also evaluated are studies examining whether factors may
result in differential exposure to SO2 and subsequent increased risk of SCh-related health
effects. Conclusions are made with respect to whether a specific factor increases the risk
of an SCh-related health effect based on the characterization of evidence framework
detailed in Table 6-1. A summary of the characterization of the evidence for each factor
considered in this chapter is presented in Section 6.6.
6-3
-------
Table 6-1 Characterization of evidence for factors potentially increasing the
risk for sulfur dioxide-related health effects.
Classification Health Effects
Adequate There is substantial, consistent evidence within a discipline to conclude that a factor results in a
evidence population or lifestage being at increased risk of air pollutant-related health effect(s) relative to
some reference population or lifestage. Where applicable, this evidence includes coherence
across disciplines. Evidence includes multiple high-quality studies.
Suggestive The collective evidence suggests that a factor results in a population or lifestage being at
evidence increased risk of air pollutant-related health effect(s) relative to some reference population or
lifestage, but the evidence is limited due to some inconsistency within a discipline or, where
applicable, a lack of coherence across disciplines.
Inadequate The collective evidence is inadequate to determine whether a factor results in a population or
evidence lifestage being at increased risk of air pollutant-related health effect(s) relative to some reference
population or lifestage. The available studies are of insufficient quantity, quality, consistency,
and/or statistical power to permit a conclusion to be drawn.
Evidence of no There is substantial, consistent evidence within a discipline to conclude that a factor does not
effect result in a population or lifestage being at increased risk of air pollutant-related health effect(s)
relative to some reference population or lifestage. Where applicable, the evidence includes
coherence across disciplines. Evidence includes multiple high-quality studies.
6.3 Pre-existing Disease
Individuals with pre-existing disease may be considered at greater risk for some air
pollution-related health effects because disease status and severity may put those
individuals in a compromised biological state. The 2008 ISA for Sulfur Oxides (U.S.
EPA. 2008d') concluded that individuals with pre-existing respiratory diseases, especially
those with asthma, were likely to be at greater risk for SCh-related health effects. Recent
epidemiologic studies evaluated effect modification of both respiratory and
cardiovascular effects by a number of pre-existing diseases, including asthma,
cardiovascular disease, diabetes, and obesity. While we evaluated the evidence from each
of these studies, the focus of this section is on effect modification of respiratory disease
by pre-existing asthma since the relationship between short-term SO2 exposure and
respiratory health effects was determined to be causal in Chapter 5. with the strongest
evidence coming from studies of asthma exacerbation. Since the relationship between
short-term SO2 exposure and cardiovascular effects (the only outcome for which evidence
is available to inform pre-existing cardiovascular disease, diabetes, or obesity) was
determined to be inadequate, we evaluated the evidence to determine if stratification by
these pre-existing diseases could explain any of the inconsistencies in study results of
cardiovascular effects. There was no indication that pre-existing disease status led to
6-4
-------
heterogeneity in study results for cardiovascular effects, thus the results for those pre-
existing diseases are not detailed in this section.
Of the recent epidemiologic studies evaluating effect modification of respiratory effects
by pre-existing disease, most focused on asthma (Section 6.3.1). Table 6-2 presents the
prevalence of asthma and other respiratory diseases according to the Centers for Disease
Control and Prevention's (CDC's) National Center for Health Statistics (Blackwell et al..
2014) and the National Health Interview Survey (Bloom et al.. 2012) including the
proportion of adults and children with a current diagnosis categorized by age and
geographic region. The large proportions of the U.S. population affected by respiratory
diseases indicates the potential public health impact, and thus, the importance of
characterizing the risk of SCh-related health effects for affected populations.
6-5
-------
Table 6-2 Prevalence of respiratory diseases among adults and children by age
and region in the U.S. in 2012.
Adults3—By Age Adults3—By Region
Numbers in
Thousands, (%)
All 18+
18-44
45-64
65-74
75+
North-
east
Midwest
South
West
Total population
234,921
111,034
82,038
23,760
18,089
42,760
53,378
85,578
53,205
"Ever
had"
29,660
(12.6)
14,929
(13.4)
10,380
(12.7)
2,863
(12.0)
1,489
(8.2)
5,686
(13.3)
7,080
(13.3)
10,071
(11.8)
6,822
(12.8)
| "Still has"
18,719
(8.0)
8,943
(8.1)
6,852
(8.4)
1,837
(7.7)
1,088
(6.0)
3,953
(9.2)
4,358
(8.2)
6,280
(7.3)
4,129
(7.8)
Chronic
bronchitis
8,658
(3.7)
2,721
(2.5)
3,831
(4.7)
1,165
(4.9)
940
(5.2)
1,446
(3.4)
2,438
(4.6)
3,449
(4.0)
1,325
(2.5)
COPD
6,790
(2.9)
512
(0.5)
3,074
(3.7)
1,646
(6.9)
1,558
(8.6)
1,013
(2.4)
1,860
(3.5)
2,781
(3.2)
1,135
(2.1)
Child renc-
-By Age
Children0—By Region
Numbers in
Thousands, (%)
All <18
0-4
5-11
12-17
North-
east
Midwest
South
West
Total population
74,518
21,210
28,845
24,463
11,956
17,651
27,170
17,741
"Ever
had"
Si
10,463
(14.0)
1,753
(8.3)
4,139
(14.3)
4,571
(18.7)
1,888
(15.8)
2,190
(12.4)
4,088
(15.0)
2,298
(13.0)
TO
J "Still has"
-1—'
CO
<
7,074
(9.5)
1,452
(6.8)
2,849
(9.9)
2,773
(11.3)
1,356
(11.3)
1,557
(8.8)
2,636
(9.7)
1,525
(8.6)
COPD = chronic obstructive pulmonary disease.
aSource: Blackwell et al. (2014): National Center for Health Statistics: Data from Table 3 of the Centers for Disease Control and
Prevention report.
bAsthma prevalence estimates are available for both "ever had asthma" and "still has asthma."
°Source: Bloom et al. (2012): Summary health statistics for U.S. children: National Health Interview Survey, Vital and Health
Statistics, Series 10, Number 254. December, 2012.
6.3.1 Asthma
Approximately 8.0% of adults and 9.5% of children (age <18 years) in the U.S. currently
have asthma (Blackwell et al.. 2014; Bloom et al.. 2012). and it is the leading chronic
illness affecting children (Bloom et al.. 2012). Based on evidence from the 2008 ISA for
Sulfur Oxides (U.S. EPA. 2008d') and recent studies, a causal relationship exists between
short-term SO2 exposure and respiratory effects, based primarily on evidence from
controlled human exposure studies demonstrating decrements in lung function in
6-6
-------
individuals with asthma (Sections 5.2.1.2 and 5.2.1.9). This is nearly the same body of
evidence evaluated in the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d). which also
concluded that individuals with asthma are more sensitive to exposures to SO2. Children
with asthma may be particularly at risk compared to adults with asthma due to (1) their
increased responsiveness to methacholine, a potential surrogate for SO2 (Section 5.2.1.2).
relative to adults; (2) children's increased ventilation rates relative to body mass
compared to adults; and (3) the increased proportion of oral breathing observed among
children, particularly boys, relative to adults (Section 4.1.2). Such oral breathing allows
greater SO2 penetration into the tracheobronchial region of the lower airways than nasal
breathing (Section 4.2.2). In addition, children tend to spend more time outdoors (where
SO2 levels are higher, compared to indoor levels), and, consequently, have the potential
for longer exposure to higher levels of SO2. While there are a number of behavioral,
environmental and physical characteristics that, in addition to SO2, could contribute to
asthma exacerbations, there is little or no empirical evidence for how these characteristics
might interact with SO2 and contribute to individuals with asthma being more at risk for
health effects attributed to SO2 than healthy individuals. This section briefly describes
evidence from the experimental studies and supporting evidence from epidemiologic
studies (Table 6-3).
Across experimental evidence, adults with asthma consistently have greater decrements
in lung function with SO2 exposure than those without asthma. Controlled human
exposure studies have evaluated respiratory outcomes among adults at SO2
concentrations ranging from 0.2 to 1 ppm and included exposures with and without
exercise. In a controlled human exposure study examining SO2 concentrations ranging
from 0.2 to 0.4 ppm in repeated exposures, while exercising, of healthy individuals and
individuals with mild, atopic and moderate/severe asthma, Linn et al. (1987) reported that
individuals having moderate and severe asthma showed the greatest SC>2-dependent
respiratory effects (airway resistance, FEVi, symptoms). Other than asthma status,
subject-level characteristics (e.g., weight, height, age and sex of those having asthma) did
not influence the response. Magnussen et al. (1990) also reported increases in sRaw in
subjects with asthma that were not observed in healthy controls, with SO2 exposures
incorporating exercise. In addition to controlled human exposure studies, a long-term
exposure study conducted in ovalbumin (OVA)-sensitized rats as an asthma model
demonstrated that 4 weeks of exposure to 2 ppm SO2 resulted in increased airway
resistance compared to normal rats (Song et al.. 2012).
6-7
-------
Table 6-3
Controlled human exposure, epidemiology, and animal toxicology
studies evaluating pre-existing asthma and sulfur dioxide exposure.
Factor
Evaluated
Direction of
Effect
Reference Modification
Category or Effect3
Outcome
Study
Population13
Study Details Study
Controlled human exposure
Asthma
(atopic)
Healthy f
Lung function
(sRaw)
n = 4 healthy
adults
0.2. 0.4. 0.6 DDm Linn et al. (1987)
SO2 for 1 h with
exercise;
Exposures were
repeated eight
times
Mild asthma
T
n - 21 atopic
adults
Moderate/
severe
asthma
T
n = 16 adults
with mild
asthma
n = 24 adults
Asthma
(atopic)
Healthy f
Lung function
(FEVi)
with moderate/
severe asthma
Mild asthma
T
Moderate/
severe
asthma
T
Asthma
(atopic)
Healthy f
Respiratory
symptoms
¦ during exposure
Mild asthma
T
Moderate/
severe
asthma
T
Asthma
Healthy f
Lung function
(sRaw)
n = 46 adults
with bronchial
asthma,
12 healthy
adults
0.5 DDm SO? for Maanussen et al.
10 min tidal (1990)
breathing,
10 min of
isocapnic
hyperventilation
(30 L/min);
histamine
challenge
Asthma
Healthy
Lung function
(FEVi, FVC,
MMEF)
n = 12 adults
with asthma,
12 healthy
adults
0.2 DDm SO2 for Tunnicliffe et al.
1 h at rest (2003)
6-8
-------
Table 6-3 (Continued): Controlled human exposure, epidemiology, and animal
toxicology studies evaluating pre existing asthma and
sulfur dioxide exposure
Direction of
Effect
Factor
Reference
Modification
Study
Evaluated
Category
or Effect3
Outcome
Population13
Study Details
Study
Epidemiology
With asthma
Without
_
Lung function
n = 506
Guadeloupe
Amadeo et al.
n = 84
asthma
(PEF)
elementary
(French West
(2015)
n = 422
school children
Indies)
ages 8-13 yr
December
2008-December
2009
Toxicology
Rat asthma
model (OVA
sensitization)
Normal
rats
AHR (metha-
choline)
IL-4 in BALF
Rats (Sprague-
Dawley),
¦ n = 10
males/group
, (4 wk)
2 ppm SO2 for
4 h/day for 4 wk
beginning at
15 days
Song etal. (2012)
IFN-y in BALF
Airway smooth
muscle cell
stiffness
(in vitro)
Airway smooth
muscle cell
contractility
(in vitro)
AHR = airway hyperresponsiveness; BALF = bronchoalveolar lavage fluid; FE\A| = forced expiratory volume in 1 sec; FVC = forced
vital capacity; IFN-y = interferon gamma; IL-4 = interleukin 4; MMEF = maximum mid-expiratory flow; n = sample size;
OVA = ovalbumin; PEF = peak expiratory flow; S02 = sulfur dioxide; sRAW = specific airway resistance.
aUp facing arrow (j) indicates that the effect of S02 is greater (e.g., larger lung function decrement, larger increase in airway
inflammation) in the group with the factor evaluated than in the reference group. Down facing arrow (I) indicates that the effect of
S02 is smaller in the group with the factor evaluated than in the reference group. A dash (-) indicates no substantial difference in
S02-related health effect between groups. In some studies, only a population with pre-existing disease was examined; therefore,
the arrow or dash represents the direction of the effect in that population after exposure to S02 relative to exposure to filtered air.
bUnless ages are indicated in the row for each study, the mean age or range was not reported in the study aside from indication of
adult subjects.
Of the recent literature included in this ISA, one epidemiologic study included
stratification by asthma status and did not find a difference for associations of short-term
SO2 exposure with changes in lung function ITable 6-3; (Amadco et al.. 2015)1. However,
evidence presented in Section 5.2.1.2 from many studies generally demonstrates
consistent positive associations between ambient SO2 concentrations and asthma-related
hospitalizations and ED visits. This is important supporting evidence because
hospitalizations and ED visits related to asthma would generally not be studied in
6-9
-------
individuals without asthma, and thus, are not amendable to stratification. In addition to
these studies, some evidence from recent panel studies (Dong et al.. 2013c; Sahsuvaroglu
et al.. 2009) and studies reviewed in the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d)
indicates that children with asthma experience respiratory symptoms associated with
exposure to ambient SO2.
In conclusion, evidence from controlled human exposure and animal toxicological studies
is consistent in demonstrating decrements in lung function with SO2 exposures in people
with asthma. The evidence for healthy individuals does not show responses at exposures
below 1 ppm (as summarized above and in Section 5.2.1.2. There is also clear biological
plausibility including key events contributing to the mode of action (Section 4.3) linking
SO2 exposure to asthma exacerbation and supporting the observed effects from
experimental studies. Furthermore, epidemiologic studies report associations between
SO2 exposure and ED visits and hospital admissions due to asthma, and that individuals
with asthma experience respiratory symptoms associated with exposure to ambient SO2.
Overall, there is adequate evidence from multiple, high-quality studies and
coherence across scientific disciplines to conclude that people with pre-existing
asthma are at increased risk of S02-induced respiratory effects.
6.4 Genetic Factors
Genetic variation in the human population is known to contribute to numerous diseases
and differential physiologic responses. The 2008 ISA for Sulfur Oxides (U.S. EPA.
2008d) discussed the biological plausibility of individuals with certain genotypes known
to result in reduced function in genes encoding antioxidant enzymes being at increased
risk for respiratory effects related to ambient air pollution. However, the evidence base
was limited to two studies demonstrating individuals with polymorphisms in GSTP1 and
tumor necrosis factor to be at increased risk for S02-related asthma and decrements in
lung function. A recently conducted study reviewed in this ISA examined effect measure
modification by genotype (Reddv et al.. 2012) and reported inconsistent results across
GSTM1 and GSTP1 genotypes in a relatively small sample of children in South Africa.
The GSTM1 null genotype and the GSTP1 Ilel05Ile and He 105 Val genotypes are
associated with reduced antioxidant enzyme function; however, effect measure
modification of these genotypes on S02-associated intra-day variability of FEVi showed
conflicting results. Despite biological plausibility characterized in the 2008 SOx ISA,
the limited and inconsistent evidence base is inadequate to determine whether
genetic background contributes to increased risk for S02-related health effects.
6-10
-------
6.5
Sociodemographic and Behavioral Factors
Recent epidemiologic evidence evaluated effect modification of respiratory,
cardiovascular, cancer, reproductive and developmental, and mortality effects associated
with SO2 exposure by a number of sociodemographic and behavioral factors, including
lifestage, sex, socioeconomic status, race/ethnicity and smoking. While we evaluated the
evidence from each of these studies within this ISA, the focus of this section is on effect
modification of the association of SO2 exposure with respiratory disease since the
relationship determined to be causal is that for short-term SO2 exposure and respiratory
health effects (see Chapter 5). Since the evidence for relationships between SO2 exposure
and cardiovascular disease, cancer, and reproductive and developmental outcomes were
determined to be inadequate in Chapter 5. we evaluated the evidence to determine if
stratification by sociodemographic or behavioral factors could explain any of the
inconsistencies in study results. There was no indication that sociodemographic and
behavioral factors led to heterogeneity in study results for cardiovascular disease,
cancer, or reproductive and developmental outcomes, thus those results are not
detailed in this section.
6.5.1 Lifestage
The 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d) discussed some evidence for
increased risk of health effects related to SO2 exposure among different lifestages
(i.e., children and older adults). Lifestage refers to a distinguishable time frame in an
individual's life characterized by unique and relatively stable behavioral or physiological
characteristics associated with development and growth (U.S. EPA. 2014c). Differential
health effects of SO2 across lifestages theoretically could be due to several factors. With
regard to children, the human respiratory system is not fully developed until 17-21 years
of age for females and 19-23 years of age for males, and therefore, children could
plausibly have intrinsic risk for respiratory effects due to potential perturbations in
normal lung development (Finkelstein and Johnston. 2004; Hankinson et al.. 1999). In
addition, children spend more time outdoors compared with adults, and, as a result, may
experience greater SO2 exposure (Section 3.4.2.1). Older adults (typically considered
those 65 years of age or greater) have weakened immune function, impaired healing,
decrements in pulmonary and cardiovascular function, and greater prevalence of chronic
disease [(Rosenthal and Kavic. 2004); Table 6-21. which may contribute to or worsen
health effects related to SO2 exposure. Also, exposure or internal dose of SO2 may vary
across lifestages due to varying ventilation rates, extent of oronasal breathing at rest, and
time-activity patterns. The following sections present the evidence comparing lifestages
6-11
-------
from the recent literature, which builds on the evidence presented in the 2008 ISA for
Sulfur Oxides (U.S. EPA. 2008d).
6.5.1.1 Children
According to the 2010 census, 24% of the U.S. population is less than 18 years of age,
with 6.5% less than age 6 (Howdcn and Mever. 2011). The large proportion of children
within the U.S. demonstrates the public health importance of characterizing the risk of
SCh-related health effects among children. This is especially so because of the causal
relationship between short-term SO2 exposure and respiratory outcomes, with strong
evidence demonstrating lung function decrements in individuals with asthma, which
affects approximately 11.3% of children 12 -17 years old and 9.5% of children less than
18 years old (Table 6-2). The 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d) presented
evidence from epidemiologic studies indicating larger effect estimates for S02-related
respiratory outcomes among children than for adults or all ages, including asthma-related
ED visits and hospitalizations; however, recent evidence is not entirely consistent with
the evidence considered previously (Table 6-4). Although Son et al. (2013) found
children (0-14 years) to be at greater risk for S02-related asthma hospital admissions
compared with adults, neither Ko et al. (2007b') nor Alhanti et al. (2016) observed
differences between children and adults when examining associations of ambient SO2
with asthma hospitalizations or ED visits. When examining evidence for different age
groups of children, Jalaludin et al. (2008) observed that associations for
respiratory-related ED visits among children ages 1-4 years were greater than for
children ages 10-14 years; however, Samoli et al. (2011) and Villeneuve et al. (2007) did
not find stronger associations for asthma-related hospital admissions or ED visits among
younger children. Similarly, Dong et al. (2013c) did not find age-related differences
among children for S02-associated asthma, and Sahsuvaroglu et al. (2009) found children
ages 6-7 years had smaller risk for S02-associated nonallergic asthma compared to
adolescents at 13-14 years.
Overall, the combined evidence from the previous and current ISA examining
respiratory outcomes across lifestages is suggestive of increased risk in children
compared to adults, given the inconsistencies across epidemiologic studies and
limited toxicological evidence to inform plausibility. In addition, there is no evidence
that younger (e.g., <7 years old) or older (e.g., 8-18 years old) children are more at risk
for asthma exacerbation-related effects due to short-term SO2 exposure than children in
general. There are biological factors (e.g., increased ventilation rates relative to body
mass among children and increased oral breathing that lead to greater SO2 penetration
and bronchial surface doses) that indicate a potential for children to receive a greater
6-12
-------
internal dose in the bronchotracheal region, and for children with asthma to be at
increased risk as compared to adults with asthma (Section 6.3). However, recent
evidence, mainly from epidemiologic studies of respiratory ED visits and hospital
admissions, does not consistently show increased risk among children (Table 6-4).
Table 6-4 Epidemiologic studies evaluating childhood lifestage and sulfur
dioxide exposure.
Direction of
Factor
Reference
Effect
Study
Evaluated
Category
Modification3
Outcome
Population
Study Details
Study
Short-term exposure
Childhood ages
All ages
1
Hospital
14 hospitals
Hong Kong,
Wona et al. (2009)
0-14 yr
n = 104.9
admissions
China
n =60.1
admissions/day
for acute
1996-2002
admissions/day
respiratory
distress
Childhood ages
Adulthood ages
_
Asthma
15 hospitals
Hong Kong,
Ko et al. (2007b)
0-14 yr
15-65 yr
hospital
n =69,176
China
n =23,596
n =21,204
admissions
admissions
2000-2005
Childhood ages
Adulthood ages
T
Asthma
Database
Eight South
Son et al. (2013)
0-14 yr
15-64 yr
hospital
accounting for
Korean cities
n = 8.7
n =4.3
admissions
48% of South
2003-2008
admissions/day
admissions/day
Korean
population
n = 19/d
Childhood ages
Childhood ages
_
Asthma
Three main
Athens,
Samoli et al. (2011)
0-4 yr
5-14 yr
hospital
children's
Greece
n = 72%
n = 28%
admissions
hospitals
2001-2004
approximately
85% of pediatric
beds of
metropolitan
area of Athens
n = 3,601
Childhood ages
Childhood ages
_
Asthma ED
Five hospitals
Edmonton,
Villeneuve et al.
2-4 yr
5-14 yr
visits
servicing more
Canada
(2007)
n = 7,247
n = 13,145
than 80% of the
1992-2002
metropolitan
area
n = 57,192
visits
6-13
-------
Table 6-4 (Continued): Epidemiologic studies evaluating childhood lifestage and
sulfur dioxide exposure
Direction of
Factor
Reference
Effect
Study
Evaluated
Category
Modification3
Outcome
Population
Study Details
Study
Childhood ages
Childhood ages
T
Respiratory-
Daily number of
Sydney,
Jalaludin et al.
1-4 yr
10-14 yr
related ED
ED visits in
Australia
(2008)
n = 109
n =25
visits
metropolitan
1997-2001
admissions/day
admissions/day
Sydney from
the New South
Wales Health
Department
n = 174/d
Childhood ages
Adulthood ages
_
Asthma ED
Daily number of
Three U.S.
Alhanti etal. (2016)
5-18 yr
19-39 yr
visits
ED visits in
cities (Atlanta,
n = 59.6
n =41.1
metropolitan
GA
admissions/day
admissions/day
area
1993-2009;
n = 62.8/day
Dallas, TX
(Atlanta)
2006-2009;
St. Louis, MO
n = 76.3/day
2001-2007)
(Dallas)
n = 50.6/day
(St. Louis)
Long-term exposure
Childhood ages
Childhood ages
_
Doctor-
n = 31,049
Seven
Dona et al. (2013c)
2-5 yr
6-14 yr
diagnosed
Children
northeastern
n = 7,508
n =23,541
asthma
ages 2-14 yr
cities study,
Liaoning
T
Respiratory
Provence,
symptoms
northeast
(cough,
China
phlegm,
2008-2009
current
wheeze)
Younger
Older children
1
Nonallergic
n~ 1,467
Hamilton,
Sahsuvaroqlu et al.
children ages
ages 13-14 yr
asthma
Children grades
Canada
(2009)
6-7 yr
n = 549
1 (ages 6-7 yr)
1994-1995
n = 918
and 8 (ages
13-14 yr)
ED = emergency department; n = sample size.
aUp facing arrow indicates that the effect of is greater (e.g., larger increase in hospital admission) in the group with the factor evaluated
than in the reference group. Down facing arrow indicates that the effect of sulfur dioxide is smaller in the group with the factor evaluated
than in the reference group. A dash indicates no substantial difference in sulfur dioxide-related health effect between groups.
6-14
-------
6.5.1.2 Older Adults
According to the 2008 National Population Projections issued by the U.S. Census
Bureau, approximately 12.9% of the U.S. population is age 65 years or older, and by
2030, this fraction is estimated to grow to 20% (Vincent and Velkoff. 2010). Thus, this
lifestage represents a substantial proportion of the U.S. population demonstrating the
public health importance of characterizing the potential for increased risk for health
effects related to SO2 exposure in this age group.
The 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d) indicated that compared with
younger adults, older adults (typically ages 65 years and older) may be at increased risk
for SC>2-related respiratory emergency department visits and hospitalizations, but limited
evidence was available to inform risk related to respiratory effects. Recently published
studies evaluating risk in older adults compared to younger adults are characterized in
Table 6-5 and generally support conclusions from the 2008 ISA for Sulfur Oxides (U.S.
EPA. 2008d). For example, Villeneuve et al. (2007) and Son et al. (2013) both reported
that asthma-related ED visits and hospital admissions were more strongly associated with
short-term ambient SO2 exposure in individuals older than 75 years than adults
65-74 years or those younger than 65. However, the handful of recent studies evaluating
asthma and nonasthma respiratory admissions or ED visits in adults older than 65 years
of age reported inconsistent results compared to the earlier literature (Alhanti et al.. 2016;
Son etal.. 2013; Arbex et al.. 2009; Wong et al.. 2009; Ko et al.. 2007b'). In addition to
these studies of short-term SO2 exposure, Forbes et al. (2009c') found older adults (45-74
and older than 75 years) to have larger decrements in lung function in association with
average annual SO2 concentrations, compared to adults aged 16-44.
Other recent studies comparing results in older and younger adults evaluated associations
between short-term SO2 exposures and mortality and generally observed inconsistent
results (Bravo et al.. 2015; Chen et al.. 2012c; Wong et al.. 2008b). Chen et al. (2012c)
and Wong et al. (2008b') both observed evidence for increased risk of total mortality with
short-term SO2 exposures in adults older than 75 years compared to other age groups,
which is consistent with age-specific evidence from studies of respiratory effects.
However, Bravo et al. (2015) did not observe any differences in risk across age groups.
Taken together, the collective evidence builds on conclusions from the 2008 ISA for
Sulfur Oxides (U.S. EPA. 2008d) and is suggestive that older adults may be at
increased risk for S02-related health effects compared to younger adults. The
evidence from the current and previous ISA related to respiratory hospitalizations and ED
visits indicates that older adults, particularly those older than 75 years, may be at
6-15
-------
increased risk for SCh-related health effects, although this evidence is not entirely
consistent.
Table 6-5
Epidemiologic studies evaluating older adult lifestage and sulfur
dioxide exposure.
Direction of
Factor
Reference
Effect
Study
Evaluated
Category
Modification3
Outcome
Study Population
Details
Study
Older
Younger
_
Asthma
15 hospitals
Hong Kong,
Ko et al. (2007b)
adulthood
adulthood
hospital
n = 69,176 admissions
China
ages >65 yr
ages
admissions
2000-2005
n =24,916
15-65 yr
n = 21,204
Older
Younger
_
Asthma ED
Five hospitals
Edmonton,
Villeneuve et al.
adulthood
adulthood
visits
n = 57,912 visits
Canada
(2007)
ages
ages
1992-2002
65-74 yr
15-64 yr
n =4,705
n = 32,815
Older
T
adulthood
ages >75 yr
n = 1,855
Adulthood
Adulthood
_
Asthma ED
Daily number of ED
Three U.S.
Alhanti et al. (2016)
ages 65+ yr
ages
visits
visits in metropolitan
cities
n =4.7
19-39 yr
area
(Atlanta, GA
admissions/
n = 41.1
n = 62.8/day (Atlanta)
1993-2009;
day
admissions/
day
n = 76.3/day (Dallas)
Dallas, TX
2006-2009;
n = 50.6/day
St. Louis,
(St. Louis)
MO
2001-2007)
Older
Younger
T
COPDED
Sao Paulo Hospital,
Sao Paulo,
Arbex et al. (2009)
adulthood
adulthood
visits
daily records for
Brazil
ages >65 yr
ages
patients >40 yr
2001-2003
n = 789
40-64 yr
n = 1,769
n = 980
Older
Younger
_
Asthma and
Hospital admission
Eight South
Son et al. (2013)
adulthood
adulthood
allergic
database accounting
Korean cities
ages
ages
disease
for 48% of Korean
2003-2008
65-74 yr
15-64 yr
hospital
population
n = 5.8
n = 8.8
admissions
n = 37.7/day
admissions/
admissions/
day
day
6-16
-------
Table 6-5 (Continued): Epidemiologic studies evaluating older adult lifestage and
sulfur dioxide exposure.
Direction of
Factor
Reference
Effect
Study
Evaluated
Category
Modification3
Outcome
Study Population
Details
Study
Older
Younger
T
adulthood
adulthood
ages >75 yr
ages
n = 5.8
15-64 yr
admissions/
n = 8.8
day
admissions/
day
Older
All ages
_
COPD
14 hospitals
Hong Kong,
Wona et al. (2009)
adulthood
n = 91.5
hospital
China
ages >65 yr
admissions
1996-2002
n = 59.6
Older
All ages
_
Respiratory
adulthood
n = 270.3
disease
ages >65 yr
hospital
n = 138.5
admissions
Older
Adulthood,
T
Total
Data from Municipal
17 Chinese
Chen et al. (2012c)
adulthood
childhood
mortality
Centers for Disease
cities
ages
ages
Control and
>65 yrb
5-64 yrb
Prevention
Older
All ages
T
Total
Data from the Ministry
Bangkok,
Wona et al. (2008b)
adulthood
(>65 yr)
mortality
of Public Health,
Thailand;
ages >75 yr
Bangkok; the Census
Hong Kong,
and Statistic
Shanghai,
Department, Hong
and Wuhan,
Kong; the Shanghai
China
Municipal Center of
1996-2004
Disease Control and
Prevention, Shanghai;
and the Wuhan Centre
for Disease Prevention
and Control
Older
Younger
T
Mortality
N = 849,127
Sao Paulo,
Bravo et al. (2015)
adulthood
adulthood
Brazil
ages
ages 35-64
May 1996-
65-74 yr
n = 315,435
December
n = 194,202
2010
Older
Younger
T
adulthood
adulthood
ages >75 yr
ages 35-64
n = 339,490
n = 315,435
COPD = chronic obstructive pulmonary disease; ED = emergency department; n = sample size.
aUp facing arrow indicates that the effect of sulfur dioxide is greater (e.g., larger risk of hospital admission, larger decrement in heart
rate variability) in the group with the factor evaluated than in the reference group. Down facing arrow indicates that the effect of
sulfur dioxide is smaller in the group with the factor evaluated than in the reference group. A dash indicates no substantial difference
in sulfur dioxide-related health effect between groups.
bSample size not reported.
6-17
-------
6.5.2 Sex
A vast number of health conditions and diseases have been shown to differ by sex, and
there is some indication of differences by sex in the relationship between air pollution
and health effects. The 2010 U.S. census indicates an approximately equal distribution of
males and females in the U.S.: 49.2% male and 50.8% female (Howdcn and Meyer.
2011). However, the distribution varies by age, with a greater prevalence of females
above 65 years of age compared to males. Thus, the public health implications of
potential sex-based differences in SC>2-related health effects may vary among age groups
within the population.
There are a number of studies evaluating sex-based differences in SCh-associated health
effects, as detailed in Table 6-6. Studies of short-term SO2 exposures and respiratory
effects in children and adults did not consistently indicate differences by sex. Ishigami et
al. (2008) found adult females to have increased respiratory symptoms with ambient SO2
exposure compared to adult males; however, Son et al. (2013) found larger associations
for asthma or allergic disease hospitalizations for males compared to females. No
differences were found between men and women for S02-related COPD ED visits (Arbex
et al.. 2009). In children, S02-associated decrements in lung function were not different
between boys and girls (Linares et al.. 2010; Dales et al.. 2009). although Samoli et al.
(2011) found boys to have higher associations between ambient SO2 exposure and asthma
hospital admissions. In a long-term SO2 exposure study, Deng et al. (2015a) observed
stronger associations with asthma incidence among boys compared to girls.
The collective body of evidence does not clearly indicate that S02-related health effects
differ between males and females. Due to the inconsistent results demonstrated across
epidemiologic studies and a lack of experimental studies examining sex-based
differences, the evidence is inadequate to determine whether males or females may
be at increased risk for SCh-related health effects.
6-18
-------
Table 6-6 Epidemiologic studies evaluating effect modification by sex and
sulfur dioxide exposure.
Direction of
Factor
Reference
Effect
Study
Evaluated
Category
Modification3
Outcome
Population
Study Details
Study
Short-term exposure
Female
Male
T
Respiratory
Healthy adult
Miyakejima Island,
Ishiqami et al.
20% person
80% person
symptoms
volunteers
Japan
(2008)
h
h
(cough, scratchy
working on an
2005
throat, sore
active volcanic
throat,
island after the
breathlessness)
evacuation order
was lifted
n = 955
Female
Male
_
Lung function
Elementary
Windsor, Canada
Dales et al.
n = 39
n = 114
(FEVi)
school children
October-December
(2009)
with asthma (no
2005
cigarette smoking
in home)
n = 182 children
(ages 9-14 yr)
Female
Male
_
Lung function
Children
Salamanca, Mexico
Linares et al.
n =235
n =229
(FEVi, FVC,
recruited from
2004-2005
(2010)
PEF,
two schools with
FEVi/FVC)
different roadway
proximity
n =464 (6-14 yr)
Female
Male
_
COPD ED visits
Sao Paulo
Sao Paulo, Brazil
Arbex et al.
n = 794
n = 875
Hospital, daily
2001-2003
(2009)
records for
patients >40 yr
n = 1,769
Female
Male
Asthma hospital
Database
Eight South Korean
Son et al. (2013)
n = 7.4
n = 8
admissions
accounting for
cities
admissions/
admissions/
48% of South
2003-2008
day
day
Korean
population
Female
Male
1
Allergic disease
n = 19/day
n = 7.1
n = 8
hospital
admissions/
admissions/
admissions
day
day
6-19
-------
Table 6-6 (Continued): Epidemiologic studies evaluating effect modification by
sex and sulfur dioxide exposure.
Factor
Evaluated
Reference
Category
Direction of
Effect
Modification3
Outcome
Study
Population
Study Details
Study
Female Male
n = 1,332 n= 2,269
Asthma hospital
admissions
Three main
children's
hospitals;
approximately
85% of pediatric
beds of
metropolitan area
of Athens
n = 3,601
Athens, Greece
2001-2004
Samoli et al.
(2011)
Long-term exposure
Asthma Children from Changsha, China Deng et al
incidence 36 different (2015a)
kindergartens
n =2,490
COPD = chronic obstructive pulmonary disease; ED = emergency department; FE\A| = forced expiratory volume in 1 sec; FVC = forced
vital capacity; n = sample size; PEF = peak expiratory flow.
aUp facing arrow indicates that the effect of S02 is greater (e.g., larger risk of hospital admission) in the group with the factor evaluated
than in the reference group. Down facing arrow indicates that the effect of sulfur dioxide is smaller in the group with the factor evaluated
than in the reference group. A dash indicates no substantial difference in sulfur dioxide-related health effect between groups.
Female Male J,
n = 1,153 n = 1,337
6.5.3 Socioeconomic Status
SES is a composite measure that usually consists of economic status indicated by income,
social status by education, and work status by occupation. Generally, persons with lower
SES have been found to have a higher prevalence of pre-existing diseases, potential
inequities in access to resources such as healthcare, and possibly increased nutritional
deficiencies, which may increase their risk for SCh-related health effects (Wong et al..
2008a; WHO. 2006). According to U.S. census data, 15.9% (approximately 48.5 million)
of Americans lived below the poverty threshold in 2011 as defined by household income,
which is one metric used to define SES (Bishaw. 2012). The wide array of SES factors
that can be used to describe or assign SES can complicate any synthesis of findings
because definitions of SES vary across countries based on population demographics,
bureaucracy, and the local economy. In addition, lower SES may coincide with proximity
to pollution sources in some locations. As a result of these complexities, the ability to
draw conclusions regarding SES as a factor for increased risk for health effects related to
SO2 exposure can be difficult.
6-20
-------
A single study (Cakmak et al.. 2016) evaluated the potential for SES (income or
education) to modify the effect of long-term exposure to SO2 on respiratory effects,
specifically measures of lung function. The authors observed greater decrements in lung
function for those in the lowest income and education groups when compared to those in
the highest. Similarly, a single studies provides limited evidence for the potential for SES
to modify the effect of short-term SO2 exposure on mortality. Chen et al. (2012c) found
low education to increase risk for mortality with short-term SO2 exposure. Overall, the
evidence for effect modification by SES on S02-related health outcomes is limited to a
single study of respiratory health effects and one on mortality. This limited evidence is
inadequate to determine whether low SES increases risk for SCh-related health
effects.
6.5.4 Smoking
Smoking is a common behavior as indicated by the 2010 National Health Interview
Survey, which estimated that approximately 19.2% of the U.S. adult population report
being current smokers and 21.5% report being former smokers (Schiller et al.. 2012).
Smoking is a well-documented risk factor for many diseases, but it is unclear whether
smoking exacerbates health effects associated with air pollutant exposures, including
S02.
Dong et al. (2012). Forbes et al. (2009c). and Smith et al. (2016) investigated effect
modification of the relationship between long-term exposure to SO2 and respiratory
endpoints by smoking status. Dong et al. (2012) found that among the few respiratory
deaths included in their retrospective cohort study, associations with long-term ambient
SO2 concentrations were only present with current smoking status. Smith et al. (2016)
observed positive associations between long-term average SO2 concentration and
pulmonary tuberculosis among ever smokers, but not with never smokers. Forbes et al.
(2009c). on the other hand, did not find current smoking to increase risk for lung function
decrements with long-term SO2 exposure compared to not smoking; however, former
smoking did appear to increase risk in this study.
Overall, the inconsistent evidence is inadequate to determine whether smoking
exacerbates S02-related health effects. A limited number of long-term exposure studies
observed positive associations among current or former smokers, but not for never
smokers for various respiratory health endpoints, including respiratory mortality. No
studies evaluated smoking as an effect modifier of the relationship between short-term
exposure to SO2 and respiratory outcomes, for which there is the most confidence in the
causal nature of the relationship.
6-21
-------
6.6
Conclusions
This chapter characterized factors that may result in populations and lifestages being at
increased risk for SCh-related health effects; a summary of at-risk factors and resulting
evidence classifications is included in Table 6-7. The evaluation of each factor focused
on the consistency, coherence, and biological plausibility of evidence integrated across
scientific disciplines: specifically, epidemiologic, controlled human exposure, and
toxicological studies using the weight-of-evidence approach detailed in Table 6-1. In
evaluating and integrating evidence related to at-risk factors, it is important to consider
additional information including exposure concentrations, dosimetry, modes of action,
and/or the independence of relationships of SO2 exposure with health effects as detailed
in Chapter 5. For many potential at-risk factors summarized in Table 6-1. there was
limited evidence of an influence on SC>2-related health effects.
Consistent with observations made in the 2008 ISA for Sulfur Oxides (U.S. EPA. 2008d).
the evidence is adequate to conclude that people with asthma are at increased risk
for S02-related health effects. Most of the evidence for this conclusion was presented in
the previous ISA, but recent studies consistently indicate increased risk across studies.
Furthermore, the evidence is based on findings for short-term SO2 exposure and
respiratory effects (specifically lung function decrements), for which a causal relationship
exists (Section 5.2.1.9). There are a limited number of epidemiologic studies evaluating
SCh-related respiratory effects in people with asthma, but there is evidence for
asthma-related hospital admissions and ED visits (Section 5.2.1.2). Further support for
increased risk in individuals with asthma is provided by biological plausibility drawn
from modes of action. Asthma prevalence in the U.S. is approximately 8-11% across age
groups (Blackwell et al.. 2014; Bloom et al.. 2012) and thus represents a substantial
fraction of the population that may be at risk for respiratory effects related to ambient air
SO2 concentrations.
There is suggestive evidence of an increased risk of S02-related respiratory effects
in children and older adults. Although the 2008 ISA for Sulfur Oxides (U.S. EPA.
2008d) discussed several studies indicating stronger associations between SO2 and
respiratory outcomes for these lifestages, the recent evidence is less consistent. For
children, studies comparing S02-associated respiratory outcomes reported mixed results,
but known age-related factors such as higher ventilation rates (Section 4.1.2.1) and
time-activity patterns (Section 3.4.2.1) provide plausibility for higher SO2 exposure
and/or dose in children. For adults, recent research generally finds similar associations for
S02-related respiratory outcomes across age groups, although individuals over 75 years
were more consistently at increased risk. In addition, there was limited toxicological
evidence to support observations made across epidemiologic studies.
6-22
-------
Table 6-7 Summary of evidence for potential increased sulfur dioxide exposure
and increased risk of sulfur dioxide-related health effects.
Factor At-Risk
Evaluated Group
Health
Evidence
Rationale for Classification
Adequate evidence
Pre-existing Individuals
disease with asthma
Respiratory
Consistent evidence for increased risk for SC>2-related lung
function decrements in controlled human exposure studies
Support provided by epidemiologic studies of hospital admissions
and ED visits for respiratory causes
Suggestive evidence
Lifestage Children
Respiratory
Evidence for increased risk among children provided in 2008 SOx
ISA; older studies provide biological plausibility; recent
epidemiologic studies provide limited support, and are not entirely
consistent
Older adults
Respiratory
Evidence for increased risk for older adults provided in 2008 SOx
ISA; mixed results in recent epidemiologic studies for
respiratory-related outcomes
Mortality
Generally inconsistent evidence from a limited number of recent
epidemiologic studies of short-term SO2 exposure and mortality
Inadequate evidence
Genetic None
background identified
Respiratory
Single epidemiologic study shows decreased risk among GSTM1
individuals and increased risk among GSTP1 lle/lle (AA)
individuals
Sex None
identified
Respiratory
Inconsistent differences in S02-related health effects, or no
observed differences; studies limited in quantity
Socioeconomic None
status identified
Respiratory
Evidence limited to single study
Mortality
Evidence limited to single study
Smoking None
identified
Respiratory
Single controlled human exposure study saw no effect on lung
function among current smokers but an increase in risk among
forger smokers compared to never smokers
Evidence of no effect
None
ED = emergency department; ISA =
Integrated Science Assessment; S02 = sulfur dioxide.
6-23
-------
For all other at-risk factors considered based on available information, evidence was
inadequate to determine whether those factors result in increased risk for
SC>2-related health effects. Generally, there was a limited number of studies available
evaluating SES, genetic background, and smoking. Many of these factors are interrelated
and are known to impact health risks related to air pollution in general, but the scientific
evidence available in the published literature specific to health effects associated with
ambient SO2 exposure is inadequate to determine whether these factors confer increased
risk.
6-24
-------
References
Abbey. DE: Nishino. N: McDonnell. WF: Burchette. RJ: Knutsea SF: Beesoa WL: Yang. JX. (1999). Long-term
inhalable particles and other air pollutants related to mortality in nonsmokers. Am J Respir Crit Care Med
159: 373-382. http://dx.doi.org/10.1164/airccm. 159.2.9806020
Abdul-Wahab. SA: Ali. S: Sardar. S: Irfan. N: Al-Damkhi. A. li. (2011). Evaluating the performance of an
integrated CALPUFF-MM5 modeling system for predicting S02 emission from a refinery. Clean Tech
EnvironPol 13: 841-854. http://dx.doi.org/10.1007/sl0098-011-0360-6
Abe. M. (1967). Effects of mixed N02-S02 gas on human pulmonary functions: effects of air pollution on the
human body. J Med Dent Sci 14: 415-433.
Abraham. WM: Oliver. W. Jr: Welker. MJ: King. MM: Wanner. A: Sackner. MA. (1981). Differences in airway
reactivity in normal and allergic sheep after exposure to sulfur dioxide. J Appl Physiol (1985) 51: 1651-
1656.
Ackermann-Liebrich. U: Leuenberger. P: Schwartz. J: Schindler. C: Mona C: Bolognini. G: Bongard. JP: Brandli.
O: Domenighetti. G: Elsasser. S: Grize. L: Karrer. W: Keller. R: Keller-Wossidlo. H: Ktinzli. N: Martin.
BW: Medici. TC: Perruchoud. AP: Schoni. MH: Tschopp. JM: Villiger. B: Wtithrich. B: Zellweger. JP:
Zemp. E. (1997). Lung function and long term exposure to air pollutants in Switzerland. Am J Respir Crit
Care Med 155: 122-129. http://dx.doi.org/10.1164/airccm. 155.1.9001300
Adamkiewicz. G: Ebelt. S: Svring. M: Slater. J: Speizer. FE: Schwartz. J: Suh. H: Gold. DR. (2004). Association
between air pollution exposure and exhaled nitric oxide in an elderly population. Thorax 59: 204-209.
http://dx.doi.org/10.1136/thorax.2003.0Q6445
Adams. C: Riggs. P: Volckens. J. (2009). Development of a method for personal, spatiotemporal exposure
assessment. J Environ Monit 11: 1331-1339. http://dx.doi.org/10.1039/b903841h
Agav-Shav. K: Friger. M: Linn. S: Peled. A: Amitai. Y: Peretz. C. (2013). Air pollution and congenital heart defects.
Environ Res 124: 28-34. http://dx.doi.Org/10.1016/i.envres.2013.03.005
Akinbami. LJ: Lynch. CD: Parker. JD: Woodruff. TJ. (2010). The association between childhood asthma prevalence
and monitored air pollutants in metropolitan areas, United States, 2001-2004. Environ Res 110: 294-301.
http://dx.doi.Org/10.1016/i.envres.2010.01.001
Alarie. Y. (1973). Sensory irritation by airborne chemicals [Review]. CRC Crit Rev Toxicol 2: 299-363.
http://dx.doi.org/10.3109/1040844730908202Q
Alarie. Y. (1981). Toxicological evaluation of airborne chemical irritants and allergens using respiratory reflex
reactions. InBKJ Leong (Ed.), Inhalation toxicology and technology (pp. 207-231). Ann Arbor, MI: Ann
Arbor Science Publishers, Inc.
Alarie. Y: Luo. JE. (1986). Sensory initiation by airborne chemicals: A basis to establish acceptable levels of
exposure. In CS Barrow (Ed.), Toxicology of the nasal passages (pp. 91-100). Washington, DC:
Hemisphere Publishing Corporation.
Alexander. B: Park. RJ: Jacob. DJ: Gong. S. (2009). Transition metal-catalyzed oxidation of atmospheric sulfur:
Global implications for the sulfur budget. J Geophys Res Atmos 114.
http://dx.doi.org/10.1029/2008JD01Q486
Alhanti. BA: Chang. HH: Winquist. A: Mulholland. JA: Darrow. LA: Sarnat. SE. (2016). Ambient air pollution and
emergency department visits for asthma: A multi-city assessment of effect modification by age. J Expo Sci
Environ Epidemiol 26: 180-188. http://dx.doi.org/10.1038/ies.2015.57
Allen. RW: Adar. SD: Avol. E: Cohen. M: Curl. CL: Larsoa T: Liu. LJ: Sheppard. L: Kaufmaa JD. (2012).
Modeling the residential infiltration of outdoor PM(2.5) in the Multi-Ethnic Study of Atherosclerosis and
Air Pollution (MESA Air). Environ Health Perspect 120: 824-830. http://dx.doi.org/10.1289/ehp. 1104447
R-l
-------
Almgvist. C: Worm. M: Levnaert. B. (2008). Impact of gender on asthma in childhood and adolescence: A
GA2LEN review [Review], Allergy 63: 47-57. http://dx.doi.org/10.1111/i. 1398-9995.2007.01524.X
Altug. H: Gaga. EO: Dogeroglu. T: Brunekreef. B: Hoek. G: Van Doom. W. (2014). Effects of ambient air pollution
on respiratory tract complaints and airway inflammation in primary school children. Sci Total Environ 479-
480: 201-209. htto://dx.doi.org/10.1016/i.scitotenv.2014.01.127
Altug. H: Gaga. EO: Dogeroglu. T: Ozdea O: Ornektekin. S: Brunekreef. B: Meliefste. K: Hoek. G: Van Doom. W.
(2013). Effects of air pollution on lung function and symptoms of asthma, rhinitis and eczema in primary
school children. Environ Sci Pollut Res Int 20: 6455-6467. http://dx.doi.org/10.1007/sll356-Q13-1674-l
Amadeo. B: Robert. C: Rondeau. V: Mounouchv. MA: Cordeau. L: Birembaux. X: Citadelle. E: Gotin. J:
Gouranton. M: Marcia G: Laurac. D: Raherison. C. (2015). Impact of close-proximity air pollution on lung
function in schoolchildren in the French West Indies. BMC Public Health 15: 45.
http://dx.doi.org/10.1186/sl2889-015-1382-5
Amdur. MO: Chen. LC: Gutv. J: Lam. HF: Miller. PP. (1988). Speciation and pulmonary effects of acidic SOx
formed on the surface of ultrafine zinc oxide aerosols. Atmos Environ 22: 557-560.
http://dx.doi.org/10.1016/0004-6981(88)90199-0
Amdur. MO: McCarthy. JF: Gill. MW. (1983). Effect of mixing conditions on irritant potency of zinc oxide and
sulfur dioxide. Am Ind Hyg Assoc J 44: 7-13. http://dx.doi.org/10.1080/15298668391404284
Amdur. MO: Melvin. WW. Jr: Drinker. P. (1953). Effects of inhalation of sulphur dioxide by man. Lancet 265: 758-
759. http://dx.doi.org/10.1016/S0140-6736(53)91455-X
American Heart Association. (2011). Cardiovascular disease statistics. Available online at
http://web.archive.org/web/20110410093532/http://www.americanheart.org/presenter.ihtml?identifier=447
8 (accessed April 12, 2011).
Ancona. C: Badaloni. C: Mataloni. F: Bolignano. A: Bucci. S: Cesaroni. G: Sozzi. R: Davoli. M: Forastiere. F.
(2015). Mortality and morbidity in a population exposed to multiple sources of air pollution: A
retrospective cohort study using air dispersion models. Environ Res 137: 467-474.
http://dx.doi.Org/10.1016/i.envres.2014.10.036
Andersen. I: Lundavist. GR: Jensea PL: Proctor. DF. (1974). Human response to controlled levels of sulfur dioxide.
Arch Environ Occup Health 28: 31-39. http://dx.doi.org/10.1080/00Q39896.1974.10666429
Anderson. HR: Armstrong. B: Haiat. S: Harrison. R: Monk. V: Poloniecki. J: Timmis. A: Wilkinson. P. (2010). Air
pollution and activation of implantable cardioverter defibrillators in London. Epidemiology 21: 405-413.
http://dx.doi.org/10.1097/EDE.0b013e3181d6160Q
Anderson. HR: Bremner. SA: Atkinson. RW: Harrisoa RM: Walters. S. (2001). Particulate matter and daily
mortality and hospital admissions in the West Midlands conurbation of the United Kingdom: associations
with fine and coarse particles, black smoke and sulphate. Occup Environ Med 58: 504-510.
http://dx.doi.Org/10.1136/oem.58.8.504
Andersson. E: Knutssoa A: Hagberg. S: Nilsson. T: Karlsson. B: Alfredssoa L: Torea K. (2006). Incidence of
asthma among workers exposed to sulphur dioxide and other irritant gases. Eur Respir J 27: 720-725.
http://dx.doi.org/10.1183/09031936.06.000343Q5
Annesi-Maesano. I: Rouve. S: Desauevroux. H: Jankovski. R: Klossek. JM: Thibaudon. M: Demolv. P: Didier. A.
(2012). Grass pollen counts, air pollution levels and allergic rhinitis severity. Int Arch Allergy Immunol
158: 397-404. http://dx.doi.org/10.1159/000332964
Anvenda. EO: Higashi. T: Kambavashi. Y: Nguyen. TT: Michigami. Y: Fuiimura. M: Hara. J: Tsuiiguchi. H:
Kitaoka. M: Asakura. H: Hori. D: Yamada. Y: Havashi. K: Havakawa. K: Nakamura. H. (2016).
Associations of cough prevalence with ambient polycyclic aromatic hydrocarbons, nitrogen and sulphur
dioxide: A longitudinal study. Int J Environ Res Public Health 13: 800.
http://dx.doi.org/10.3390/iierphl308080Q
R-2
-------
Arbex. MA: de Souza Conceicao. GM: Cendoa SP: Arbex. FF: Lopes. AC: Movses. EP: Santiago. SL: Saldiva.
PHN: Pereira. LAA: Braga. ALF. (2009). Urban air pollution and chronic obstructive pulmonary disease-
related emergency department visits. J Epidemiol Community Health 63: 777-783.
http://dx.doi.org/10.1136/iech.2008.07836Q
Armstrong. B. (2008). Measurement error: Consequences and design issues. In D Baker; MJ Nieuwenhuijsen (Eds.),
Environmental epidemiology: Study methods and application (1st ed., pp. 93-112). New York, NY: Oxford
University Press.
Armstrong. BK: White. E: Saracci. R. (1992). Principles of exposure measurement in epidemiology. New York,
NY: Oxford University Press.
Arnedo-Pena. A: Garcia-Marcos. L: Carvaial Uruefia. I: Busauets Monge. R: Morales Suarez-Varela. M: Miner
Canflanca. I: Batlles Garrido. J: Blanco Ouiros. A: Lopez-Silvarrev Varela. A: Garcia Hernandez. G:
Aguinaga Ontoso. I: Gonzalez Diaz. C. (2009). Air pollution and recent symptoms of asthma, allergic
rhinitis, and atopic eczema in schoolchildren aged between 6 and 7 years. Arch Bronconeumol 45: 224-
229. http://dx.doi.Org/10.1016/S 1579-2129(09)72152-4
Arts. JH: de Heer. C: Woutersen. RA. (2006). Local effects in the respiratory tract: relevance of subjectively
measured irritation for setting occupational exposure limits [Review]. Int Arch Occup Environ Health 79:
283-298. http://dx.doi.org/10.1007/s00420-005-0Q44-9
Assibev-Mensah. V: Liu. K: Thurstoa SW: Stevens. TP: Zhang. J: Zhang. J: Kane. C: Pan. Y: Weinberger. B:
Ohman-Strickland. P: Woodruff. T: Rich. DO. (2015). Impact of the 2008 Beijing Olympics on the risk of
pregnancy complications. Arch Environ Occup Health 71: 208-215.
http://dx.doi.org/10.1080/19338244.2Q15.1058236
Atabi. F: Jafarigol. F: Moattar. F: Nouri. J. (2016). Comparison of AERMOD and CALPUFF models for simulating
S02 concentrations in a gas refinery. Environ Monit Assess 188: 516. http://dx.doi.org/10.1007/s 10661 -
016-5508-8
Atari. DO: Luginaah. I: Xu. X: Fung. K. (2008). Spatial variability of ambient nitrogen dioxide and sulfur dioxide in
Sarnia, "Chemical Valley," Ontario, Canada. J Toxicol Environ Health A 71: 1572-1581.
http://dx.doi.org/10.1080/152873908Q2414158
Atari. DO: Luginaah. IN: Fung. K. (2009). The relationship between odour annoyance scores and modelled ambient
air pollution in Sarnia, "Chemical Valley", Ontario. Int J Environ Res Public Health 6: 2655-2675.
http://dx.doi.org/10.3390/iierph6102655
Atkinsoa RW: Bremner. SA: Anderson. HR: Strachaa DP: Bland. JM: Ponce de Leon. A. (1999). Short-term
associations between emergency hospital admissions for respiratory and cardiovascular disease and outdoor
air pollution in London. Arch Environ Health 54: 398-411. http://dx.doi.org/10.1080/000398999096Q3371
Atkinson. RW: Carey. IM: Kent. AJ: van Staa. TP: Anderson. HR: Cook. DG. (2013). Long-term exposure to
outdoor air pollution and incidence of cardiovascular diseases. Epidemiology 24: 44-53.
http://dx.doi.org/10.1097/EDE.0b013e318276ccb8
Atkinson. RW: Carey. IM: Kent. AJ: van Staa. TP: Andersoa HR: Cook. DG. (2015). Long-term exposure to
outdoor air pollution and the incidence of chronic obstructive pulmonary disease in a national English
cohort. Occup Environ Med 72: 42-48. http://dx.doi.org/10.1136/oemed-2014-102266
Atkinson. RW: Cohen. A: Mehta. S: Andersoa HR. (2012). Systematic review and meta-analysis of
epidemiological time-series studies on outdoor air pollution and health in Asia. Air Qual Atmos Health 5:
383-391. http://dx.doi.org/10.1007/sll869-010-Q123-2
ATS (American Thoracic Society). (2000a). Guidelines for methacholine and exercise challenge testing-1999. Am J
Respir Crit Care Med 161: 309-329. http://dx.doi.Org/10.1164/airccm.161.l.atsll-99
ATS (American Thoracic Society). (2000b). What constitutes an adverse health effect of air pollution? Am J Respir
Crit Care Med 161: 665-673. http://dx.doi.Org/10.1164/airccm.161.2.ats4-00
R-3
-------
ATSDR (Agency for Toxic Substances and Disease Registry). (2006). A study of ambient air contaminants and
asthma in New York City: Part A and B. Atlanta, GA: U.S. Department of Health and Human Services.
http://permanent.access.gpo.gov/lps88357/ASTHMA BRONX FINAL REPORT.pdf
Atzori. L: Bannenberg. G: Corriga. AM: Lou. YP: Lundberg. JM: Rvrfeldt. A: Moldeus. P. (1992). Sulfur dioxide -
induced bronchoconstriction via ruthenium red-sensitive activation of sensory nerves. Respiration 59: 272-
278. http://dx.doi.org/10.1159/000196Q72
Avital. A: Noviski. N: Bar-Yishav. E: Springer. C: Lew. M: Godfrey. S. (1991). Nonspecific bronchial reactivity in
asthmatic children depends on severity but not on age. Am Rev Respir Dis 144: 36-38.
http://dx.doi.org/10.1164/airccm/144.1.36
Baccarelli. A: Zanobetti. A: Martinelli. I: Grillo. P: Hou. L: Giacomini. S: Bonzini. M: Lanzani. G: Mannucci. PM:
Bertazzi. PA: Schwartz. J. (2007a). Effects of exposure to air pollution on blood coagulation. J Thromb
Haemost 5: 252-260. http://dx.doi.Org/10.llll/i.1538-7836.2007.02300.x
Baccarelli. A: Zanobetti. A: Martinelli. I: Grillo. P: Hou. L: Lanzani. G: Mannucci. PM: Bertazzi. PA: Schwartz. J.
(2007b). Air pollution, smoking, and plasma homocysteine. Environ Health Perspect 115: 176-181.
http://dx.doi.org/10.1289/ehp.9517
Bai. J: Meng. Z. (2005). Effects of sulfur dioxide on apoptosis-related gene expressions in lungs from rats. Regul
Toxicol Pharmacol 43: 272-279. http://dx.doi.Org/10.1016/i.vrtph.2005.09.002
Baia. ES: Schwartz. JD: Wellenius. GA: Coull. BA: Zanobetti. A: Vokonas. PS: Suh. HH. (2010). Traffic-related air
pollution and QT interval: Modification by diabetes, obesity, and oxidative stress gene polymorphisms in
the Normative Aging Study. Environ Health Perspect 118: 840-846. http://dx.doi.org/10.1289/ehp.0901396
Balazv. M: Abu-Yousef. IA: Harpp. DN: Park. J. (2003). Identification of carbonyl sulfide and sulfur dioxide in
porcine coronary artery by gas chromatography/mass spectrometry, possible relevance to EDHF. Biochem
Biophys Res Commun 311: 728-734. http://dx.doi.Org/10.1016/i.bbrc.2003.10.055
Balchum. OJ: Dvbicki. J: Meneelv. GR. (1959). Absorption and distribution of S3502 inhaled through the nose and
mouth by dogs. Am J Physiol 197: 1317-1321.
Balchum. OJ: Dvbicki. J: Meneelv. GR. (1960). The dynamics of sulfur dioxide inhalation: Absorption, distribution,
and retention. AMA Arch Ind Health 21: 564-569.
Baldasano. JM: Soret. A: Guevara. M: Martinez. F: Gasso. S. (2014). Integrated assessment of air pollution using
observations and modelling in Santa Cruz de Tenerife (Canary Islands). Sci Total Environ 473-474: 576-
588. http://dx.doi.Org/10.1016/i.scitotenv.2013.12.062
Ballester. F: Rodriguez. P: Iniguez. C: Saez. M: Daponte. A: Galaa I: Taracido. M: Arribas. F: Bellido. J: Cirarda.
FB: Canada. A: Guillea JJ: Guillen-Grima. F: Lopez. E: Perez-Hovos. S: Lertxundi. A: Toro. S. (2006).
Air pollution and cardiovascular admisisons association in Spain: Results within the EMECAS project. J
Epidemiol Community Health 60: 328-336. http://dx.doi.org/10.1136/iech.2005.037978
Ballester. F: Tenias. JM: Perez-Hovos. S. (2001). Air pollution and emergency hospital admissions for
cardiovascular diseases in Valencia, Spain. J Epidemiol Community Health 55: 57-65.
http://dx.doi.org/10.1136/iech.55.1.57
Balmes. JR: Fine. JM: Sheppard. D. (1987). Symptomatic bronchoconstriction after short-term inhalation of sulfur
dioxide. Am J Respir Crit Care Med 136: 1117-1121. http://dx.doi.org/10.1164/airccm/136.5.1117
Baneriee. T: Singh. SB: Srivastava. RK. (2011). Development and performance evaluation of statistical models
correlating air pollutants and meteorological variables at Pantnagar, India. Atmos Res 99: 505-517.
http://dx.doi.Org/10.1016/i.atmosres.2010.12.003
Bannenberg. G: Atzori. L: Xue. J: Auberson. S: Kimland. M: Rvrfeldt. A: Lundberg. JM: Moldeus. P. (1994). Sulfur
dioxide and sodium metabisulfite induce bronchoconstriction in the isolated perfused and ventilated guinea
pig lung via stimulation of capsaicin-sensitive sensory nerves. Respiration 61: 130-137.
http://dx.doi.org/10.1159/00Q196324
Barnes. I: Hiorth. J: Mihalopoulos. N. (2006). Dimethyl sulfide and dimethyl sulfoxide and their oxidation in the
atmosphere [Review]. ChemRev 106: 940-975. http://dx.doi.org/10.102l/cr020529+
R-4
-------
Barr. RG: Herbstmaa J: Speizer. FE: Camargo. CA. Jr. (2002). Validation of self-reported chronic obstructive
pulmonary disease in a cohort study of nurses. Am J Epidemiol 155: 965-971.
http://dx.doi.org/10.1093/aie/155.lQ.965
Barrett. EG: Day. KC: Gigliotti. AP: Reed. MP: McDonald. JD: Mauderlv. JL: Seilkop. SK. (2011). Effects of
simulated downwind coal combustion emissions on pre-existing allergic airway responses in mice. Inhal
Toxicol 23: 792-804. http://dx.doi.org/10.3109/08958378.2011.6Q9917
Barthelemv. P: Badier. M: Jammes. Y. (1988). Interaction between S02 and cold-induced bronchospasm in
anesthetized rabbits. Respir Physiol 71: 1-10. http://dx.doi.org/10.1016/0034-5687(88)90110-7
Baskurt. OK. (1988). Acute hematologic and hemorheologic effects of sulfur dioxide inhalation. Arch Environ
Occup Health 43: 344-348. http://dx.doi.org/10.1080/00039896.1988.9934946
Bateson. TF: Coull. BA: Hubbell. B: Ito. K: Jerrett. M: Lumlev. T: Thomas. D: Vedal. S: Ross. M. (2007). Panel
discussion review: Session three - issues involved in interpretation of epidemiologic analyses - statistical
modeling. J Expo Sci Environ Epidemiol 17: S90-S96. http://dx.doi.org/10.1038/si.ies.7500631
Baur. X: Bakehe. P: Vellguth. H. (2012). Bronchial asthma and COPD due to irritants in the workplace - an
evidence-based approach. J Occup Med Toxicol 7: 19. http://dx.doi.org/10.1186/1745-6673-7-19
Baxter. LK: Sacks. JD. (2014). Clustering cities with similar fine particulate matter exposure characteristics based
on residential infiltration and in-vehicle commuting factors. Sci Total Environ 470-471: 631-638.
http://dx.doi.Org/10.1016/i.scitotenv.2013.10.019
Becauemin. MM: Bertholoa JF: Bouchikhi. A: Malarbet. JL: Roy. M. (1999). Oronasal ventilation partitioning in
adults and children: Effect on aerosol deposition in airways. Radiat Prot Dosimetry 81: 221-228.
Bedi. JF: Folinsbee. LJ: Horvath. SM: Ebensteia RS. (1979). Human exposure to sulfur dioxide and ozone: absence
of a synergistic effect. Arch Environ Occup Health 34: 233-239.
http://dx.doi.org/10.1080/00039896.1979.106674Q5
Beelea R: Hoek. G: Fischer. P: van den Brandt. PA: Brunekreef. B. (2007). Estimated long-term outdoor air
pollution concentrations in a cohort study. Atmos Environ 41: 13431358.
http://dx.doi.Org/10.1016/i.atmosenv.2006.10.020
Beelea R: Hoek. G: van den Brandt. PA: Goldbohm. RA: Fischer. P: Schouten. LJ: Armstrong. B: Brunekreef. B.
(2008a). Long-term exposure to traffic-related air pollution and lung cancer risk. Epidemiology 19: 702-
710. http://dx.doi.org/10.1097/EDE.Qb013e318181b3ca
Beelea R: Hoek. G: van den Brandt. PA: Goldbohm. RA: Fischer. P: Schouten. LJ: Jerrett. M: Hughes. E:
Armstrong. B: Brunekreef. B. (2008b). Long-term effects of traffic-related air pollution on mortality in a
Dutch cohort (NLCS-AIR study). Environ Health Perspect 116: 196-202.
http://dx.doi.org/10.1289/ehp. 10767
Beirle. S: Hormann. C: Penning de Vries. M: Dorner. S: Kern. C: Wagner. T. (2013). Estimating the volcanic
emission rate and atmospheric lifetime of S02 from space: a case study for Kilauea volcano, Hawai'i.
Atmos ChemPhys Discuss 13: 28695-28727. http://dx.doi.org/10.5194/acpd-13-28695-2013
Bell. ML: Ebisu. K: Belanger. K. (2007). Ambient air pollution and low birth weight in Connecticut and
Massachusetts. Environ Health Perspect 115: 1118-1124. http://dx.doi.org/10.1289/ehp.9759
Bell. ML: Lew. JK: Lin. Z. (2008). The effect of sandstorms and air pollution on cause-specific hospital admissions
in Taipei, Taiwan. Occup Environ Med 65: 104-111. http://dx.doi.org/10.1136/oem.2006.031500
Bellini. P: Baccini. M: Biggeri. A: Terracini. B. (2007). The meta-analysis of the Italian studies on short-term effects
of air pollution (MISA): Old and new issues on the interpretation of the statistical evidences.
Environmetrics 18: 219-229. http://dx.doi.org/10.10Q2/env.836
Bennett. O: Kandala. NB: Ji. C: Linnane. J: Clarke. A. (2014). Spatial variation of heart failure and air pollution in
Warwickshire, UK: an investigation of small scale variation at the ward-level. BMJ Open 4: e006028.
http://dx.doi.org/10.1136/bmiopen-2014-006Q28
R-5
-------
Bennett. W: Zemaa K: Jarabek. A. (2003). Nasal contribution to breathing with exercise: Effect of race and gender.
J Appl Physiol (1985) 95: 497-503. http://dx.doi.org/10.1152/iapplphYsiol.00718.2002
Bennett. WD: Zemaa KL: Jarabek. AM. (2008). Nasal contribution to breathing and fine particle deposition in
children versus adults. J Toxicol Environ Health A 71: 227-237.
http://dx.doi.org/10.1080/1528739070159820Q
Bentaveb. M: Wagner. V: Stempfelet. M: Zins. M: Goldberg. M: Pascal. M: Larrieu. S: Beaudeau. P: Cassadou. S:
Eilstein. D: Filleul. L: Le Tertre. A: Medina. S: Pascal. L: Prouvost. H: Ouenel. P: Zeghnoun. A: Lefranc.
A. (2015). Association between long-term exposure to air pollution and mortality in France: A 25-year
follow-up study. Environ Int 85: 5-14. http://dx.doi.Org/10.1016/i.envint.2015.08.006
Bergea S: Szpiro. AA. (2015). Mitigating the impact of measurement error when using penalized regression to
model exposure in two-stage air pollution epidemiology studies. Environ Ecol Stat 22: 601-631.
http://dx.doi.org/10.1007/slQ651-015-0314-v
Berglind. N: Bellander. T: Forastiere. F: vonKlot. S: Aalto. P: Elosua. R: Kulmala. M: Lanki. T: Lowel. H: Peters.
A: Picciotto. S: Salomaa. V: Stafoggia. M: Sunver. J: Nvberg. F. (2009). Ambient air pollution and daily
mortality among survivors of myocardial infarction. Epidemiology 20: 110-118.
http://dx.doi.org/10.1097/EDE.0b013e3181878b5Q
Berndt. T: Jokinen. T: Mauldia R. Ill: Petaia. T: Herrmana H: Junninen. H: Paasonea P: Worsnop. PR: Sipila. M.
(2012). Gas-phase ozonolysis of selected olefins: The yield of stabilized criegee intermediate and the
reactivity toward S02. J Phys Chem Lett 3: 2892-2896. http://dx.doi.org/10.1021/iz301158u
Bethel. RA: Epstein. J: Sheppard. D: Nadel. JA: Boushev. HA. (1983). Sulfur dioxide-induced bronchoconstriction
in freely breathing, exercising, asthmatic subjects. Am Rev Respir Dis 128: 987-990.
Bethel. RA: Sheppard. D: Epsteia J: Tam. E: Nadel. JA: Boushev. HA. (1984). Interaction of sulfur dioxide and dry
cold air in causing bronchoconstriction in asthmatic subjects. J Appl Physiol (1985) 57: 419-423.
Bethel. RA: Sheppard. D: Geffrov. B: Tam E: Nadel. JA: Boushev. HA. (1985). Effect of 0.25 ppm sulfur dioxide
on airway resistance in freely breathing, heavily exercising, asthmatic subjects. Am Rev Respir Dis 131:
659-661.
Bhaskaran. K: Haiat. S: Armstrong. B: Haines. A: Herrett. E: Wilkinson. P: Smeeth. L. (2011). The effects of hourly
differences in air pollution on the risk of myocardial infarction: case crossover analysis of the MINAP
database. B M J 343: d5531. http://dx.doi.org/10.1136/bmi.d5531
Bhattacharwa. N: Shapiro. NL. (2010). Air quality improvement and the prevalence of frequent ear infections in
children. Otolaryngol Head Neck Surg 142: 242-246. http://dx.doi.Org/10.1016/i.otohns.2009.10.052
Bigbv. B: Boushev. H. (1993). Effects of nedocromil sodium on the bronchomotor response to sulfur dioxide in
asthmatic patients. J Allergy Clin Immunol 92: 195-197. http://dx.doi.org/10.1016/0091-6749(93)90106-P
Biggeri. A: Baccini. M: Bellini. P: Terracini. B. (2005). Meta-analysis of the Italian studies of short-term effects of
air pollution (MISA), 1990-1999. Int J Occup Environ Health 11: 107-122.
http://dx.doi.org/10.1179/oeh.2005.11.1.107
Bishaw. A. (2012). Poverty: 2010 and 2011. American community survey briefs. Washington, DC: U.S. Department
of Commerce, U.S. Census Bureau. http://www.census.gov/prod/2012pubs/acsbrl 1-01 .pdf
Blackwell. PL: Lucas. JW: Clarke. TC. (2014). Summary health statistics for U.S. adults: National Health Interview
Survey, 2012. (Vital and Health Statistics: Series 10, No. 260). Hyattsville, MD: National Center for Health
Statistics, http://www.cdc.gov/nchs/data/series/sr 10/srlQ 260.pdf
Blanchard. CL: Hidv. GM: Tanenbaum. S: Edgerton. ES: Hartsell. BE. (2013). The Southeastern Aerosol Research
and Characterization (SEARCH) study: Temporal trends in gas and PM concentrations and composition,
1999-2010. J Air Waste Manag Assoc 63: 247-259. http://dx.doi.org/10.1080/10962247.2Q12.748523
Bloom. B: Cohen. RA: Freemaa G. (2012). Summary health statistics for u.s. Children: national health interview
survey, 2011. Vital Health Stat 101-88.
R-6
-------
Bluett. J: Gimson. N: Fisher. G: Hevdenrvch. C: Freeman. T: Godfrey. J. (2004). Good practice guide for
atmospheric dispersion modelling: 2. Which dispersion model to use? In E New Zealand Ministry for the
(Ed.), Good practice guide for atmospheric dispersion modelling (pp. 8-21). Wellington, New Zealand:
Ministry for the Environment, http://www.mfe.govt.nz/publications/air/atmospheric-dispersion-modelling-
iun04/html/page5. html
Bobak. M. (2000). Outdoor air pollution, low birth weight, and prematurity. Environ Health Perspect 108: 173-176.
http://dx.doi.org/10.2307/3454517
Bobrowski. N: Kern. C: Piatt. U: Hormann. C: Wagner. T. (2010). Novel S02 spectral evaluation scheme using the
360390 nm wavelength range. Atmos Meas Tech 3: 879-891. http://dx.doi.org/10.5194/amt-3-879-2010
Boezea HM: Van Per Zee. SC: Postma. DS: Vonk. JM: Gerritsea J: Hoek. G: Brunekreef. B: Riicken. B:
Schouten. JP. (1999). Effects of ambient air pollution on upper and lower respiratory symptoms and peak
expiratory flow in children [Comment]. Lancet 353: 874-878. http://dx.doi.org/10.1016/S014Q-
6736(98)06311-9
Boezea HM: Vonk. JM: Van Per Zee. SC: Gerritsea J: Hoek. G: Brunekreef. B: Schouten. JP: Postma. DS. (2005).
Susceptibility to air pollution in elderly males and females. Eur Respir J 25: 1018-1024.
http://dx.doi.org/10.1183/09031936.05.000761Q4
Bogumil. K: Orphal. J: Homana T: Voigt. S: Spietz. P: Fleischmann. OC: Vogel. A: Hartmann. M: Kromminga. H:
Bovensmana H: Frerick. J: Burrows. JP. (2003). Measurements of molecular absorption spectra with the
SCIAMACHY pre-flight model: instrument characterization and reference data for atmospheric remote-
sensing in the 230-2380 nm region. J Photochem Photobiol A 157: 167-184.
http://dx.doi.Org/10.1016/S 1010-6030(03)00062-5
Borrell. LN: Nguyen. EA: Roth. LA: Oh. SS: Tcheurekdiiaa H: Sen. S: Davis. A: Farber. HJ: Avila. PC: Brigino-
Buenaventura. E: Lenoir. MA: Lurmann. F: Meade. K: Serebriskv. D: Rodriguez-Cintroa W: Kumar. R:
Rodriguez-Santana. JR: Thvne. SM: Burchard. EG. (2013). Childhood obesity and asthma control in the
GALA II and SAGE II studies. Am J Respir Crit Care Med 187: 697-702.
http://dx.doi.org/10.1164/rccm.201211-2116QC
Bovnard. A: Clerbaux. C: Clarisse. L: Safieddine. S: Pommier. M: Van Damme. M: Bauduia S: Oudot. C: Hadii-
Lazaro. J: Hurtmans. D: Coheur. PF. (2014). First simultaneous space measurements of atmospheric
pollutants in the boundary layer from IASI: A case study in the North China Plain. Geophys Res Lett 41:
645-651. http://dx.doi.org/10.1002/2013GLQ58333
Bozkurt. Z: Dogan. G: Arslanbas. D: Pekev. B: Pekev. H: Dumanoglu. Y: Bavram. A: Tuncel. G. (2015).
Determination of the personal, indoor and outdoor exposure levels of inorganic gaseous pollutants in
different microenvironments in an industrial city. Environ Monit Assess 187: 590.
http://dx.doi.org/10.1007/slQ661-015-4816-8
Brain. JD. (1970). The uptake of inhaled gases by the nose. Ann Otol Rhinol Laryngol 79: 529-539.
http://dx.doi.org/10.1177/00034894700790Q315
Brand. A: Mcleaa KE: Henderson. SB: Fournier. M: Liu. L: Kosatskv. T: Smargiassi. A. (2016). Respiratory
hospital admissions in young children living near metal smelters, pulp mills and oil refineries in two
Canadian provinces. Environ Int 94: 24-32. http://dx.doi.Org/10.1016/i.envint.2016.05.002
Brauer. M: Koutrakis. P: Spengler. J. (1989). Personal exposures to acidic aerosols and gases. Environ Sci Technol
23: 1408-1412. http://dx.doi.org/10.1021/esQ0069a013
Brauer. M: Lencar. C: Tamburic. L: Koehoora M: Demers. P: Karr. C. (2008). A cohort study of traffic-related air
pollution impacts on birth outcomes. Environ Health Perspect 116: 680-686.
http://dx.doi.org/10.1289/ehp. 10952
Bravo. MA: Son. J: de Freitas. CU: Gouveia. N: Bell. ML. (2015). Air pollution and mortality in Sao Paulo, Brazil:
Effects of multiple pollutants and analysis of susceptible populations. J Expo Sci Environ Epidemiol 26:
150-161. http://dx.doi.org/10.1038/ies.2014.9Q
R-7
-------
Breen. MS: Long. TC: Schultz. BP: Crooks. J: Breea M: Langstaff. JE: Isaacs. KK: Taa YM; Williams. RW: Cao.
Y: Geller. AM: Devlin. RB: Battermaa SA: Buckley. TJ. (2014a). GPS-based microenvironment tracker
(MicroTrac) model to estimate time-location of individuals for air pollution exposure assessments: Model
evaluation in central North Carolina. J Expo Sci Environ Epidemiol 24: 412-420.
http://dx.doi.org/10.1038/ies.2Q14.13
Breen. MS: Schultz. BP: Soha MP: Long. T: Langstaff. J: Williams. R: Isaacs. K: Meng. OY: Stallings. C: Smith.
L. (2014b). A review of air exchange rate models for air pollution exposure assessments [Review]. J Expo
Sci Environ Epidemiol 24: 555-563. http://dx.doi.org/10.1038/ies.2013.3Q
Briggs. GA. (1993). Plume dispersion in the convective boundary layer. Part II: Analysis of CONDORS field
experiment data. J Appl Meteorol 32: 1388-1425. http://dx.doi.org/10.1175/152Q-
0450(1993)032<1388:PDITCB>2.0.CQ:2
Brimblecombe. P. (2003). The global sulfur cycle. In HD Holland; KK Turekian (Eds.), Treatise on geochemistry
(Second editions) (2nd ed., pp. 559-591). Amsterdam, The Netherlands: Elsevier Inc.
http://dx.doi.org/10.1016/B978-0-08-095975-7.0Q814-7
Brochu. P: Bouchard. M: Haddad. S. (2014). Physiological daily inhalation rates for health risk assessment in
overweight/obese children, adults, and elderly. Risk Anal 34: 567-582. http://dx.doi.org/10. Ill 1/risa. 12125
Brochu. P: Brodeur. J: Krishnan. K. (2011). Derivation of physiological inhalation rates in children, adults, and
elderly based on nighttime and daytime respiratory parameters. Inhal Toxicol 23: 74-94.
http://dx.doi.org/10.3109/08958378.201Q.543439
Brodv. DJ: Zhang. X: Kit. BK: Dilloa CF. (2013). Reference values and factors associated with exhaled nitric
oxide: U.S. youth and adults. Respir Med 107: 1682-1691. http://dx.doi.Org/10.1016/i.rmed.2013.07.006
Brook. RD: Kousha. T. (2015). Air pollution and emergency department visits for hypertension in Edmonton and
Calgary, Canada: A case-crossover study. Am J Hypertens 28: 1121-1126.
http://dx.doi.org/10.1093/aih/hpu302
Browa JS. (2015). Nitrogen dioxide exposure and airway responsiveness in individuals with asthma. Inhal Toxicol
27: 1-14. http://dx.doi.org/10.3109/08958378.2014.97996Q
Brown. KW: Sarnat. JA: Suh. HH: Coull. BA: Koutrakis. P. (2009). Factors influencing relationships between
personal and ambient concentrations of gaseous and particulate pollutants. Sci Total Environ 407: 3754-
3765. http://dx.doi.Org/10.1016/i.scitotenv.2009.02.016
Brunekreef. B: Beelen. R: Hoek. G: Schouten. L: Bausch-Goldbohm. S: Fischer. P: Armstrong. B: Hughes. E:
Jerrett. M: van den Brandt. P. (2009). Effects of long-term exposure to traffic-related air pollution on
respiratory and cardiovascular mortality in the Netherlands: The NLCS-AIR Study. (139). Boston, MA:
Health Effects Institute. http://www.n65.nl/NCLS-AIR-Studv-2009.pdf
Briiske. I: Hampel. R: Baumgartner. Z: Rtickerl. R: Grevea S: Koenig. W: Peters. A: Schneider. A. (2011). Ambient
air pollution and lipoprotein-associated phospholipase A2 in survivors of myocardial infarction. Environ
Health Perspect 119: 921-926. http://dx.doi.org/10.1289/ehp. 1002681
Burke. JM: Zufall. MJ: Ozkavnak. H. (2001). A population exposure model for particulate matter: Case study results
forPM2.5 in Philadelphia, PA. J Expo Anal Environ Epidemiol 11: 470-489.
http://dx.doi.org/10.1038/si.iea.7500188
Burnett. RT: Brook. J: Dana T: Delocla. C: Philips. O: Cakmak. S: Vincent. R: Goldberg. MS: Krewski. D. (2000).
Association between particulate- and gas-phase components of urban air pollution and daily mortality in
eight Canadian cities. Inhal Toxicol 12: 15-39. http://dx.doi.org/10.1080/0895837005Q164851
Burnett. RT: Cakmak. S: Brook. JR: Krewski. D. (1997). The role of particulate size and chemistry in the
association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
Environ Health Perspect 105: 614-620. http://dx.doi.org/10.1289/ehp.97105614
Burnett. RT: Smith-Doiron. M: Stieb. D: Cakmak. S: Brook. JR. (1999). Effects of particulate and gaseous air
pollution on cardiorespiratory hospitalizations. Arch Environ Health 54: 130-139.
http://dx.doi.org/10.1080/000398999096Q2248
R-8
-------
Burnett. RT: Stieb. D: Brook. JR: Cakmak. S: Dales. R: Raizenne. M: Vincent. R: Dana T. (2004). Associations
between short-term changes in nitrogen dioxide and mortality in Canadian cities. Arch Environ Occup
Health 59: 228-236. http://dx.doi.Org/10.3200/AEQH.59.5.228-236
Burnev. PG: Laitinen. LA: Perdrizet. S: Huckauf. H: Tattersfield. AE: China S: Poissoa N: Heeren. A: Britton. JR:
Jones. T. (1989). Validity and repeatability of the IUATLD (1984) Bronchial Symptoms Questionnaire: an
international comparison. Eur Respir J 2: 940-945.
Burr. ML. (1992). Diagnosing asthma by questionnaire in epidemiological surveys [Editorial]. Clin Exp Allergy 22:
509-510. http://dx.doi.org/10.1111/i. 1365-2222.1992.tb00158.x
Burra. TA: Moineddin. R: Agha. MM: Glazier. RH. (2009). Social disadvantage, air pollution, and asthma physician
visits in Toronto, Canada. Environ Res 109: 567-574. http://dx.doi.Org/10.1016/i.envres.2009.03.004
Burrows. B: Sears. MR: Flannerv. EM: Herbisoa GP: Holdawav. MP: Silva. PA. (1995). Relation of the course of
bronchial responsiveness from age 9 to age 15 to allergy. Am J Respir Crit Care Med 152: 1302-1308.
http://dx.doi.Org/10.1164/airccm.152.4.7551386
Burstvn. I: Cherry. NM: Yasui. Y: Kim. HM. (2008). Relative performance of different exposure modeling
approaches for sulfur dioxide concentrations in the air in rural western Canada. BMC Med Res Methodol 8:
43. http://dx.doi.org/10.1186/1471-2288-8-43
Bvers. N: Ritchev. M: Vaidvanathaa A: Brandt. AJ: Yip. F. (2015). Short-term effects of ambient air pollutants on
asthma-related emergency department visits in Indianapolis, Indiana, 2007-2011. J Asthma 53: 1-8.
http://dx.doi.org/10.3109/02770903.2015.10910Q6
Bvua D: Schere. KL. (2006). Review of the governing equations, computational algorithms, and other components
of the models-3 community multiscale air quality (CMAQ) modeling system [Review]. Appl MechRev 59:
51-77. http://dx.doi.org/10.1115/1.2128636
CAA. Clean Air Act, as amended by Pub. L. No. 101-549. section 108: Air quality criteria and control techniques.
42 USC 7408. 42 USC (1990a). http://www.law.cornell.edu/uscode/text/42/7408
CAA (Clean Air Act). (1990b). Clean Air Act, as amended by Pub. L. No. 101-549, section 109: National primary
and secondary ambient air quality standards, 42 USC 7409.
https://www.law.cornell.edu/uscode/text/42/7409
CAA. Clean Air Act, section 302: Definitions. 42 USC 7602. 42 USC (2005).
http://www.gpo.gOv/fdsvs/pkg/USCODE-2005-title42/pdf/USCODE-2005-title42-chap85-subchapIII-
sec7602.pdf
Cakmak. S: Dales. RE: Judek. S. (2006). Respiratory health effects of air pollution gases: Modification by education
and income. Arch Environ Occup Health 61: 5-10. http://dx.doi.Org/10.3200/AEOH.61.l.5-10
Cakmak. S: Hebbera C: Cakmak. JD: Vanos. J. (2016). The modifying effect of socioeconomic status on the
relationship between traffic, air pollution and respiratory health in elementary schoolchildren. J Environ
Manage 177: 1-8. http://dx.doi.Org/10.1016/i.ienvman.2016.03.051
Calkins. WH. (1994). The chemical forms of sulfur in coal - a review. Fuel 73: 475-484.
http://dx.doi.org/10.1016/0016-2361(94)90028-0
Cambra. K: Martinez-Rueda. T: Alonso-Fustel. E: Cirarda. FB: Ibanez. B: Esnaola. S: Calvo. M: Aldasoro. E:
Montova. I. (2011). Mortality in small geographical areas and proximity to air polluting industries in the
Basque Country (Spain). Occup Environ Med 68: 140-147. http://dx.doi.org/10.1136/oem.2009.048215
CAMP Research Group (Childhood Asthma Management Program Research Group). (1999). Recruitment of
participants in the childhood Asthma Management Program (CAMP). I. Description of methods. J Asthma
36: 217-237. http://dx.doi.org/10.3109/027709099090754Q6
Canova. C: Torresaa S: Simonato. L: Scapellato. ML: Tessari. R: Visentin. A: Lotti. M: Maestrelli. P. (2010).
Carbon monoxide pollution is associated with decreased lung function in asthmatic adults. Eur Respir J 35:
266-272. http://dx.doi.org/10.1183/09031936.000437Q9
R-9
-------
Cao. J: Yang. C: Li. J: Chen. R: Chea B: Gu. B: Kaa H. (2011). Association between long-term exposure to
outdoor air pollution and mortality in China: A cohort study. J Hazard Mater 186: 1594-1600.
http://dx.doi.Org/10.1016/i.ihazmat.2010.12.036
Capobussi. M: Tettamanti. R: Marcolin. L: Piovesaa L: Bronzin. S: Gattoni. ME: Polloni. I: Sabatino. G: Tersalvi.
CA: Auxilia. F: Castaldi. S. (2016). Air pollution impact on pregnancy outcomes in Como, Italy. J Occup
Environ Med 58: 47-52. http://dx.doi.org/10.1097/JOM.000000000000063Q
Carey. IM: Atkinson. RW: Kent. AJ: van Staa. T: Cook. DG: Anderson. HR. (2013). Mortality associations with
long-term exposure to outdoor air pollution in a national English cohort. Am J Respir Crit Care Med 187:
1226-1233. http://dx.doi.org/10.1164/rccm.201210-1758QC
Cam. SA: Krueger. AJ: Krotkov. NA: Yang. K: Levelt. PF. (2007). Sulfur dioxide emissions from Peruvian copper
smelters detected by the Ozone Monitoring Instrument. Geophys Res Lett 34: L09801.
http://dx.doi.org/10.1029/2006glQ29020
Carraro. S: Gottardi. G: Bonetto. G: Baraldi. E. (2007). Exhaled nitric oxide in children with asthma and sinusitis
[Review]. Pediatric Allergy and Immunology 18: 28-30. http://dx.doi.org/10.Ill 1/i. 1399-
3038.2007.00629.x
Carruthers. DJ: Davies. BM: Edmunds. HA: Ellis. KL: McHugh. CA: Thomson. DJ. (1995). The Atmospheric
Dispersion Modelling System (ADMS): Comparisons with data from the Kincaid experiment. Int J Environ
Pollut 5: 382-400.
Carson. JL: Brighton. LE: Collier. AM: Bromberg. PA. (2013). Correlative ultrastructural investigations of airway
epithelium following experimental exposure to defined air pollutants and lifestyle exposure to tobacco
smoke. Inhal Toxicol 25: 134-140. http://dx.doi.org/10.3109/08958378.2013.763314
Castro. HA: Cunha. MF: Mendonca. GA: Junger. WL: Cunha-Cruz. J: Leon. AP. (2009). Effect of air pollution on
lung function in schoolchildren in Rio de Janeiro, Brazil. Rev Saude Publica 43: 26-34.
http://dx.doi.org/10.1590/S0034-891020090001000Q4
Cendoa S: Pereira. LA: Braga. AL: Conceicao. GM: Curv Junior. A: Romaldini. H: Lopes. AC: Saldiva. PH.
(2006). Air pollution effects on myocardial infarction. Rev Saude Publica 40: 414-419.
http://dx.doi.org/10.1590/S0034-891020060003000Q8
Chadha. TS: Birch. S: Sackner. MA. (1987). Oronasal distribution of ventilation during exercise in normal subjects
and patients with asthma and rhinitis. Chest 92: 1037-1041. http://dx.doi.org/10.1378/chest.92.6.1037
Chambers. JM: Cleveland. WS: Kleiner. B: Tukev. JA. (1983). Comparing data distributions. In Graphical methods
for data analysis. Belmont, California; Boston, Massachusetts: Wadsworth international Group, Duxbury
Press.
Chan. L: OiHong. D: CuiYun. O: WeiWei. L: Sundell. J. an. (2013). Effects of ambient air pollution on allergic
rhinitis among preschool children in Changsha, China. Chin Sci Bull 58: 4252-4258.
http://dx.doi.org/10.1007/sll434-013-5725-2
Chandra. A: Copen. CE: Stephen. EH. (2013). Infertility and impaired fecundity in the United States, 1982-2010:
data from the National Survey of Family Growth. Atlanta, GA: Centers for Disease Control.
http://www. cdc. gov/nchs/data/nhsr/nhsr067.pdf
Chang. CC: Tsai. SS: Ho. SC: Yang. CY. (2005). Air pollution and hospital admissions for cardiovascular disease in
Taipei, Taiwan. Environ Res 98: 114-119. http://dx.doi.Org/10.1016/i.envres.2004.07.005
Chang. HH: Fuentes. M: Frev. HC. (2012a). Time series analysis of personal exposure to ambient air pollution and
mortality using an exposure simulator. J Expo Sci Environ Epidemiol 22: 483-488.
http://dx.doi.org/10.1038/ies.2012.53
Chang. JC: Hanna. SR. (2004). Air quality model performance evaluation. Meteorol Atmos Phys 87: 167-196.
http://dx.doi.org/10.1007/s00703-003-007Q-7
Chang. JC: Hanna. SR. (2005). Technical descriptions and users guide for the BOOT statistical model evaluation
software package, Version 2.0. Available online at http://www.harmo.org/kit/Download/BOOT UG.pdf
R-10
-------
Chang. LT; Koutrakis. P: Catalano. PJ: Suh. HH. (2000). Hourly personal exposures to fine particles and gaseous
pollutants-Results from Baltimore, Maryland. J Air Waste Manag Assoc 50: 1223-1235.
http://dx.doi.org/10.1080/10473289.2000.10464151
Chang. YK: Wu. CC: Lee. LT: Lin. RS: Yu. YH: Chea YC. (2012b). The short-term effects of air pollution on
adolescent lung function in Taiwan. Chemosphere 87: 26-30.
http://dx.doi.Org/10.1016/i.chemosphere.2011.11.048
Chapmaa RS: Calafiore. DC: Hasselblad. V. (1985). Prevalence of persistent cough and phlegm in young adults in
relation to long-term ambient sulfur oxide exposure. Am Rev Respir Dis 132: 261-267.
Che. WW: Frev. HC: Lau. AK. (2014). Assessment of the effect of population and diary sampling methods on
estimation of school-age children exposure to fine particles. Risk Anal 34: 2066-2079.
http://dx.doi.org/10. Ill 1/risa. 12238
Chea BY: Chan. CC: Lee. CT: Cheng. TJ: Huang. WC: Jhou. JC: Han. YY: Chea CC: Guo. YL. (2012a). The
association of ambient air pollution with airway inflammation in schoolchildren. Am J Epidemiol 175: 764-
774. http://dx.doi.org/10.1093/aie/kwr380
Chen. G: Li. J: Ying. O: Shermaa S: Perkins. N: Raieshwari. S: Mendola. P. (2014a). Evaluation of observation-
fused regional air quality model results for population air pollution exposure estimation. Sci Total Environ
485-486: 563-574. http://dx.doi.Org/10.1016/i.scitotenv.2014.03.107
Chea G: Wan. X. ia: Yang. G: Zou. X. (2015a). Traffic-related air pollution and lung cancer: A meta-analysis. 6:
307-318. http://dx.doi.org/10.1111/1759-7714.12185
Chen. L: Bell. EM: Caton. AR: Druschel. CM: Liq S. (2010a). Residential mobility during pregnancy and the
potential for ambient air pollution exposure misclassification. Environ Res 110: 162-168.
http://dx.doi.Org/10.1016/i.envres.2009.ll.001
Chen. L: Villeneuve. PJ: Rowe. BH: Liu. L: Stieb. DM. (2014b). The Air Quality Health Index as a predictor of
emergency department visits for ischemic stroke in Edmonton, Canada. J Expo Sci Environ Epidemiol 24:
358-364. http://dx.doi.org/10.1038/ies.2013.82
Chen. L: Zhou. Y: Li. S: Williams. G: Kan. H: Marks. GB: Morawska. L: Abramson. MJ: Chen. S: Yao. T: Oiq T:
Wu. S: Guo. Y. (2015b). Air pollution and fasting blood glucose: A longitudinal study in China. Sci Total
Environ 541: 750-755. http://dx.doi.org/10.1016/i.scitotenv.2015.09.132
Chea R: Chu. C: Tan. J: Cao. J: Song. W: Xu. X: Jiang. C: Ma. W: Yang. C: Chen. B: Gui. Y: Kaa H. (2010b).
Ambient air pollution and hospital admission in Shanghai, China. J Hazard Mater 181: 234-240.
http://dx.doi.Org/10.1016/i.ihazmat.2010.05.002
Chea R: Huang. W: Wong. CM: Wang. Z: Thach. TO: Chea B: Kaa H. (2012b). Short-term exposure to sulfur
dioxide and daily mortality in 17 Chinese cities: The China air pollution and health effects study (CAPES).
Environ Res 118: 101-106. http://dx.doi.Org/10.1016/i.envres.2012.07.003
Chea R: Samoli. E: Wong. CM: Huang. W: Wang. Z: Chen. B: Kaa H. (2012c). Associations between short-term
exposure to nitrogen dioxide and mortality in 17 Chinese cities: The China Air Pollution and Health Effects
Study (CAPES). Environ Int 45: 32-38. http://dx.doi.Org/10.1016/i.envint.2012.04.008
Chea R: Zhang. Y: Yang. C: Zhao. Z: Xu. X: Kan. H. (2013). Acute effect of ambient air pollution on stroke
mortality in the china air pollution and health effects study. Stroke 44: 954-960.
http://dx.doi.Org/10.l 161/STROKEAHA. 111.673442
Chea SY: Su. TC: Liq YL: Chan. CC. (2012d). Short-term effects of air pollution on pulse pressure among
nonsmoking adults. Epidemiology 23: 341-348. http://dx.doi.org/10.1097/EDE.0b013e3182452fld
Cheq X: Zhang. LW: Huang. JJ: Song. FJ: Zhang. LP: Oian. ZM: Trevathaq E: Mao. HJ: Haq B: Vaughn. M:
Cheq KX: Liu. YM: Chen. J: Zhao. BX: Jiang. GH: Gu. O: Bai. ZP: Dong. GH: Tang. N.T. (2016). Long-
term exposure to urban air pollution and lung cancer mortality: A 12-year cohort study in Northern China.
Sci Total Environ 571: 855-861. http://dx.doi.Org/10.1016/i.scitotenv.2016.07.064
R-ll
-------
Cheng. MF: Tsai. SS: Yang. CY. (2009). Air pollution and hospital admissions for myocardial infarction in a
tropical city: Kaohsiung, Taiwan. J Toxicol Environ Health A 72: 1135-1140.
http://dx.doi.org/10.1080/15287390903Q91756
Cherniack. R: Adkinsoa NF: Strunk. R: Szefler. S: Tonascia. J: Weiss. S. (1999). The Childhood Asthma
Management Program (CAMP): Design, rationale, and methods. Childhood Asthma Management Program
Research Group. Contr Clin Trials 20: 91-120. http://dx.doi.org/10.1016/S0197-2456(98)00044-0
Chiang. TY: Yuan. TH: Shie. RH: Chea CF: Chan. CC. (2016a). Increased incidence of allergic rhinitis, bronchitis
and asthma, in children living near a petrochemical complex with S02 pollution. Environ Int 96: 1-7.
http://dx.doi.Org/10.1016/i.envint.2016.08.009
Chiang. TY: Yuan. TH: Shie. RH: Chea CF: Chan. CC. (2016b). Increased incidence of allergic rhinitis, bronchitis
and asthma, in children living near a petrochemical complex with S02 pollution - Supplementary data
[Supplemental Data]. Environ Int 96.
Chin. M: Savoie. PL: Huebert. BJ: Bandy. AR: Thornton. DC: Bates. TS: Ouinn. PK: Saltzman. ES: De Bruvn. WJ.
(2000). Atmospheric sulfur cycle simulated in the global model GOCART: Comparison with field
observations and regional budgets. J Geophys Res Atmos 105: 24689-24712.
http://dx.doi.org/10.1029/2000JD90Q385
Ching. J: Herwehe. J: Swall. J. (2006). On joint deterministic grid modeling and sub-grid variability conceptual
framework for model evaluation. Atmos Environ 40: 4935-4945.
Chowdhurv. B: Svkes. I: Henn. D: Knipping. E: Karamchandani. P. (2012). Summary of updates to SCICHEM-
2012 model: comparison of results with observations and previous version results. Presentation presented at
11th Annual CMAS Conference, October 15-17, 2012, UNC-Chapel Hill.
Chuang. KJ: Chaa CC: Su. TC: Lin. LY: Lee. CT. (2007). Associations between particulate sulfate and organic
carbon exposures and heart rate variability in patients with or at risk for cardiovascular diseases. J Occup
Environ Med 49: 610-617. http://dx.doi.org/10.1097/JQM.0b013e318058205b
Chuang. KJ: Coull. BA: Zanobetti. A: Suh. H: Schwartz. J: Stone. PH: Litoniua. A: Speizer. FE: Gold. DR. (2008).
Particulate air pollution as a risk factor for ST-segment depression in patients with coronary artery disease.
Circulation 118: 1314-1320. http://dx.doi.org/10.1161/CIRCULATIQNAHA.108.765669
Chuang. KJ: Yan. YH: Cheng. TJ. (2010). Effect of air pollution on blood pressure, blood lipids, and blood sugar: A
population-based approach. J Occup Environ Med 52: 258-262.
http://dx.doi.org/10.1097/JQM.0b013e3181ceff7a
Chuang. KJ: Yan. YH: Chiu. SY: Cheng. TJ. (2011). Long-term air pollution exposure and risk factors for
cardiovascular diseases among the elderly in Taiwan. Occup Environ Med 68: 64-68.
http://dx.doi.org/10.1136/oem.2009.0527Q4
Chung. MK: Lao. TT: Ting. YH: Wong. TW: Leung. TY. (2014). Seasonality of fetal trisomy 21 - Have ambient air
pollutants played a role? J Matern Fetal Neonatal Med 28: 552-557.
http://dx.doi.org/10.3109/14767058.2014.9241Q4
Cimorelli. AJ: Perry. SG: Venkatram. A: Weil. JC: Paine. R: Wilson. RB: Lee. RF: Peters. WD: Brode. RW. (2005).
AERMOD: A dispersion model for industrial source applications. Part I: General model formulation and
boundary layer characterization. J Appl Meteorol 44: 682-693. http://dx.doi.org/10.1175/JAM2227.1
Clarisse. L: Fromm. M: Ngadi. Y: Emmons. L: Clerbaux. C: Hurtmans. D: Coheur. PF. (2011). Intercontinental
transport of anthropogenic sulfur dioxide and other pollutants: An infrared remote sensing case study.
Geophys Res Lett 38: L19806. http://dx.doi.org/10.1029/2011GL048976
Clarisse. L: Hurtmans. D: Clerbaux. C: Hadii-Lazaro. J: Ngadi. Y: Coheur. PF. (2012). Retrieval of sulphur dioxide
from the infrared atmospheric sounding interferometer (IASI). Atmos Meas Tech 5: 581-594.
http://dx.doi.org/10.5194/amt-5-581-2012
R-12
-------
Clark. NA: Demers. PA: Karr. CJ: Koehoora M: Lencar. C: Tamburic. L: Brauer. M. (2010). Effect of early life
exposure to air pollution on development of childhood asthma. Environ Health Perspect 118: 284-290.
http://dx.doi.org/10.1289/ehp.0900916
Clarke. JF: Edgertoa ES: Martia BE. (1997). Dry deposition calculations for the clean air status and trends
network. Atmos Environ 31: 3667-3678. http://dx.doi.Org/10.1016/S 1352-2310(97)00141-6
Cloughertv. JE: Kheirbek. I: Eisl. HM: Ross. Z: Pezeshki. G: Gorczvnski. JE: Johnsoa S: Markowitz. S: Kass. D:
Matte. T. (2013). Intra-urban spatial variability in wintertime street-level concentrations of multiple
combustion-related air pollutants: The New York City Community Air Survey (NYCCAS). J Expo Sci
Environ Epidemiol 23: 232-240. http://dx.doi.org/10.1038/ies.2012.125
Coletta. C: Papapetropoulos. A: Erdelvi. K: Olah. G: Modis. K: Panopoulos. P: Asimakopoulou. A: Geroe. D:
Sharina. I: Martia E: Szabo. C. (2012). Hydrogen sulfide and nitric oxide are mutually dependent in the
regulation of angiogenesis and endothelium-dependent vasorelaxation. Proc Natl Acad Sci USA 109: 9161-
9166. http://dx.doi.org/10.1073/pnas. 1202916109
Collaco. CR: Hochmaa DJ: Goldblum. RM: Brooks. EG. (2006). Effect of sodium sulfite on mast cell
degranulation and oxidant stress. Ann Allergy Asthma Immunol 96: 550-556.
http://dx.doi.Org/10.1016/S 1081-1206(10)63549-1
Conner. MW: Flood. WH: Rogers. AE: Amdur. MO. (1989). Changes in pulmonary lavage fluid of guinea pigs
exposed to ultrafine zinc oxide with adsorbed sulfuric acid. J Toxicol Environ Health 26: 223-234.
http://dx.doi.org/10.1080/15287398909531247
Conner. MW: Lam. HF: Rogers. AE: Fitzgerald. S: Amdur. MO. (1985). Lung injury in guinea pigs caused by
multiple exposures to submicron zinc oxide mixed with sulfur dioxide in a humidified furnace. J Toxicol
Environ Health 16: 101-114. http://dx.doi.org/10.1080/1528739850953Q722
Correia-Deur. J: Claudio. L: Imazawa. AT: Eluf-Neto. J. (2012). Variations in peak expiratory flow measurements
associated to air pollution and allergic sensitization in children in Sao Paulo, Brazil. Am J Ind Med 55:
1087-1098. http://dx.doi.org/10.1002/aiim.2206Q
Costa Nascimento. LF: Francisco. JB: Patto. MBR: Antunes. AM. (2012). Environmental pollutants and stroke-
related hospital admissions. Cad Saude Publica 28: 1319-1324. http://dx.doi.org/10.1590/S0102-
311X2012000700010
Cox. WM: Tikvart. JA. (1990). A statistical procedure for determining the best performing air quality simulation
model. Atmos Environ A 24: 2387-2395. http://dx.doi.org/10.1016/0960-1686(90)90331-G
Crouse. U: Laine-Alava. MT: Warren. DW: Wood. CL. (1999). A longitudinal study of nasal airway size from age 9
to age 13. Angle Orthod 69: 413-418.
Dabberdt. WF: Miller. E. (2000). Uncertainty, ensembles and air quality dispersion modeling: applications and
challenges. Atmos Environ 34: 4667-4673. http://dx.doi.Org/10.1016/S 1352-2310(00)00141-2
Dadvand. P: Rankin. J: Rushton. S: Pless-Mulloli. T. (2011a). Ambient air pollution and congenital heart disease: A
register-based study. Environ Res 111: 435-441. http://dx.doi.Org/10.1016/i.envres.2011.01.022
Dadvand. P: Rankin. J: Rushton. S: Pless-Mulloli. T. (2011b). Association between maternal exposure to ambient air
pollution and congenital heart disease: A register-based spatiotemporal analysis. Am J Epidemiol 173: 171-
182. http://dx.doi.org/10.1093/aie/kwa342
Dales. R: Chen. L: Frescura. AM: Liu. L: Villeneuve. PJ. (2009). Acute effects of outdoor air pollution on forced
expiratory volume in 1 s: A panel study of schoolchildren with asthma. Eur Respir J 34: 316-323.
http://dx.doi.org/10.1183/09031936.001389Q8
Dales. R: Kauri. LM: Cakmak. S: Mahmud. M: Weichenthal. SA: VanRvswvk. K: Kumarathasan. P: Thomsoa E:
Vincent. R: Broad. G: Liu. L. (2013). Acute changes in lung function associated with proximity to a steel
plant: A randomized study. Environ Int 55: 15-19. http://dx.doi.org/10.1016/i.envint.2013.01.014
R-13
-------
Dales. R: Wheeler. A: Mahmud. M: Frescura. AM: Smith-Doiroa M: Netherv. E: Liu. L. (2008). The influence of
living near roadways on spirometry and exhaled nitric oxide in elementary schoolchildren. Environ Health
Perspect 116: 1423-1427. http://dx.doi.org/10.1289/ehp. 10943
Dales. RE: Cakmak. S: Doiroa MS. (2006). Gaseous air pollutants and hospitalization for respiratory disease in the
neonatal period. Environ Health Perspect 114: 1751-1754. http://dx.doi.org/10.1289/ehp.9044
Dales. RE: Cakmak. S: Vidal. CB. (2010). Air pollution and hospitalization for venous thromboembolic disease in
Chile. J Thromb Haemost 8: 669-674. http://dx.doi.org/10.1111/i. 1538-7836.2010.03760.X
Darrow. LA: Kleia M: Flanders. WD: Waller. LA: Correa. A: Marcus. M: Mulholland. JA: Russell. AG: Tolbert.
PE. (2009). Ambient air pollution and preterm birth: A time-series analysis. Epidemiology 20: 689-698.
http://dx.doi.org/10.1097/EDE.0b013e3181a7128f
Darrow. LA: Kleia M: Strickland. MJ: Mulholland. JA: Tolbert. PE. (2011). Ambient air pollution and birth weight
in full-term infants JsEpjin Atlanta, 1994-2004. Environ Health Perspect 119: 731-737.
http://dx.doi.org/10.1289/ehp. 1002785
Deger. L: Plante. C: Jacques. L: Goudreau. S: Perron. S: Hicks. J: Kosatskv. T: Smargiassi. A. (2012). Active and
uncontrolled asthma among children exposed to air stack emissions of sulphur dioxide from petroleum
refineries in Montreal, Quebec: Across-sectional study. CanRespir J 19: 97-102.
Delfino. RJ: Coate. BP: Zeiger. RS: Seltzer. JM: Street. DH: Koutrakis. P. (1996). Daily asthma severity in relation
to personal ozone exposure and outdoor fungal spores. Am J Respir Crit Care Med 154: 633-641.
http://dx.doi.Org/10.1164/airccm.154.3.8810598
Delfino. RJ: Gone. H: Lina WS: Pellizzari. ED: Hu. Y. (2003a). Asthma symptoms in Hispanic children and daily
ambient exposures to toxic and criteria air pollutants. Environ Health Perspect 111: 647-656.
http://dx.doi.org/10.1289/ehp.5992
Delfino. RJ: Gong. H: Linn. WS: Hu. Y: Pellizzari. ED. (2003b). Respiratory symptoms and peak expiratory flow in
children with asthma in relation to volatile organic compounds in exhaled breath and ambient air. J Expo
Anal Environ Epidemiol 13: 348-363. http://dx.doi.org/10.1038/si.iea.7500287
Demokritou. P: Kavouras. IG: Ferguson. ST: Koutrakis. P. (2001). Development and laboratory performance
evaluation of a personal multipollutant sampler for simultaneous measurements for particulate and gaseous
pollutants. Aerosol Sci Technol 35: 741-752. http://dx.doi.org/10.1080/0278682Q152546789
Deng. O: Lu. C: Norback. D: Bornehag. CG: Zhang. Y: Liu. W: Yuan. H: Sundell. J. (2015a). Early life exposure to
ambient air pollution and childhood asthma in China. Environ Res 143: 83-92.
http://dx.doi.Org/10.1016/i.envres.2015.09.032
Deng. O: Lu. C: Ou. C: Liu. W. (2015b). Effects of early life exposure to outdoor air pollution and indoor
renovation on childhood asthma in China. Build Environ 93: 84-91.
http://dx.doi.Org/10.1016/i.buildenv.2015.01.019
Dennekamp. M: Akram. M: Abramsoa MJ: Tonkin. A: Sim. MR: Fridman. M: Erbas. B. (2010). Outdoor air
pollution as a trigger for out-of-hospital cardiac arrests. Epidemiology 21: 494-500.
http://dx.doi.org/10.1097/EDE.Qb013e3181e093db
Devalia. JL: Rusznak. C: Herdman. MJ: Trigg. CJ: Tarraf. H: Davies. RJ. (1994). Effect of nitrogen-dioxide and
sulfur-dioxide on airway response of mild asthmatic-patients to allergen inhalation. Lancet 344: 1668-1671.
http://dx.doi.org/10.1016/SQ 140-6736(94)90458-8
Dibben. C: Clemens. T. (2015). Place of work and residential exposure to ambient air pollution and birth outcomes
in Scotland, using geographically fine pollution climate mapping estimates. Environ Res 140: 535-541.
http://dx.doi.Org/10.1016/i.envres.2015.05.010
Dionisio. KL: Baxter. LK: Chang. HH. (2014). An empirical assessment of exposure measurement error and effect
attenuation in bipollutant epidemiologic models. Environ Health Perspect 122: 1216-1224.
http://dx.doi.org/10.1289/ehp. 1307772
R-14
-------
Dockerv. DW: Pope. CA. Ill; Xu. X: Spengler. JD: Ware. JH: Fay. ME: Ferris. BG. Jr: Speizer. FE. (1993). An
association between air pollution and mortality in six US cities. N Engl J Med 329: 1753-1759.
http://dx.doi.org/10.1056/NEJM199312Q93292401
Dockerv. DW: Speizer. FE; Stram. DO; Ware. JH; Spengler. JD; Ferris. BG. Jr. (1989). Effects of inhalable particles
on respiratory health of children. Am J Respir Crit Care Med 139: 587-594.
http://dx.doi.Org/10.1164/airccm/139.3.587
Dodge. R: Solomon. P; Movers. J: Haves. C. (1985). A longitudinal study of children exposed to sulfur oxides. Am
J Epidemiol 121: 720-736. http://dx.doi.Org/10.1093/aie/121.5.720
Polk. H; Armstrong. B; Lachowvcz. K; Vriiheid. M; Rankin. J; Abramskv. L; Bovd. PA; Welleslev. D. (2010).
Ambient air pollution and risk of congenital anomalies in England, 1991-1999. Occup Environ Med 67:
223-227. http://dx.doi.org/10.1136/oem.2009.045997
Pollard. GJ: Unsworth. MH; Harve. MJ. (1983). Pollutant transfer in upland regions by occult precipitation. Nature
302: 241-243. http://dx.doi.org/10.1038/302241a0
Pominici. F; McPermott. A; Paniels. M; Zeger. SL; Samet. JM. (2003). Mortality among residents of 90 cities
[HEI], In Revised analyses of time-series studies of air pollution and health (pp. 9-24). Boston, MA: Health
Effects Institute. http://pubs.healtheffects.org/view.php?id=4
Pong. G: Oian. Z; Wang. J; Chen. W: Ma. W: Trevathaa E; Xaverius. PK; PeClue. R: Wiese. A; Langstoa M; Liu.
MM; Wang. P; Rea W. (2013a). Associations between ambient air pollution and prevalence of stroke and
cardiovascular diseases in 33 Chinese communities. Atmos Environ 77: 968-973.
http://dx.doi.Org/10.1016/i.atmosenv.2013.06.034
Pong. GH; Oian. Z; Liu. MM; Wang. P; Ren. WH; Fu. O: Wang. J; Simckes. M; Ferguson. TF; Trevathaa. E.
(2013b). Obesity enhanced respiratory health effects of ambient air pollution in Chinese children: the Seven
Northeastern Cities study. Int J Obes (Lond) 37: 94-100. http://dx.doi.org/10.1038/iio.2012.125
Pong. GH; Oian. ZM; Liu. MM; Wang. P; Rea WH; Bawa. S: Fu. J: Wang. J; Lewis. R: Zelicoff. A; Simckes. M;
Trevathaa E. (2013c). Breastfeeding as a modifier of the respiratory effects of air pollution in children.
Epidemiology 24: 387-394. http://dx.doi.org/10.1097/EPE.0b013e3182877eb8
Pong. GH; Oian. ZM; Trevathaa E; Zeng. XW: Vaugha MG: Wang. J; Zhao. Y; Liu. YO; Rea WH; Oia XP.
(2014). Air pollution associated hypertension and increased blood pressure may be reduced by
breastfeeding in Chinese children: the Seven Northeastern Cities Chinese Children's Study. Int J Cardiol
176: 956-961. http://dx.doi.Org/10.1016/i.iicard.2014.08.099
Pong. GH; Oian. ZM; Xaverius. PK; Trevathaa E; Maalouf. S; Parker. J; Yang. L; Liu. MM; Wang. P; Ren. WH;
Ma. W: Wang. J; Zelicoff. A; Fu. O: Simckes. M. (2013d). Association between long-term air pollution and
increased blood pressure and hypertension in China. Hypertension 61: 578-584.
http://dx.doi.org/10.1161/HYPERTENSIONAHA. 111.00003
Pong. GH; Wang. J: Zeng. XW: Chea L; Oia XP; Zhou. Y; Li. M: Yang. M: Zhao. Y; Rea WH; Hu. OS. (2015).
Interactions between air pollution and obesity on blood pressure and hypertension in Chinese children.
Epidemiology 26: 740-747. http://dx.doi.org/10.1097/EPE.000000000000Q336
Pong. GH; Zhang. P; Sun. B; Zhang. L; Chea X; Ma. N; Yu. F; Guo. H; Huang. H; Lee. YL; Tang. N; Chea J.
(2012). Long-term exposure to ambient air pollution and respiratory disease mortality in Shenyang, China:
A 12-year population-based retrospective cohort study. Respiration 84: 360-368.
http://dx.doi.org/10.1159/00033293Q
Pourado. H; Santos. J: Reis. N. Jr; Melo. AMY. (2012). The effects of atmospheric turbulence on peak-to-mean
concentration ratio and its consequence on the odour impact assessment using dispersion models. Chemical
Engineering Transactions 30: 163-168. http://dx.doi.org/10.3303/CET1230Q28
Praxler. RR. (1999). HYSPLIT 4 user's guide. Available online at http://readv.arl.noaa.gov/HYSPLIT.php
Pubowskv. SP: Suh. H; Schwartz. J; Coull. BA; Gold. PR. (2006). Piabetes, obesity, and hypertension may
enhance associations between air pollution and markers of systemic inflammation. Environ Health Perspect
114: 992-998. http://dx.doi.org/10.1289/ehp.8469
R-15
-------
Pugandzic. R: Dodds. L: Stieb. D: Smith-Doiron. M. (2006). The association between low level exposures to
ambient air pollution and term low birth weight: A retrospective cohort study. Environ Health 5:3.
http://dx.doi.org/10.1186/1476-069X-5-3
Eatough. DJ: Arthur. RJ: Eatough. NL: Hill. MW: Mangelsoa NF: Richter. BE: Hansen. LP: Cooper. JA. (1984).
Rapid conversion of S02(g) to sulfate in a fog bank. Environ Sci Technol 18: 855-859.
http://dx.doi.org/10.1021/esQ0129a009
Eatough. DJ: Eatough. NL: Hill. MW: Mangelson. NF: Ryder. J: Hansen. LP: Meisenheimer. R: Fischer. J. (1979).
The chemical composition of smelter flue dusts. Atmos Environ 13: 489-506.
http://dx.doi.org/10.1016/0004-6981(79)90142-2
Eatough. DJ: Hansen. LP. (1980). S(IV) chemistry in smelter produced particulate matter [Review]. Am J Ind Med
1: 435-448. http://dx.doi.org/10.1002/aiim.470001Q322
Eatough. PJ: Hansen. LP. (1983). Organic and inorganic S(IV) compounds in airborne particulate matter. In Trace
atmospheric constituents: Properties, transformations, and fates. New York, NY: John Wiley and Sons.
Eatough. PJ: Hansen. LP. (1984). Bis-hydroxymethyl sulfone: A major aerosol product of atmospheric reactions of
S02(g). Sci Total Environ 36: 319-328. http://dx.doi.org/10.1016/0048-9697(84)90283-3
Eatough. PJ: Lee. ML: Later. PW: Richter. BE: Eatough. NL: Hansea LP. (1981). Pimethyl sulfate in particulate
matter from coal- and oil-fired power plants. Environ Sci Technol 15: 1502-1506.
http://dx.doi.org/10.1021/esQ0094a014
Eatough. PJ: Major. T: Ryder. J: Hill. M: Mangelsoa NF: Eatough. NL: Hansea LP: Meisenheimer. RG: Fischer.
JW. (1978). The formation and stability of sulfite species in aerosols. Atmos Environ (1967) 12: 263-271.
http://dx.doi.org/10.1016/0004-6981(78)90206-8
Eatough. PJ: White. VF: Hansen. LP: Eatough. NL: Cheney. JL. (1986). Identification of gas-phase dimethyl
sulfate and monoethyl sulfate in the Los Angeles atmosphere. Environ Sci Technol 20: 867-872.
http://dx.doi.org/10.1021/esQ0151a003
Ebisu. K: Bell. ML. (2012). Airborne PM2.5 chemical components and low birth weight in the Northeastern and
Mid-Atlantic regions of the United States. Environ Health Perspect 120: 1746-1752.
http://dx.doi.org/10.1289/ehp. 1104763
Eguchi. K: Schwartz. JE: Pickering. TG: Hoshide. S: Ishikawa. J: Shimada. K: Kario. K. (2010). Increased heart rate
variability during sleep is a predictor for future cardiovascular events in patients with type 2 diabetes.
Hypertens Res 33: 737-742. http://dx.doi.org/10.1038/hr.2010.61
Eitaa O: Yuval. O: Barchana. M: Pubnov. J: Lina S: Carmel. Y: Brodav. PM. (2010). Spatial analysis of air
pollution and cancer incidence rates in Haifa Bay, Israel. Sci Total Environ 408: 4429-4439.
http://dx.doi.Org/10.1016/i.scitotenv.2010.06.031
Elgethun. K: Yost. MG: Fitzpatrick. CTE: Nverges. TL: Fenske. RA. (2007). Comparison of global positioning
system (GPS) tracking and parent-report diaries to characterize children's time-location patterns. J Expo Sci
Environ Epidemiol 17: 196-206. http://dx.doi.org/10.1038/si.ies.7500496
Elliott. P: Shaddick. G: Wakefield. JC: de Hoogh. C: Briggs. PJ. (2007). Long-term associations of outdoor air
pollution with mortality in Great Britain. Thorax 62: 1088-1094. http://dx.doi.org/10.1136/thx.2006.076851
Enkhmaa. P: Warburtoa N: Javzandulam B: Uvanga. J: Khishigsurea Y: Lodovsamba. S: Enkhtur. S: Warburtoa
P. (2014). Seasonal ambient air pollution correlates strongly with spontaneous abortion in Mongolia. BMC
Pregnancy Childbirth 14: 146. http://dx.doi.org/10.1186/1471-2393-14-146
Etlik. O: Tomur. A: Kutmaa MN: Yorukaa S: Puman. O. (1995). The effects of sulfur dioxide inhalation and
antioxidant vitamins on red blood cell lipoperoxidation. Environ Res 71: 25-28.
http://dx.doi.org/10.1006/enrs. 1995.1063
Faiz. AS: Rhoads. GG: Pemissie. K: Kruse. L: Lin. Y: Rich. PO. (2012). Ambient air pollution and the risk of
stillbirth. Am J Epidemiol 176: 308-316. http://dx.doi.org/10.1093/aie/kws029
R-16
-------
Faiz. AS: Rhoads. GG: Demissie. K: Lin. Y: Kruse. L: Rich. DO. (2013). Does ambient air pollution trigger
stillbirth? Epidemiology 24: 538-544. http://dx.doi.org/10.1097/EDE.0b013e3182949ce5
Farhi. A: Bovko. V: Almagor. J: Benensoa I: Segre. E: Rudich. Y: Stern. E: Lerner-Geva. L. (2014). The possible
association between exposure to air pollution and the risk for congenital malformations. Environ Res 135C:
173-180. http://dx.doi.Org/10.1016/i.envres.2014.08.024
Fattore. E: Davoli. E: Castiglioni. S: Bosetti. C: Re Depaolini. A: Marzona. I: Zuccato. E: Fanelli. R. (2016).
Wastewater-based epidemiological evaluation of the effect of air pollution on short-acting beta-agonist
consumption for acute asthma treatment. Environ Res 150: 106-111.
http://dx.doi.Org/10.1016/i.envres.2016.05.051
Feichter. J: Kiellstrom. E: Rodhe. H: Dentener. F: Lelieveld. J: Roelofs. GJ. (1996). Simulation of the tropospheric
sulfur cycle in a global climate model. Atmos Environ 30: 1693-1707. http://dx.doi.org/10.lQ16/1352-
2310(95)00394-0
Ferek. RJ: Covert. PA: Luke. W. (1997). Intercomparison of measurements of sulfur dioxide in ambient air by
carbonate-impregnated filters and Teco pulsed-fluorescence analyzers. J Geophys Res Atmos 102:
1626716272. http://dx.doi.org/10.1029/96JD03587
Ferris. BG. (1978). Epidemiology standardization project II: Recommended respiratory disease questionnaires for
use with adults and children in epidemiological research. Am J Respir Crit Care Med 118: 7-53.
Field. PI: Simmul. R: Bell. SC: Allea DH: Berend. N. (1996). Evidence for opioid modulation and generation of
prostaglandins in sulphur dioxide (SO)2-inducedbronchoconstriction. Thorax 51: 159-163.
http://dx.doi.Org/10.1136/thx.51.2.159
Filho. MAP: Pereira. LAA: Arbex. FF: Arbex. M: Conceicao. GM: Santos. UP: Lopes. AC: Saldiva. PHN: Braga.
ALF: Cendoa S. (2008). Effect of air pollution on diabetes and cardiovascular diseases in Sao Paulo,
Brazil Braz J Med Biol Res 41: 526-532. http://dx.doi.org/10.1590/S0100-879X200800500002Q
Filipovic. MR: Milikovic. J: Nauser. T: Rovzea M: Klos. K: Shubina. T: Koppenol. WH: Lippard. SJ: Ivanovic-
Burmazovic. I. (2012). Chemical characterization of the smallest S-nitrosothiol, HSNO; cellular cross-talk
of H2S and S-nitrosothiols. J Am Chem Soc 134: 12016-12027. http://dx.doi.org/10.1021/ia3009693
Filleul. L: Rondeau. V: Vandentorren. S: Le Moual. N: Cantagrel. A: Annesi-Maesano. I: Charpia D: Declerca. C:
Neukirch. F: Paris. C: Vervloet. D: Brochard. P: Tessier. JF: Kauffmana F: Baldi. I. (2005). Twenty five
year mortality and air pollution: Results from the French PAARC survey. Occup Environ Med 62: 453-460.
http://dx.doi.org/10.1136/oem.2004.014746
Finkelsteia JN: Johnston. CJ. (2004). Enhanced sensitivity of the postnatal lung to environmental insults and
oxidant stress [Review]. Pediatrics 113: 1092-1096.
Finlavson-Pitts. BJ: Pitts. JN. Jr. (2000). Chemistry of the upper and lower atmosphere: Theory, experiments and
applications. In Chemistry of the upper and lower atmosphere: Theory, experiments and applications. San
Diego, CA: Academic Press. http://www.sciencedirect.com/science/book/97801225706Q5
Fioletov. YE: McLindea CA: Krotkov. N: Moran. MP: Yang. K. (2011). Estimation of S02 emissions using OMI
retrievals. Geophys Res Lett 38: 1-5. http://dx.doi.org/10.1029/2011GL049402
Fioletov. YE: McLindea CA: Krotkov. N: Yang. K: Loyola. DG: Valks. P: Thevs. N: Van Roozendael. M: Nowlaa
CR: Chance. K: Liu. X: Lee. C: Martin. RV. (2013). Application of OMI, SCIAMACHY, and GOME-2
satellite S02 retrievals for detection of large emission sources. J Geophys Res Atmos 118: 11399-11418.
http://dx.doi.org/10.1002/igrd.50826
Forbes. LJ: Patel. MP: Rudnicka. AR: Cook. DG: Bush. T: Stedmaa JR: Whincup. PH: Strachan. DP: Anderson.
HR. (2009a). Chronic exposure to outdoor air pollution and diagnosed cardiovascular disease: meta-
analysis of three large cross-sectional surveys. Environ Health 8: 30. http://dx.doi.org/10.1186/1476-069X-
8-30
Forbes. LJ: Patel. MP: Rudnicka. AR: Cook. DG: Bush. T: Stedmaa JR: Whincup. PH: Strachan. DP: Anderson.
RH. (2009b). Chronic exposure to outdoor air pollution and markers of systemic inflammation.
Epidemiology 20: 245-253. http://dx.doi.org/10.1097/EPE.0b013e318190ea3f
R-17
-------
Forbes. LJL: Kapetanakis. V: Rudnicka. AR: Cook. DG: Bush. T: Stedmaa JR: Whincup. PH: Strachan. DP:
Anderson. HR. (2009c). Chronic exposure to outdoor air pollution and lung function in adults. Thorax 64:
657-663. http://dx.doi.org/10.1136/thx.20Q8.109389
Frank. NR: Amdur. MO: Worcester. J: Whittenberger. JL. (1962). Effects of acute controlled exposure to S02 on
respiratory mechanics in healthy male adults. J Appl Physiol (1985) 17: 252-258.
Frank. NR: Yoder. RE: Braia JD: Yokovama. E. (1969). S02 (35S labeled) absorption by the nose and mouth under
conditions of varying concentration and flow. Arch Environ Occup Health 18: 315-322.
http://dx.doi.org/10.1080/00Q39896.1969.10665414
Frank. NR: Yoder. RE: Yokovama. E: Speizer. FE. (1967). The diffusion of 35S02 from tissue fluids into the lungs
following exposure of dogs to 35S02. Health Phys 13: 31-38.
Frev. HC: Zhao. Y. (2004). Quantification of variability and uncertainty for air toxic emission inventories with
censored emission factor data. Environ Sci Technol 38: 6094-6100. http://dx.doi.org/10.1021/es035096m
Friedman. B: Brophv. P: Brune. WH: Farmer. DK. (2016). Anthropogenic sulfur perturbations on biogenic
oxidation: S02 additions impact gas-phase OH oxidation products of a- and (3-pinene. Environ Sci Technol
50: 1269-1279. http://dx.doi.org/10.1021/acs.est.5b05010
Frischer. T: Studnicka. M: Gartner. C: Tauber. E: Horak. F: Veiter. A: Spengler. J: Kuhr. J: Urbanek. R. (1999).
Lung function growth and ambient ozone: A three-year population study in school children. Am J Respir
Crit Care Med 160: 390-396. http://dx.doi.Org/10.1164/airccm.160.2.9809075
Frost. KD. (2014). AERMOD performance evaluation for three coal-fired electrical generating units in Southwest
Indiana. J Air Waste Manag Assoc 64: 280-290. http://dx.doi.org/10.1080/10962247.2Q13.858651
Fruia SA: Hudda. N: Sioutas. C: Defino. RJ. (2011). Predictive model for vehicle air exchange rates based on a
large, representative sample. Environ Sci Technol 45: 3569-3575. http://dx.doi.org/10.1021/esl03897u
Frve. C: Hoelscher. B: Cvrvs. J: Wist. M: Wichmann. HE: Heinrich. J. (2003). Association of lung function with
declining ambient air pollution. Environ Health Perspect 111: 383-387.
Fung. KY: Khaa S: Krewski. D: Chea Y. (2006). Association between air pollution and multiple respiratory
hospitalizations among the elderly in Vancouver, Canada. Inhal Toxicol 18: 1005-1011.
http://dx.doi.org/10.1080/089583706009Q4538
Fung. KY: Luginaah. I: Gorev. KM: Webster. G. (2005). Air pollution and daily hospital admissions for
cardiovascular diseases in Windsor, Ontario. Can J Public Health 96: 29-33.
Gandhi. SK: Rich. DO: Ohman-Strickland. PA: Kipen. HM: Gow. A. (2014). Plasma nitrite is an indicator of acute
changes in ambient air pollutant concentrations. Inhal Toxicol 26: 426-434.
http://dx.doi.org/10.3109/08958378.2014.913216
Ganguly. R: Battermaa S: Isakov. V: Snyder. M: Breea M: Brakefield-Caldwell. W. (2015). Effect of geocoding
errors on traffic-related air pollutant exposure and concentration estimates. J Expo Sci Environ Epidemiol
25: 490-498. http://dx.doi.Org/10.1038/ies.2015.l
Gariazzo. C: Pelliccioni. A: Bogliolo. M: Scalisi. G. (2004). Evaluation of a Lagrangian particle model (SPRAY) to
assess environmental impact of an industrial facility in complex terrain. Water Air Soil Pollut 155: 137-
158. http://dx.doi.Org/10.1023/B:WATE.0000026525.82039.ef
Geer. LA: Weedoa J: Bell. ML. (2012). Ambient air pollution and term birth weight in Texas from 1998 to 2004. J
Air Waste Manag Assoc 62: 1285-1295. http://dx.doi.org/10.1080/10962247.2012.7Q7632
Gent. JF: Koutrakis. P: Belanger. K: Triche. E: Holford. TR: Bracken. MB: Leaderer. BP. (2009). Symptoms and
medication use in children with asthma and traffic-related sources of fine particle pollution. Environ Health
Perspect 117: 1168-1174. http://dx.doi.org/10.1289/ehp.0800335
Gevh. AS: Xue. J: Ozkavnak. H: Spengler. JD. (2000). The Harvard Southern California chronic ozone exposure
study: Assessing ozone exposure of grade-school-age children in two southern California communities.
Environ Health Perspect 108: 265-270. http://dx.doi.org/10.1289/ehp.00108265
R-18
-------
Gianicolo. EAL: Mangia. C: Cervino. M: Bruni. A: Andreassi. MG: Latini. G. (2014). Congenital anomalies among
live births in a high environmental risk area-A case-control study in Brindisi (southern Italy). Environ Res
128: 9-14. http://dx.doi.Org/10.1016/i.envres.2013.ll.002
Gilford. F. (1960). Peak to average concentration ratios according to a fluctuating plume dispersion model. Int J Air
Pollut 3: 253-260.
Gilbert. WM: Nesbitt. TS: Danielsen. B. (2003). The cost of prematurity: Quantification by gestational age and birth
weight. Obstet Gynecol 102: 488-492.
Gilboa. SM: Mendola. P: Olshaa AF: Langlois. PH: Savitz. DA: Loomis. D: Herring. AH: Fixler. DE. (2005).
Relation between ambient air quality and selected birth defects, seven county study, Texas, 1997-2000. Am
J Epidemiol 162: 238-252. http://dx.doi.org/10.1093/aie/kwil89
Gilliland. FD: Berhane. K: Islam. T: Mcconnell. R: Gauderman. WJ: Gilliland. SS: Avol. E: Peters. JM. (2003).
Obesity and the risk of newly diagnosed asthma in school-age children. Am J Epidemiol 158: 406-415.
http://dx.doi.org/10.1093/aie/kwgl75
Giordano. L: Brunner. D: Flemming. J: Hogrefe. C: Im U: Bianconi. R: Badia. A: Balzarini. A: Baro. R: Chemel. C:
Curci. G: Forkel. R: Jimenez-Guerrero. P: Hirtl. M: Hodzic. A: Honzak. L: Jorba. O: Knote. C: Kuenen.
JJP: Makar. PA: Manders-Groot. A: Neal. L: Perez. JL: Pirovano. G: Pouliot. G: San Jose. R: Savage. N:
Schroeder. W: Sokhi. RS: Svrakov. D: Toriaa A: Tuccella. P: Werhahn. J: Wolke. R: Yahva. K: Zabkar.
R: Zhang. Y: Galmarini. S. (2015). Assessment of the MACC reanalysis and its influence as chemical
boundary conditions for regional air quality modeling in AQMEII-2. Atmos Environ 115: 371-388.
http://dx.doi.Org/10.1016/i.atmosenv.2015.02.034
Glasgow. ML: Rudra. CB: Yoo. EH: Demirbas. M: Merriman. J: Navak. P: Crabtree-Ide. C: Szpiro. AA: Rudra. A:
Wactawski-Wende. J: Mu. L. (2014). Using smartphones to collect time-activity data for long-term
personal-level air pollution exposure assessment. J Expo Sci Environ Epidemiol 26: 356-364.
http://dx.doi.org/10.1038/ies.2014.78
Godoi. RH: Godoi. AF: Goncalves Junior. SJ: Paralovo. SL: Borillo. GC: Goncalves Gregorio Barbosa. C: Arantes.
MG: Charello. RC: Rosario Filho. NA: Grassi. MT: Yamamoto. CI: Potgieter-Vermaak. S: Rotondo. GG:
De Wael. K: van Griekea R. (2013). Healthy environment - indoor air quality of Brazilian elementary
schools nearby petrochemical industry. Sci Total Environ 463-464: 639-646.
http://dx.doi.Org/10.1016/i.scitotenv.2013.06.043
Gold. PR: Damokosh. AI: Dockerv. DW: Berkev. CS. (2003). Body-mass index as a predictor of incident asthma in
a prospective cohort of children. Pediatr Pulmonol 36: 514-521. http://dx.doi.org/10.1002/ppul. 10376
Goldberg. MS: Giannetti. N: Burnett. RT: Mavo. NE: Valois. MF: Brophv. JM. (2008). A panel study in congestive
heart failure to estimate the short-term effects from personal factors and environmental conditions on
oxygen saturation and pulse rate. Occup Environ Med 65: 659-666.
http://dx.doi.org/10.1136/oem.2007.034934
Goldmaa GT: Mulholland. JA: Russell. AG: Gass. K: Strickland. MJ: Tolbert. PE. (2012). Characterization of
ambient air pollution measurement error in a time-series health study using a geostatistical simulation
approach. Atmos Environ 57: 101-108. http://dx.doi.Org/10.1016/i.atmosenv.2012.04.045
Goldmaa GT: Mulholland. JA: Russell. AG: Srivastava. A: Strickland. MJ: Klein. M: Waller. LA: Tolbert. PE:
Edgertoa ES. (2010). Ambient air pollutant measurement error: characterization and impacts in a time-
series epidemiologic study in Atlanta. Environ Sci Technol 44: 7692-7698.
http://dx.doi.org/10.1021/esl01386r
Goldman. GT: Mulholland. JA: Russell. AG: Strickland. MJ: Kleia M: Waller. LA: Tolbert. PE. (2011). Impact of
exposure measurement error in air pollution epidemiology: Effect of error type in time-series studies.
Environ Health 10: 61. http://dx.doi.org/10.1186/1476-069X-10-61
Gong. H. Jr: Lachenbruch. PA: Harber. P: Linn. WS. (1995). Comparative short-term health responses to sulfur
dioxide exposure and other common stresses in a panel of asthmatics. Toxicol Ind Health 11: 467-487.
Gong. H. Jr: Linn. WS: Shamoo. DA: Andersoa KR: Nugent. CA: Clark. KW: Lia AE. (1996). Effect of inhaled
salmeterol on sulfur dioxide-induced bronchoconstriction in asthmatic subjects. Chest 110: 1229-1235.
R-19
-------
Gong. H. Jr: Linn. WS: Terrell. SL: Anderson. KR: Clark. KW. (2001). Anti-inflammatory and lung function effects
of montelukast in asthmatic volunteers exposed to sulfur dioxide. Chest 119: 402-408.
http://dx.doi.Org/10.1378/chest.119.2.402
Goodman. JE: Seelev. M: Mattuck. R: Thakali. S. (2015). Do group responses mask the effects of air pollutants on
potentially sensitive individuals in controlled human exposure studies? [Review]. Regul Toxicol Pharmacol
71: 552-564. http://dx.doi.Org/10.1016/i.vrtph.2015.02.002
Green. R: Sarovar. V: Malig. B: Basu. R. (2015). Association of stillbirth with ambient air pollution in a California
cohort study. Am J Epidemiol 181: 874-882. http://dx.doi.org/10.1093/aie/kwu460
Greenberg. N: Carel. RS: Derazne. E: Bibi. H: Shpriz. M: Tzur. D: Portnov. BA. (2016). Different effects of long-
term exposures to S02 and N02 air pollutants on asthma severity in young adults. J Toxicol Environ
Health A 79: 1-10. http://dx.doi.org/10.1080/15287394.2016.1153548
Greenwald. R: Sarnat. SE: Ravsoni. AU: Li. WW: Johnsoa BA: Stock. TH: Holguia F: Sosa. T: Sarnat. JA. (2013).
Associations between source-indicative pollution metrics and increases in pulmonary inflammation and
reduced lung function in a panel of asthmatic children. Air Qual Atmos Health 6: 487-499.
http://dx.doi.org/10.1007/sll869-012-Q186-3
Gregory. RE: Gunnisoa AF. (1984). Identification of plasma-proteins containing sulfite-reactive disulfide bonds.
Chem Biol Interact 49: 55-69. http://dx.doi.org/10.1016/0009-2797(84)90052-8
Groneberg. DA: Ouarcoo. D: Frossard. N: Fischer. A. (2004). Neurogenic mechanisms in bronchial inflammatory
diseases [Review]. Allergy 59: 1139-1152. http://dx.doi.org/10.Ill 1/i. 1398-9995.2004.00665.x
Grontoft. T: Ravchaudhuri. MR. (2004). Compilation of tables of surface deposition velocities for 03, N02 and
S02 to a range of indoor surfaces. Atmos Environ 38: 533-544.
http://dx.doi.Org/10.1016/i.atmosenv.2003.10.010
Grunsteia MM: Hazucha. M: Sorli. J: Milic-Emili. J. (1977). Effect of S02 on control of breathing in anesthetized
cats. J Appl Physiol (1985) 43: 844-851.
Gryparis. A: Coull. BA: Schwartz. J. (2007). Controlling for confounding in the presence of measurement error in
hierarchical models: A Bayesian approach. J Expo Sci Environ Epidemiol 17: S20-S28.
http://dx.doi.org/10.1038/si.ies.7500624
Guav. M: Stieb. DM: Smith-Doiroa M. (2011). Assessment of long-term exposure to air pollution in a longitudinal
national health survey. J Expo Sci Environ Epidemiol 21: 337-342. http://dx.doi.org/10.1038/ies.2010.37
Gulliver. J: Morris. C: Lee. K: Vienneau. D: Briggs. D: Hansell. A. (2011). Land use regression modeling to
estimate historic (1962-1991) concentrations of black smoke and sulfur dioxide for Great Britain. Environ
Sci Techno145: 3526-3532. http://dx.doi.org/10.1021/esl03821v
Gunnison. AF. (1981). Sulfite toxicity - a critical-review of invitro and invivo data. Food Cosmet Toxicol 19: 667-
682. http://dx.doi.org/10.1016/0015-6264(81)90519-8
Gunnison. AF: Bentoa AW. (1971). Sulfur dioxide: sulfite interaction with mammalian serum and plasma. Arch
Environ Occup Health 22: 381-388.
Gunnison. AF: Jacobsea DW: Schwartz. HJ. (1987a). Sulfite hypersensitivity: A critical review [Review]. CRC Crit
Rev Toxicol 17: 185-214. http://dx.doi.org/10.3109/104084487090712Q8
Gunnison. AF: Palmes. ED. (1973). Persistence of plasma S-sulfonates following exposure of rabbits to sulfite and
sulfur dioxide. Toxicol Appl Pharmacol 24: 266-278.
Gunnison. AF: Palmes. ED. (1974). S-sulfonates in human plasma following inhalation of sulfur dioxide. Am Ind
Hyg Assoc J 35: 288-291. http://dx.doi.org/10.1080/0002889748507Q36
Gunnison. AF: Sellakumar. A: Currie. D: Snyder. EA. (1987b). Distribution, metabolism and toxicity of inhaled
sulfur dioxide and endogenously generated sulfite in the respiratory tract of normal and sulfite oxidase-
deficient rats. J Toxicol Environ Health 21: 141-162. http://dx.doi.org/10.1080/152873987095310Q8
R-20
-------
Gunnison. AF: Zaccardi. J: Dulak. L: Chiang. G. (1981). Tissue distribution of s-sulfonate metabolites following
exposure to sulfur-dioxide. Environ Res 24: 432-443.
Guo. Y: Jia. Y: Paa X: Liu. L: Wichmann. HE. (2009). The association between fine particulate air pollution and
hospital emergency room visits for cardiovascular diseases in Beijing, China. Sci Total Environ 407: 4826-
4830. http://dx.doi.Org/10.1016/i.scitotenv.2009.05.022
Guo. Y: Tong. S: Li. S: Barnett. AG: Yu. W: Zhang. Y: Paa X. (2010). Gaseous air pollution and emergency
hospital visits for hypertension in Beijing, China: a time-stratified case-crossover study. Environ Health 9:
57. http://dx.doi.org/10.1186/1476-069X-9-57
Gwvna RC: Burnett. RT: Thurston. GD. (2000). A time-series analysis of acidic particulate matter and daily
mortality and morbidity in the Buffalo, New York, region. Environ Health Perspect 108: 125-133.
Ha. EH: Hong. YC: Lee. BE: Woo. BH: Schwartz. J: Christian! DC. (2001). Is air pollution a risk factor for low
birth weight in Seoul? Epidemiology 12: 643-648.
Haider. SS: Hasaa M: Khan. NH. (1982). Air pollutant sulfur dioxide-induced alterations on the levels of lipids,
lipid peroxidation and lipase activity in various regions of the rat brain. Basic Clin Pharmacol Toxicol 51:
45-50. http://dx.doi.org/10.1111/i. 1600-0773.1982.tb01061.x
Haii. AM: Burki. NK: Lee. LY. (1996). Role of tachykinins in sulfur dioxide-induced bronchoconstriction in
anesthetized guinea pigs. J Appl Physiol (1985) 80: 2044-2050.
Hales. JM: Sutter. SL. (1973). Solubility of sulfur dioxide in water at low concentrations. Atmos Environ 7: 997-
1001. http://dx.doi.org/10.1016/0004-6981(73)90049-8
Halinea AI: Salonea RO: Pennanea AS. (2000a). Combined respiratory effects of cold air with S02 or N02 in
single 1-hour exposures of hyperventilating guinea pigs. Inhal Toxicol 12: 693-713.
http://dx.doi.org/10.1080/08958370050085147
Halinen. AI: Salonea RO: Pennanen. AS: Kosma. VM. (2000b). Combined respiratory effects of cold air with S02
or N02 in repeated 10-minute exposures of hyperventilating guinea pigs. Inhal Toxicol 12: 671-691.
http://dx.doi.org/10.1080/08958370050Q85138
Halios. CC: Helmis. CG: Eleftheriadis. K: Flocas. HA: Assimakopoulos. YD. (2009). A Comparative Study of the
Main Mechanisms Controlling Indoor Air Pollution in Residential Flats. Water Air Soil Pollut 204: 333-
350. http://dx.doi.org/10.1007/sll270-009-0Q48-2
Halios. CH: Helmis. CG: Deligianni. K: Vratolis. S: Eleftheriadis. K. (2014). Determining the ventilation and
aerosol deposition rates from routine indoor-air measurements. Environ Monit Assess 186: 151-163.
http://dx.doi.org/10.1007/slQ661-013-3362-5
Hand. JL: Schichtel. BA: Malm. WC: Pitchford. ML. (2012). Particulate sulfate ion concentration and S02 emission
trends in the United States from the early 1990s through 2010. Atmos ChemPhys 12: 10353-10365.
http://dx.doi.org/10.5194/acp-12-10353-2012
Hankinson. JL: Odencrantz. JR: Fedaa KB. (1999). Spirometric reference values from a sample of the general US
population. Am J Respir Crit Care Med 159: 179-187. http://dx.doi.org/10.1164/airccm. 159.1.9712108
Hanna. SR. (2007). Chapter 4.0. A review of uncertainty and sensitivity analyses of atmospheric transport and
dispersion models. In Developments in environmental science: Air pollution modeling and its application
XVIII. Boston, Massachusetts: Elsevier Publishing Co. http://dx.doi.Org/10.1016/S 1474-8177(07)06040-8
Hanna. SR: Briggs. GA: Hosker. RP. (1982). Handbook on atmospheric diffusion. (DOE/TIC-11223; DE82002045).
Springfield, VA: U.S. Department of Energy.
Hanna. SR: Brown. MJ: Camelli. FE: Chan. ST: Coirier. WJ: Hansen. OR: Huber. AH: Kim. S: Reynolds. RM.
(2006). Detailed simulations of atmospheric flow and dispersion in downtown Manhattan: An application
of five computational fluid dynamics models. Bull AmMeteorol Soc 87: 1713-1726.
http://dx.doi.org/10.1175/BAMS-87-12-1713
Hanna. SR: Chang. JC. (1993). Hybrid plume dispersion model (HPDM) improvements and testing at three field
sites. Atmos Environ A 27: 1491-1508. http://dx.doi.org/10.1016/0960-1686(93)90135-L
R-21
-------
Hanna. SR: Chang. JC: Strimaitis. DG. (1993). Hazardous gas model evaluation with field observations. Atmos
Environ A 27: 2265-2285. http://dx.doi.org/10.1016/0960-1686(93)90397-H
Hanna. SR: Egaa BA: Purdum. J: Wagler. J. (2001). Evaluation of the ADMS, AERMOD, and ISC3 dispersion
models with the OPTEX, Duke Forest, Kincaid, Indianapolis and Lovett field datasets. Int J Environ Pollut
16: 301-314. http://dx.doi.org/10.1504/IJEP.2001.000626
Hansell. A: Ghosh. RE: Blangiardo. M: Perkins. C: Vienneau. D: Goffe. K: Briggs. D: Gulliver. J. (2016). Historic
air pollution exposure and long-term mortality risks in England and Wales: prospective longitudinal cohort
study. Thorax 71: 330-338. http://dx.doi.Org/10.l 136/thoraxinl-2015-207111
Hansen. CA: Barnett. AG: Pritchard. G. (2008). The effect of ambient air pollution during early pregnancy on fetal
ultrasonic measurements during mid-pregnancy. Environ Health Perspect 116: 362-369.
http://dx.doi.org/10.1289/ehp. 10720
Hansen. M: Bower. C: Milne. E: de Klerk. N: Kurinczuk. JJ. (2005). Assisted reproductive technologies and the risk
of birth defects~a systematic review [Review]. Hum Reprod 20: 328-338.
http://dx.doi.org/10.1093/humrep/deh593
Harre. ESM: Price. PD: Avrev. RB: Toop. LJ: Martin. IR: Towa GI. (1997). Respiratory effects of air pollution in
chronic obstructive pulmonary disease: A three month prospective study. Thorax 52: 1040-1044.
http://dx.doi.org/10.1136/thx.52.12.1040
Harries. MG: Parkes. PEG: Lessof. MH: Orr. TSC. (1981). Role of bronchial irritant receptors in asthma. Lancet
317: 5-7. http://dx.doi.org/10.1016/S0140-6736(81)90113-6
Hart. JE: Garshick. E: Dockerv. DW: Smith. TJ: Rvaa L: Ladea F. (2011). Long-term ambient multi-pollutant
exposures and mortality. Am J Respir Crit Care Med 183: 73-78. http://dx.doi.org/10.1164/rccm.200912-
1903QC
Hazucha. M: Bates. DV. (1975). Combined effect of ozone and sulphur dioxide on human pulmonary function.
Nature 257: 50-51. http://dx.doi.org/10.1038/257050a0
Heckbert. SR: Kooperberg. C: Safford. MM: Psatv. BM: Hsia. J: McTiernaa A: Gaziano. JM: Frishmaa WH: Curb.
JD. (2004). Comparison of self-report, hospital discharge codes, and adjudication of cardiovascular events
in the Women's Health Initiative. Am J Epidemiol 160: 1152-1158. http://dx.doi.org/10.1093/aie/kwh314
Hedlev. AJ: Wong. CM: Thach. TO: Ma. S: Lam. TH: Anderson. HR. (2002). Cardiorespiratory and all-cause
mortality after restrictions on sulphur content of fuel in Hong Kong: an intervention study. Lancet 360:
1646-1652.
HEI (Health Effects Institute). (2004). Health effects of outdoor air pollution in developing countries of Asia: a
literature review. (Special report 15). Boston, MA: Health Effects Institute (HEI).
https://ntrl.ntis.gov/NTRL/dashboard/searchResults.xhtml?searchQuery=PB2005103805
HEI (Health Effects Institute). (2012). Effects of short-term exposure to air pollution on hospital admissions of
young children for acute lower respiratory infections in Ho Chi Minh City, Vietnam (pp. 5-72; discussion
73-83). (ISSN 1041-5505
Research Report 169). Boston, MA: Health Effects Institute, HEI Collaborative Working Group on Air Pollution,
Poverty, and Health in Ho Chi Minh City, https://www.healtheffects.org/publication/effects-short-term-
exposure-air-pollution-hospital-admissions-voung-children-acute-lower
Heinrich. U: Mohr. U: Fuhst. R: Brockmever. C. (1989). Investigation of a potential cotumorigenic effect of the
dioxides of nitrogen and sulfur, and of diesel-engine exhaust, on the respiratory tract of Syrian golden
hamsters (pp. 1-27). (Research report number26). Boston, MA: Health Effects Institute.
http://pubs.healtheffects.org/getfile.php?u=769
Helmerhorst. FM: Perquin. DA: Donker. D: Keirse. MJ. (2004). Perinatal outcome of singletons and twins after
assisted conception: a systematic review of controlled studies [Review]. B M J (Online) 328: 261-265.
http://dx.doi.org/10.1136/bmi.37957.560278.EE
R-22
-------
Henneberger. A: Zareba. W: Ibald-Mulli. A: Rtickerl. R: Cvrvs. J: Couderc. JP: Mvkins. B: Woelke. G: Wichmann.
HE: Peters. A. (2005). Repolarization changes induced by air pollution in ischemic heart disease patients.
Environ Health Perspect 113: 440-446. http://dx.doi.org/10.1289/ehp.7579
Henrotia JB: Besancenot. JP: Beiot. Y: Giroud. M. (2007). Short-term effects of ozone air pollution on ischaemic
stroke occurrence: A case-crossover analysis from a 10-year population-based study in Dijon, France.
Occup Environ Med 64: 439-445. http://dx.doi.org/10.1136/oem.2006.029306
Hertz-Picciotto. I: Cassadv. D: Lee. K: Bennett. DH: Ritz. B: Vogt. R. (2010). Study of Use of Products and
Exposure-Related Behaviors (SUPERB): study design, methods, and demographic characteristics of
cohorts. Environ Health 9. http://dx.doi.org/10.1186/1476-069X-9-54
HEW (U.S. Department of Health, Education and Welfare). (1969). Air quality criteria for sulfur oxides.
Washington, DC: National Air Pollution Control Administration.
Hidv. GM. (1994). Atmospheric sulfur and nitrogen oxides: Eastern North American source-receptor relationships.
In Atmospheric sulfur and nitrogen oxides: Eastern North American source-receptor relationships. San
Diego: Academic Press.
Hildebrandt. K: Rtickerl. R: Koenig. W: Schneider. A: Pitz. M: Heinrich. J: Marder. V: Frampton. M: Oberdorster.
G: Wichmann. HE: Peters. A. (2009). Short-term effects of air pollution: A panel study of blood markers in
patients with chronic pulmonary disease. Part Fibre Toxicol 6: 25. http://dx.doi.org/10.1186/1743-8977-6-
25
Hirsch. JA: Swensoa EW: Wanner. A. (1975). Tracheal mucous transport in beagles after long-term exposure to 1
ppm sulfur dioxide. Arch Environ Health 30: 249-253.
Hoek. G. (2003). Daily mortality and air pollution in The Netherlands [HEI], In Revised analyses of time-series
studies of air pollution and health (pp. 133-141). Boston, MA: Health Effects Institute.
http://pubs.healtheffects.org/getfile.php?u=21
Holnicki. P: Kaluszko. A: Trapp. W. (2016). An urban scale application and validation of the CALPUFF model.
Atmos Pollut Res 7: 393-402. http://dx.doi.Org/10.1016/i.apr.2015.10.016
Honninger. G: von Friedeburg. C: Piatt. U. (2004). Multi axis differential optical absorption spectroscopy (MAX-
DOAS). Atmos ChemPhys 4: 231-254.
Hopp. RJ: Bewtra. A: Nair. NM: Townlev. RG. (1985). The effect of age on methacholine response. J Allergy Clin
Immunol 76: 609-613. http://dx.doi.org/10.1016/0091-6749(85)90783-3
Hopp. RJ: Bewtra. AK: Nair. NM: Watt. GD: Townlev. RG. (1986). Methacholine inhalation challenge studies in a
selected pediatric population. Am Rev RespirDis 134: 994-998.
http://dx.doi.Org/10.1164/arrd.1986.134.5.994
Hoppel. WA: Caffrev. PF. (2005). Oxidation of S(IV) in sea-salt aerosol at high pH: ozone versus aerobic reaction. J
Geophys Res 110: D23202. http://dx.doi.org/10.1029/2005JD0Q6239
Horowitz. LW: Walters. S: Mauzerall. PL: Emmons. LK: Rasch. PJ: Granier. C: Tie. X: Lamarque. J. -F: Schultz.
MG: Tvndall. GS: Orlando. JJ: Brasseur. GP. (2003). A global simulation of tropospheric ozone and related
tracers: Description and evaluation of MOZART, version 2. J Geophys Res 108.
http://dx.doi.org/10.1029/2002JD0Q2853
Horstman. D: Roger. LJ: Kehrl. H: Hazucha. M. (1986). Airway sensitivity of asthmatics to sulfur dioxide. Toxicol
Ind Health 2: 289-298.
Horstmaa DH: Seal. E. Jr: Folinsbee. LJ: Ives. P: Roger. LJ. (1988). The relationship between exposure duration
and sulfur dioxide-induced bronchoconstriction in asthmatic subjects. AIHA J 49: 38-47.
http://dx.doi.org/10.108Q/15298668891379341
Hou. HY: Wang. D: Zou. XP: Yang. ZH: Li. TC: Chen. YO. (2014). Does ambient air pollutants increase the risk of
fetal loss? A case-control study. Arch Gynecol Obstet 289: 285-291. http://dx.doi.org/10.1007/sQ0404-013-
2962-1
R-23
-------
Howden. LM: Mever. JA. (2011). Age and sex composition: 2010. (2010 Census Briefs, C2010BR-03).
Washington, DC: U.S. Department of Commerce, Economics and Statistics Administration, U.S. Census
Bureau. htto://www.census.gov/prod/cen2010/briefs/c2010br-03.pdf
Hovert. PL: Xu. J. (2012). Deaths: Preliminary data for 2011. National Vital Statistics Reports 61: 1-51.
Hsieh. YL: Yang. YH: Wu. TN: Yang. CY. (2010). Air pollution and hospital admissions for myocardial infarction
in a subtropical city: Taipei, Taiwan. J Toxicol Environ Health A 73: 757-765.
http://dx.doi.org/10.1080/152873910Q3684789
Huang. CC: Wen. HJ: Chea PC: Chiang. TL: Lia SJ: Guo. YL. (2015a). Prenatal air pollutant exposure and
occurrence of atopic dermatitis. Br J Dermatol 173: 981-988. http://dx.doi.org/10. Ill 1/bid. 14039
Huang. L: Zhao. Y: Li. H: Chea Z. (2015b). Kinetics of heterogeneous reaction of sulfur dioxide on authentic
mineral dust: Effects of relative humidity and hydrogen peroxide. Environ Sci Technol 49: 10797-10805.
http://dx.doi.org/10.1021/acs.est.5b03930
Huang. W: Zhu. T: Pan. X: Hu. M: Lu. SE: Lin. Y: Wang. T: Zhang. Y: Tang. X. (2012). Air pollution and
autonomic and vascular dysfunction in patients with cardiovascular disease: Interactions of systemic
inflammation, overweight, and gender. Am J Epidemiol 176: 117-126.
http://dx.doi.org/10.1093/aie/kwr511
Hudda. N: Kostenidou. E: Sioutas. C: Delfino. RJ: Fruia SA. (2011). Vehicle and driving characteristics that
influence in-cabin particle number concentrations. Environ Sci Technol 45: 8691-8697.
http://dx.doi.org/10.1021/es202025m
Hung. HM: Hoffmann. MR. (2015). Oxidation of gas-phase S02 on the surfaces of acidic microdroplets:
Implications for sulfate and sulfate radical anion formation in the atmospheric liquid phase. Environ Sci
Technol 49: 13768-13776. http://dx.doi.org/10.1021/acs.est.5b01658
Hurley. PJ. (2006). An evaluation and inter-comparison of AUSPLUME, AERMOD and TAPM for seven field
datasets of point source dispersion. Clean Air Environ Qual 40: 45-50.
Hwang. BF: Chea YH: Lia YT: Wu. XT: Leo Lee. Y. (2015a). Relationship between exposure to fine particulates
and ozone and reduced lung function in children. Environ Res 137: 382-390.
http://dx.doi.Org/10.1016/i.envres.2015.01.009
Hwang. BF: Jaakkola. JJ. (2008). Ozone and other air pollutants and the risk of oral clefts. Environ Health Perspect
116: 1411-1415. http://dx.doi.org/10.1289/ehp. 11311
Hwang. BF: Lee. YL: Jaakkola. JJ. (2011). Air pollution and stillbirth: A population-based case-control study in
Taiwan. Environ Health Perspect 119: 1345-1349. http://dx.doi.org/10.1289/ehp. 1003056
Hwang. BF: Lee. YL: Jaakkola. JJ. (2015b). Air Pollution and the Risk of Cardiac Defects: A Population-Based
Case-Control Study. Medicine (Baltimore) 94: el883. http://dx.doi.org/10.1097/MD.00000000000Q1883
Hwang. SS: Kang. S: Lee. JY: Lee. JS: Kim. HJ: Haa SK: Yim. JJ. (2014). Impact of outdoor air pollution on the
incidence of tuberculosis in the Seoul metropolitan area, South Korea. Korean J Intern Med 29: 183-190.
http://dx.doi.Org/10.3904/kiim.2014.29.2.183
ICRP (International Commission on Radiological Protection). (1994). Human respiratory tract model for
radiological protection: A report of a task group of the International Commission on Radiological
Protection. ICRP Publication 66. New York, NY: Pergamon Press.
ICRP (International Commission on Radiological Protection). (2002). Basic anatomical and physiological data for
use in radiological protection: Reference values (pp. 1-277). (ICRP Publication 89). New York, NY:
Pergamon Press. http://dx.doi.org/10.1016/S0146-6453(03)00002-2
Ierodiakonou. D: Zanobetti. A: Coull. BA: Mellv. S: Postma. PS: Boezen. HM: Vonk. JM: Williams. PV: Shapiro.
GG: Mckone. EF: Hallstrand. TS: Koenig. JO: Schildcrout. JS: Lumlev. T: Fuhlbrigge. AN: Koutrakis. P:
Schwartz. J: Weiss. ST: Gold. PR: Group. CAMPR. (2015). Ambient air pollution, lung function, and
airway responsiveness in asthmatic children. J Allergy Clin Immunol 137: 390-399.
http://dx.doi.Org/10.1016/i.iaci.2015.05.028
R-24
-------
IOM (Institute of Medicine). (2005). DRI, dietary reference intakes for water, potassium, sodium, chloride, and
sulfate. Washington, DC: National Academies Press. https://www.nap.cdu/rcad/10925/chapter/1 #ii
IOM (Institute of Medicine). (2007). Preterm birth: Causes, consequences, and prevention. In Preterm birth: Causes,
consequences, and prevention. Washington, DC: National Academies Press (US).
http://dx.doi.org/10.17226/11622
Irwia JS. (2014). A suggested method for dispersion model evaluation. J Air Waste Manag Assoc 64: 255-264.
http://dx.doi.org/10.1080/10962247.2Q13.833147
Isaacs. K. (2014). The consolidated human activity database - master version (CHAD-Master) technical
memorandum. Washington, DC: U.S. Environmental Protection Agency, National Exposure Research
Laboratory, https://www.epa. gov/sites/production/files/2015-02/documents/chadmaster 091814 1 .pdf
Isaacs. K: McCurdv. T: Glen. G: Nvsewander. M: Errickson. A: Forbes. S: Graham. S: McCurdv. L: Smith. L:
Tulve. N: Vallero. D. (2013). Statistical properties of longitudinal time-activity data for use in human
exposure modeling. J Expo Sci Environ Epidemiol 23: 328-336. http://dx.doi.org/10.1038/ies.2012.94
Isakov. V: Venkatram A: Touma. JS: Koracin. D: Otte. TL. (2007). Evaluating the use of outputs from
comprehensive meteorological models in air quality modeling applications. Atmos Environ 41: 1689-1705.
http://dx.doi.Org/10.1016/i.atmosenv.2006.10.043
Ishigami. A: Kikuchi. Y: Iwasawa. S: Nishiwaki. Y: Takebavashi. T: Tanaka. S: Omae. K. (2008). Volcanic sulfur
dioxide and acute respiratory symptoms onMiyakejima island. Occup Environ Med 65: 701-707.
http://dx.doi.org/10.1136/oem.2007.033456
Ito. K: Mathes. R: Ross. Z: Nadas. A: Thurston. G: Matte. T. (2011). Fine particulate matter constituents associated
with cardiovascular hospitalizations and mortality in New York City. Environ Health Perspect 119: 467-
473. http://dx.doi.org/10.1289/ehp. 1002667
Ito. K: Thurston. GD: Silvermaa RA. (2007). Characterization of PM2.5, gaseous pollutants, and meteorological
interactions in the context of time-series health effects models. J Expo Sci Environ Epidemiol 17: S45-S60.
http://dx.doi.org/10.1038/si.ies.7500627
Ito. T: Ohvama. KI: Kusano. T: Usuda. Y: Nozawa. A: Havashi. H: Ohii. H: Kitamura. H: Kanisawa. M. (1997).
Pulmonary endocrine cell hyperplasia and papilloma in rats induced by intratracheal injections of extract
from particulate air pollutants. Exp Toxicol Pathol 49: 65-70. http://dx.doi.org/10.1016/S0940-
2993(97)80066-8
Iwasawa. S: Kikuchi. Y: Nishiwaki. Y: Nakano. M: Michikawa. T: Tsuboi. T: Tanaka. S: Uemura. T: Ishigami. A:
Nakashima. H: Takebavashi. T: Adachi. M: Morikawa. A: Maruvama. K: Kudo. S: Uchivama. I: Omae. K.
(2009). Effects of S02 on respiratory system of adult Miyakejima resident 2 years after returning to the
island. J Occup Health 51: 38-47. http://dx.doi.org/10.1539/ioh.L8075
Iwasawa. S: Nakano. M: Tsuboi. T: Kochi. T: Tanaka. S: Katsunuma. T: Morikawa. A: Omae. K. (2015). Effects of
sulfur dioxide on the respiratory system of Miyakejima child residents 6years after returning to the island.
Int Arch Occup Environ Health 88: 1111-1118. http://dx.doi.org/10.1007/s00420-Q15-1037-v
Jackson. RA: Gibson. KA: Wu. YW: Croughan. MS. (2004). Perinatal outcomes in singletons following in vitro
fertilization: a meta-analysis. Obstet Gynecol 103: 551-563.
http://dx.doi.Org/10.1097/01.AOG.0000114989.84822.51
Jacob. DJ. (1999). Introduction to atmospheric chemistry. Princeton, NJ: Princeton University Press.
http://press.princeton.edu/titles/6767.html
Jacobsoa MZ. (2002). Atmospheric pollution: History, science, and regulation. In Atmospheric Pollution: History,
Science, and Regulation. New York: Cambridge University Press.
http://www.cambridge.org/us/academic/subiects/earth-and-environmental-science/atmospheric-science-
and-meteorology/atmospheric-pollution-history-science-and-regulation
Jaffe. DH: Singer. ME: Rimm. AA. (2003). Air pollution and emergency department visits for asthma among Ohio
Medicaid recipients, 1991-1996. EnvironRes 91: 21-28. http://dx.doi.org/10.1016/S0Q 13-9351(02)00004-
X
R-25
-------
Jalaludia B: Khalai. B: Sheppeard. V: Morgan. G. (2008). Air pollution and ED visits for asthma in Australian
children: A case-crossover analysis. Int Arch Occup Environ Health 81: 967-974.
http://dx.doi.org/10.1007/s00420-007-029Q-0
Jalaludia B: Mannes. T: Morgan. G: Lincola D: Sheppeard. V: Corbett. S. (2007). Impact of ambient air pollution
on gestational age is modified by season in Sydney, Australia. Environ Health 6: 16.
http://dx.doi.org/10.1186/1476-Q69X-6-16
Jalaludia B: Morgaa G: Lincola D: Sheppeard. V: Simpson. R: Corbett. S. (2006). Associations between ambient
air pollution and daily emergency department attendances for cardiovascular disease in the elderly (65+
years), Sydney, Australia. J Expo Sci Environ Epidemiol 16: 225-237.
http://dx.doi.org/10.1038/si .iea.7500451
James. PS: Stidlev. CA: Lambert. WE: Chick. TW: Mermier. CM: Samet. JM. (1997). Oronasal distribution of
ventilation at different ages. Arch Environ Occup Health 52: 118-123.
http://dx.doi.org/10.1080/000398997096Q2874
Japar. SM: Wallington. TJ: Andino. JM: Ball. JC. (1990). Atmospheric reactivity of gaseous dimethyl sulfate.
Environ Sci Technol 24: 313-315. http://dx.doi.org/10.1021/esQ0073a004
Jedrvchowski. W: Flak. E: Mroz. E. (1999). The adverse effect of low levels of ambient air pollutants on lung
function growth in preadolescent children. Environ Health Perspect 107: 669-674.
Jerrett. M: Burnett. RT: Pope. CA. Ill: Ito. K: Thurstoa G: Krewski. D: Shi. Y: Calle. E: Thua M. (2009). Long-
term ozone exposure and mortality. N Engl J Med 360: 1085-1095.
http://dx.doi.org/10.1056/NEJMoa0803894
Jerrett. M: Burnett. RT: Willis. A: Krewski. D: Goldberg. MS: Deluca. P: Finkelsteia N. (2003). Spatial analysis of
the air pollution-mortality relationship in the context of ecologic confounders. J Toxicol Environ Health A
66: 1735-1777. http://dx.doi.org/10.1080/15287390306438
Jiao. W. an: Frev. HC: Cao. Y. e. (2012). Assessment of Inter-Individual, Geographic, and Seasonal Variability in
Estimated Human Exposure to Fine Particles. Environ Sci Technol 46: 12519-12526.
http://dx.doi.org/10.1021/es302803g
Johns. DO: Svendsgaard. D: Lina WS. (2010). Analysis of the concentration-respiratory response among asthmatics
following controlled short-term exposures to sulfur dioxide. Inhal Toxicol 22: 1184-1193.
http://dx.doi.org/10.3109/08958378.2010.53522Q
Johnson-Winters. K: Tollin. G: Enemark. JH. (2010). Elucidating the catalytic mechanism of sulfite oxidizing
enzymes using structural, spectroscopic, and kinetic analyses [Review]. Biochemistry 49: 7242-7254.
http://dx.doi.org/10.1021/bil008485
Johnson. D: Lewia AG: Marston. G. (2001). The effect of Criegee-intermediate scavengers on the OH yield from
the reaction of ozone with 2-methylbut-2-ene. J Phys Chem A 105: 2933-2935.
http://dx.doi.org/10.1021/ip003975e
Johnson. JY: Rowe. BH: Villeneuve. PJ. (2010). Ecological analysis of long-term exposure to ambient air pollution
and the incidence of stroke in Edmonton, Alberta, Canada. Stroke 41: 1319-1325.
http://dx.doi.org/10.1161/STROKEAHA. 110.580571
Jones. SL: Kittelsoa J: Cowaa JO: Flannerv. EM: Hancox. RJ: McLachlan. CR: Taylor. DR. (2001). The predictive
value of exhaled nitric oxide measurements in assessing changes in asthma control. Am J Respir Crit Care
Med 164: 738-743. http://dx.doi.org/10.1164/airccm. 164.5.2012125
Jorres. R: Magnussea H. (1990). Airways response of asthmatics after a 30 min exposure, at resting ventilation, to
0.25 ppmN02 or 0.5 ppm S02. Eur Respir J 3: 132-137.
Jurewicz. J: Radwan. M: Sobala. W: Polariska. K: Radwan. P: Jakubowski. L: Ulanska. A: Hanke. W. (2014). The
relationship between exposure to air pollution and sperm disomy. Environ Mol Mutagen 56: 50-59.
http://dx.doi.org/10.1002/em.21883
R-26
-------
Kaa H: Wong. CM: Vichit-Vadakaa N: Oiaa Z. (2010). Short-term association between sulfur dioxide and daily
mortality: The Public Health and Air Pollution in Asia (PAPA) study. Environ Res 110: 258-264.
http://dx.doi.Org/10.1016/i.envres.2010.01.006
Kanaroglou. PS: Adams. MP: De Luca. PF: Corr. D: Sohel. N. (2013). Estimation of sulfur dioxide air pollution
concentrations with a spatial autoregressive model. Atmos Environ 79: 421-427.
http://dx.doi.Org/10.1016/i.atmosenv.2013.07.014
Kane. LE: Barrow. CS: Alarie. Y. (1979). A short-term test to predict acceptable levels of exposure to airborne
sensory irritants. Am Ind Hyg Assoc J 40: 207-229. http://dx.doi.org/10.108Q/15298667991429516
Kang. SH: Heo. J: Oh. IY: Kim. .1: Lim. WH: Cho. Y: Choi. EK: Yi. SM: Do Shin. S: Kim. H: Oh. S. (2016).
Ambient air pollution and out-of-hospital cardiac arrest. Int J Cardiol 203: 1086-1092.
http://dx.doi.Org/10.1016/i.iicard.2015.ll.100
Kara. E: Ozdilek. HG: Kara. EE. (2013). Ambient air quality and asthma cases in Nigde, Turkey. Environ Sci Pollut
Res Int 20: 4225-4234. http://dx.doi.org/10.1007/sll356-Q12-1376-0
Karamchandani. P: Viiavaraghavaa K: Chea S: Balmori-Bronsoa R: Knipping. EM. (2010). Development and
application of a parallelized version of the advanced modeling system for transport, emissions, reactions
and deposition of atmospheric matter (AMSTERDAM): 1. Model performance evaluation and impacts of
plume-in-grid treatment. Atmos Pollut Res 1 SI: 260-270. http://dx.doi.org/10.5094/APR.2010.Q33
Karr. CJ: Demers. PA: Koehoora MW: Lencar. CC: Tamburic. L: Brauer. M. (2009). Influence of ambient air
pollutant sources on clinical encounters for infant bronchiolitis. Am J Respir Crit Care Med 180: 995-1001.
http://dx.doi.org/10.1164/rccm.200901-0117QC
Katanoda. K: Sobue. T: Satoh. H: Taiima. K: Suzuki. T: Nakatsuka. H: Takezaki. T: Nakavama. T: Nitta. H:
Tanabe. K: Tominaga. S. (2011). An association between long-term exposure to ambient air pollution and
mortality from lung cancer and respiratory diseases in Japan. J Epidemiol 21: 132-143.
http://dx.doi.org/10.2188/iea.JE20100Q98
Katsouvanni. K: Touloumi. G: Spix. C: Schwartz. J: Balducci. F: Medina. S: Rossi. G: Woitvniak. B: Sunver. J:
Bacharova. L: Schouten. JP: Ponka. A: Anderson. HR. (1997). Short term effects of ambient sulphur
dioxide and particulate matter on mortality in 12 European cities: Results from time series data from the
APHEA project. B M J 314: 1658-1663. http://dx.doi.org/10.1136/bmi.314.7095.1658
Kehrl. HR: Roger. LJ: Hazucha. MJ: Horstman. PH. (1987). Differing response of asthmatics to sulfur dioxide
exposure with continuous and intermittent exercise. Am J Respir Crit Care Med 135: 350-355.
Kelishadi. R: Mirghaffari. N: Poursafa. P: Gidding. SS. (2009). Lifestyle and environmental factors associated with
inflammation, oxidative stress and insulin resistance in children. Atherosclerosis 203: 311-319.
http://dx.doi.Org/10.1016/i.atherosclerosis.2008.06.022
Kerr. WJ: McWilliam JS: Linder-Aronson. S. (1989). Mandibular form and position related to changed mode of
breathing~a five-year longitudinal study. Angle Orthod 59: 91-96.
Khafaie. MA: Salvi. SS: Oiha. A: Khafaie. B: Gore. SS: Yainik. CS. (2013). Systemic inflammation (C-reactive
protein) in type 2 diabetic patients is associated with ambient air pollution in Pune City, India. Diabetes
Care 36: 625-630. http://dx.doi.org/10.2337/dcl2-0388
Kharitonov. SA: Barnes. PJ. (2000). Clinical aspects of exhaled nitric oxide [Review]. Eur Respir J 16: 781-792.
Khoder. MI. (2002). Atmospheric conversion of sulfur dioxide to particulate sulfat and nitrogen dioxide to
particulate nitrate and gaseous nitric acid in an urban area. Chemosphere 49: 675-684.
http://dx.doi.org/10.1016/S0045-6535(02)00391-0
Khokhar. MF: Piatt. U: Wagner. T. (2008). Temporal trends of anthropogenic S02 emitted by non-ferrous metal
smelters in Peru and Russia estimated from Satellite observations. Atmos Chem Phys Discuss 8: 17393-
17422. http://dx.doi.org/10.5194/acpd-8-17393-2008
Kikuva. M: Hozawa. A: Ohokubo. T: Tsuii. I: Michimata. M: Matsubara. M: Ota. M: Nagai. K: Araki. T: Satoh. H:
Ito. S: Hisamichi. S: Imai. Y. (2000). Prognostic significance of blood pressure and heart rate variabilities:
the Ohasama study. Hypertension 36: 901-906. http://dx.doi.Org/10.1161/01.HYP.36.5.901
R-27
-------
Kilkenny. C: Browne. WJ: Cuthill. 1C: Emerson. M: Altman. DG. (2010). Improving bioscience research reporting:
The ARRIVE guidelines for reporting animal research [Review]. PLoS Biol 8: el000412.
http://dx.doi.org/10.1371/iournal.pbio.1000412
Kim. H: Park. Y: Park. K: Yoo. B. (2016a). Association between pollen risk indexes, air pollutants, and allergic
diseases in Korea. Osong Public Health Res Perspect 7: 172-179.
http://dx.doi.Org/10.1016/i.phrp.2016.04.003
Kim. JH: Hong. YC. (2012). GSTM1, GSTT1, and GSTP1 polymorphisms and associations between air pollutants
and markers of insulin resistance in elderly Koreans. Environ Health Perspect 120: 1378-1384.
http://dx.doi.org/10.1289/ehp. 1104406
Kim. KH: Kim. MY. (2001). Comparison of an open path differential optical absorption spectroscopy system and a
conventional in situ monitoring system on the basis of long-term measurements of S02, N02, and 03.
Atmos Environ 35: 4059-4072. http://dx.doi.org/10.1016/S1352-2310(01)00216-3
Kim. KN: Kim. JH: Jung. K: Hong. YC. (2016b). Associations of air pollution exposure with blood pressure and
heart rate variability are modified by oxidative stress genes: A repeated-measures panel among elderly
urban residents. Environ Health 15: 47. http://dx.doi.org/10.1186/sl2940-016-0130-3
Kim. PS: Jacob. DJ: Fisher. JA: Travis. K: Yu. K: Zhu. L: Yantosca. RM: Sulprizio. MP: Jimenez. JL: Campuzano-
Jost. P: Frovd. KD: Liao. J: Hair. JW: Fenn. MA: Butler. CF: Wagner. NL: Gordon. TP: Welti. A:
Wennberg. PO: Crounse. JD: St. Clair. JM: Teng. AP: Millet. DB: Schwarz. JP: Markovic. MZ: Perring.
AE. (2015). Sources, seasonality, and trends of southeast US aerosol: An integrated analysis of surface,
aircraft, and satellite observations with the GEOS-Chem chemical transport model. Atmos Chem Phys 15:
10411-10433. http://dx.doi.org/10.5194/acp-15-10411-2015
Kim. T: Lee. K: Yang. W: Yu. SD. (2012). A new analytical method for the classification of time-location data
obtained from the global positioning system (GPS). J Environ Monit 14: 2270-2274.
http://dx.doi.org/10.1039/c2em30190c
Kimbell. JS: Miller. FJ. (1999). Regional respiratory-tract absorption of inhaled reactive gases: A modeling
approach. In DE Gardner; JD Crapo; RO McClellan (Eds.), Toxicology of the lung (3rd ed ed., pp. 557-
598). New York, NY: Taylor and Francis.
Klepeis. NE. (1999). An introduction to the indirect exposure assessment approach: Modeling human exposure
using microenvironmental measurements and the recent National Human Activity Pattern Survey [Review].
Environ Health Perspect 107: 365-374. http://dx.doi.org/10.2307/3434429
Klepeis. NE: Nelson. WC: Ott. WR: Robinson. JP: Tsang. AM: Switzer. P: Behar. JV: Hera SC: Engelmann. WH.
(2001). The National Human Activity Pattern Survey (NHAPS): A resource for assessing exposure to
environmental pollutants. J Expo Anal Environ Epidemiol 11: 231-252.
http://dx.doi.org/10.1038/si.iea.750Q165
Klepeis. NE: Tsang. AM: Behar. JV. (1996). Analysis of the national human activity pattern survey (NHAPS)
respondents from a standpoint of exposure assessment [EPA Report]. (EPA/600/R-96/074). Washington,
DC: U.S. Environmental Protection Agency.
http://exposurescience.org/pub/reports/NHAPS Reportl.pdf#....Local SettingsTemporary Internet
FilesContent.Outlook3JQ221FPB_Approaches_Population_Tables.docx
Klimisch. HJ: Andreae. M: Tillmann. U. (1997). A systematic approach for evaluating the quality of experimental
toxicological and ecotoxicological data. Regul Toxicol Pharmacol 25: 1-5.
http://dx.doi.org/10.1006/rtph. 1996.1076
Kloster. S: Six. KD: Feichter. J: Maier-Reimer. E: Roeckner. E: Wetzel. P: Stier. P: Esch. M. (2007). Response of
dimethylsulfide (DMS) in the ocean and atmosphere to global warming. J Geophys Res 112: G03005.
http://dx.doi.org/10.1029/2006JG00Q224
Knowledge Networks. (2009). Field report: National-scale activity survey (NSAS). Research Triangle Park, NC:
Research Triangle Institute.
R-28
-------
Ko. FW: Tam. W: Wong. TW: Chan. DP: Tung. AH: Lai. CKW: Hui. DSC. (2007a). Temporal relationship between
air pollutants and hospital admissions for chronic obstructive pulmonary disease in Hong Kong. Thorax 62:
780-785. http://dx.doi.org/10.1136/thx.2006.Q76166
Ko. FWS: Tam W: Wong. TW: Lai. CKW: Wong. GWK: Leung. TF: Ng. SSS: Hui. DSC. (2007b). Effects of air
pollution on asthma hospitalization rates in different age groups in Hong Kong. Clin Exp Allergy 37: 1312-
1319. http://dx.doi.org/10.1111/i .1365-2222.2007.02791 ,x
Koenig. JO: Covert. PS: Hanlev. OS: Van Belle. G: Piersoa WE. (1990). Prior exposure to ozone potentiates
subsequent response to sulfur dioxide in adolescent asthmatic subjects. Am J Respir Crit Care Med 141:
377-380. http://dx.doi.Org/10.1164/airccm/141.2.377
Koenig. JO: Dumler. K: Rebolledo. V: Williams. PV: Piersoa WE. (1992). Theophylline mitigates the
bronchoconstrictor effects of sulfur dioxide in subjects with asthma. J Allergy Clin Immunol 89: 789-794.
http://dx.doi.org/10.1016/0091-6749(92)90432-2
Koenig. JO: Marshall. SG: Horike. M: Shapiro. GG: Furukawa. CT: Bierman. CW: Piersoa WE. (1987). The effects
of albuterol on sulfur dioxide-induced bronchoconstriction in allergic adolescents. J Allergy Clin Immunol
79: 54-58. http://dx.doi.org/10.1016/S0091-6749(87)80016-7
Koenig. JO: Marshall. SG: van Belle. G: McManus. MS: Bierman. CW: Shapiro. GG: Furukawa. CT: Pierson. WE.
(1988). Therapeutic range cromolyn dose-response inhibition and complete obliteration of S02-induced
bronchoconstriction in atopic adolescents. J Allergy Clin Immunol 81: 897-901.
Koenig. JO: Piersoa WE: Frank. R. (1980). Acute effects of inhaled S02 plus NaCl droplet aerosol on pulmonary
function in asthmatic adolescents. Environ Res 22: 145-153.
Koenig. JO: Piersoa WE: Horike. M: Frank. R. (1981). Effects of S02 plus NaCl aerosol combined with moderate
exercise on pulmonary function in asthmatic adolescents. Environ Res 25: 340-348.
Koenig. JO: Pierson. WE: Horike. M: Frank. R. (1983). A comparison of the pulmonary effects of 0.5 ppm versus
1.0 ppm sulfur dioxide plus sodium chloride droplets in asthmatic adolescents. J Toxicol Environ Health
11: 129-139. http://dx.doi.org/10.1080/1528739830953Q327
Koken. PJM: Piver. WT: Ye. F: Elixhauser. A: Olsea LM: Portier. CJ. (2003). Temperature, air pollution, and
hospitalization for cardiovascular diseases among elderly people in Denver. Environ Health Perspect 111:
1312-1317.
Koutrakis. P: Wolfsoa JM: Slater. JL: Brauer. M: Spengler. JD: Stevens. RK: Stone. CL. (1988). Evaluation of an
annular denuder/filter pack system to collect acidic aerosols and gases. Environ Sci Technol 22: 1463-
1468. http://dx.doi.org/10.1021/esQ0177a013
Kraft. J: van Eldik. R. (1989a). Kinetics and mechanism of the iron(III)-catalyzed autoxidation of sulfur(IV) oxides
in aqueous solution. 1. Formation of transient iron(III)-sulfur(IV) complexes. Inorg Chem 28: 2297-2305.
http://dx.doi.org/10.1021/ic0031 laOl 1
Kraft. J: van Eldik. R. (1989b). Kinetics and mechanism of the iron(III)-catalyzed autoxidation of sulfur(IV) oxides
in aqueous solution. 2. Decomposition of transient iron(III)-sulfur(IV) complexes. Inorg Chem 28: 2306-
2312. http://dx.doi.org/10.1021/ic0031 la012
Krewski. D: Burnett. RT: Goldberg. MS: Hoover. K: Siemiatvcki. J: Jerrett. M: Abrahamowicz. M: White. WH.
(2000). Reanalysis of the Harvard Six Cities study and the American Cancer Society study of particulate air
pollution and mortality. Cambridge, MA: Health Effects Institute.
http://pubs.healtheffects.org/view.php?id=6
Krewski. D: Jerrett. M: Burnett. RT: Ma. R: Hughes. E: Shi. Y: Turner. MC: Pope. CA. Ill: Thurstoa G: Calle. EE:
Thun. MJ: Beckerman. B: Deluca. P: Finkelstein. N: Ito. K: Moore. DK: Newbold. KB: Ramsay. T: Ross.
Z: Shin. H: Tempalski. B. (2009). Extended follow-up and spatial analysis of the American Cancer Society
study linking particulate air pollution and mortality (pp. 5-114; discussion 115-136). (ISSN 1041-5505,
HEI Research Report 140). Boston, MA: Health Effects Institute.
http://pubs.healtheffects.org/view.php?id=315
R-29
-------
Kumar. N. (2012). Uncertainty in the relationship between criteria pollutants and low birth weight in Chicago.
Atmos Environ49: 171-179. http://dx.doi.Org/10.1016/i.atmosenv.2011.12.001
La Rovere. MT: Pinna. GD: Maestri. R: Mortara. A: Capomolla. S: Febo. O: Ferrari. R: Franchini. M: Gnemmi. M:
Qpasich. C: Riccardi. PG: Traversi. E: Cobelli. F. (2003). Short-term heart rate variability strongly predicts
sudden cardiac death in chronic heart failure patients. Circulation 107: 565-570.
http://dx.doi.org/10.1161/01.CIR.0000Q47275.25795.17
Lane. KJ: Scammell. MK: Lew. JI: Fuller. CH: Parambi. R. on: Zamore. W. ig: Mwamburi. M: Brugge. D. (2013).
Positional error and time-activity patterns in near-highway proximity studies: an exposure misclassification
analysis. Environ Health 12: 75. http://dx.doi.org/10.1186/1476-069X-12-75
Langner. J: Rodhe. H. (1991). A global three-dimensional model of the tropospheric sulfur cycle. J Atmos Chem 13:
225-263.
Laskia S: Kuschner. M: Sellakumar. A: Katz. GV. (1976). Combined carcinogen-irritant animal inhalation studies.
In EF Aharonson; A Ben-David; MA Klingberg (Eds.), Air pollution and the lung: proceedings of the
twentieth annual (pp. 190-213). New York, NY: John Wiley & Sons.
Lawther. PJ. (1955). Effect of inhalation of sulphur dioxide on respiration and pulse-rate in normal subjects. Lancet
269: 745-748.
Lawther. PJ: MacFarlane. AJ: Waller. RE: Brooks. AGF. (1975). Pulmonary function and sulphur dioxide, some
preliminary findings. Environ Res 10: 355-367. http://dx.doi.org/10.1016/0013-9351(75)90031-6
Lazarus. SC: Boushev. HA: Fahv. JV: Chinchilli. VM: Lemanske. RF: Sorkness. CA: Kraft. M: Fish. JE: Peters. SP:
Craig. T: Drazea JM: Ford. JG: Israel. E: Martia RJ: Mauger. EA: Nachman. SA: Spahn. JD: Szefler. SJ:
Asthma Clinical Research Network for the National Heart. L. (2001). Long-acting beta2-agonist
monotherapy vs continued therapy with inhaled corticosteroids in patients with persistent asthma: a
randomized controlled trial. JAMA 285: 2583-2593. http://dx.doi.org/10.1001/iama.285.20.2583
Lazarus. SC: Wong. HH: Watts. MJ: Boushev. HA: Lavins. BJ: Minkwitz. MC. (1997). The leukotriene receptor
antagonist zafirlukast inhibits sulfur dioxide-induced bronchoconstriction in patients with asthma. Am J
Respir Crit Care Med 156: 1725-1730. http://dx.doi.Org/10.1164/airccm.156.6.9608006
Le. HO: Batterman. SA: Wirth. JJ: Wahl. RL: Hoggatt. KJ: Sadeghneiad. A: Hultin. ML: Depa. M. (2012). Air
pollutant exposure and preterm and term small-for-gestational-age births in Detroit, Michigan: Long-term
trends and associations. Environ Int 44: 7-17. http://dx.doi.Org/10.1016/i.envint.2012.01.003
Le Souef. PN: Sears. MR: Sherrill. D. (1995). The effect of size and age of subject on airway responsiveness in
children. Am J Respir Crit Care Med 152: 576-579. http://dx.doi.Org/10.1164/airccm.152.2.7633710
Lee. BE: Ha. EH: Park. HS: Kim. YJ: Hong. YC: Kim. H: Lee. JT. (2003). Exposure to air pollution during different
gestational phases contributes to risks of low birth weight. Hum Reprod 18: 638-643.
http://dx.doi.org/10.1093/humrep/degl02
Lee. C: Martia RV: van Donkelaar. A: Lee. H: Dickerson. RR: Hains. JC: Krotkov. N: Richter. A: Vinnikov. K:
Schwab. JJ. (2011a). S02 emissions and lifetimes: Estimates from inverse modeling using in situ and
global, space-based (SCIAMACHY and OMI) observations. J Geophys Res 116: D06304.
http://dx.doi.org/10.1029/2010JDQ14758
Lee. CL: Brimblecombe. P. (2016). Anthropogenic contributions to global carbonyl sulfide, carbon disulfide and
organosulfides fluxes. Earth Sci Rev 160: 1-18. http://dx.doi.Org/10.1016/i.earscirev.2016.06.005
Lee. HP. on: Yoo. JW. oo: Kang. M. inK: Kang. J. iS: Jung. JH: Oh. KJ. (2014). Evaluation of concentrations and
source contribution of PM10 and S02 emitted from industrial complexes in Ulsan, Korea: Interfacing of
the WRF-CALPUFF modeling tools. Atmos Pollut Res 5: 664-676.
http://dx.doi.org/10.5094/APR.2014.076
Lee. LY: Widdicombe. JG. (2001). Modulation of airway sensitivity to inhaled irritants: role of inflammatory
mediators [Review]. Environ Health Perspect 4: 585-589.
R-30
-------
Lee. ML: Later. DW: Rollins. DK: Eatough. DJ: Hansea LP. (1980). Dimethyl and monomethyl sulfate: Presence
in coal fly ash and airborne particulate matter. Science 207: 186-188.
http://dx.doi.org/10.1126/science.7350652
Lee. PC: Talbott. EO: Roberts. JM: Catov. JM: Bilonick. RA: Stone. RA: Sharma. RK: Ritz. B. (2012). Ambient air
pollution exposure and blood pressure changes during pregnancy. Environ Res 117: 46-53.
http://dx.doi.Org/10.1016/i.envres.2012.05.011
Lee. PC: Talbott. EO: Roberts. JM: Catov. JM: Sharma. RK: Ritz. B. (201 lb). Particulate air pollution exposure and
C-reactive protein during early pregnancy. Epidemiology 22: 524-531.
http://dx.doi.org/10.1097/EDE.0b013e31821c6c58
Lee. YL: Wang. WH: Lu. CW: Lia YH: Hwang. BF. (201 lc). Effects of ambient air pollution on pulmonary
function among schoolchildren. Int J Hyg Environ Health 214: 369-375.
http://dx.doi.Org/10.1016/i.iiheh.2011.05.004
Lee. YP. (2015). Perspective: Spectroscopy and kinetics of small gaseous Criegee intermediates. J ChemPhys 143:
020901. http://dx.doi.org/10.1063/1.4923165
Leem. JH: Kaplan. BM: Shim. YK: Pohl. HR: Gotwav. CA: Bullard. SM: Rogers. JF: Smith. MM: Tvlenda. CA.
(2006). Exposures to air pollutants during pregnancy and preterm delivery. Environ Health Perspect 114:
905-910. http://dx.doi.org/10.1289/ehp.8733
Legro. RS: Sauer. MY: Mottla. GL: Richter. KS: Li. X: Dodson. WC: Liao. D. (2010). Effect of air quality on
assisted human reproduction. Hum Reprod 25: 1317-1324. http://dx.doi.org/10.1093/humrep/dea021
Leiberman. A: Ohki. M: Forte. V: Fraschetti. J: Cole. P. (1990). Nose/mouth distribution of respiratory airflow in
'mouthbreathing' children. Acta Otolaryngol 109: 454-460. http://dx.doi.org/10.3109/000164890Q9125169
Leidos Inc. (2016). Air quality monitoring program at the port of Los Angeles. Year eleven data summary, May
2015 - April 2016. San Pedro, CA: Port of Los Angeles, Environmental Planning Division.
Leiter. JC: Baker. GL. (1989). Partitioning of ventilation between nose and mouth: The role of nasal resistance. Am
J Orthod Dentofacial Orthop 95: 432-438. http://dx.doi.org/10.1016/0889-5406(89)90305-3
Lenters. V: Uiterwaal. CS: Beelen. R: Bots. ML: Fischer. P: Brunekreef. B: Hoek. G. (2010). Long-term exposure to
air pollution and vascular damage in young adults. Epidemiology 21:512-520.
http://dx.doi.org/10.1097/EDE.0b013e3181dec3a7
Leppanea S: Anttila. P: Lattila. H: Makkonea U. (2005). Long-term comparison of filter method and sensitive
analyser in monitoring of sulphur dioxide. Atmos Environ 39: 2683-2693.
http://dx.doi.Org/10.1016/i.atmosenv.2005.02.002
Lew. D: Sheppard. L: Checkowav. H: Kaufmaa J: Lumlev. T: Koenig. J: Siscovick. D. (2001). A case-crossover
analysis of particulate matter air pollution and out-of-hospital primary cardiac arrest. Epidemiology 12:
193-199.
Li. C: Zhang. O: Krotkov. NA: Streets. DG: He. K: Tsav. SC: Gleason. JF. (2010). Recent large reduction in sulfur
dioxide emissions from Chinese power plants observed by the Ozone Monitoring Instrument. Geophys Res
Lett 37: L08807. http://dx.doi.org/10.1029/2010GLQ42594
Li. R: Kou. X: Tiaa J: Meng. Z: Cai. Z: Cheng. F: Dong. C. (2014). Effect of sulfur dioxide on inflammatory and
immune regulation in asthmatic rats. Chemosphere 112: 296-304.
http://dx.doi.Org/10.1016/i.chemosphere.2014.04.065
Li. R: Meng. Z: Xie. J. (2007). Effects of sulfur dioxide on the expressions of MUC5AC and ICAM-1 in airway of
asthmatic rats. Regul Toxicol Pharmacol 48: 284-291. http://dx.doi.Org/10.1016/i.vrtph.2007.04.009
Li. R: Meng. Z: Xie. J. (2008). Effects of sulfur dioxide on the expressions of EGF, EGFR, and COX-2 in airway of
asthmatic rats. Arch Environ Contam Toxicol 54: 748-757. http://dx.doi.org/10.1007/s00244-007-9Q54-9
R-31
-------
Li. S: Batterman. S: Wasilevich. E: Wahl. R: Wirth. J: Su. FC: Mukheriee. B. (2011). Association of daily asthma
emergency department visits and hospital admissions with ambient air pollutants among the pediatric
Medicaid population in Detroit: Time-series and time-stratified case-crossover analyses with threshold
effects. Environ Res 111: 1137-1147. http://dx.doi.Org/10.1016/i.envres.2011.06.002
Li. S: Guo. Y: Williams. G. (2016). Acute impact of hourly ambient air pollution on preterm birth. Environ Health
Perspect 124: 1623-1629. http://dx.doi.org/10.1289/EHP200
Li. X: Bazer. FW: Gao. H: Jobgea W: Johnsoa GA: Li. P: McKnight. JR: Satterfield. MC: Spencer. TE: Wu. G.
(2009). Amino acids and gaseous signaling [Review]. Amino Acids 37: 65-78.
http://dx.doi.org/10.1007/s00726-009-Q264-5
Liang. Z: Wu. L: Fan. L: Zhao. O. (2014). Ambient air pollution and birth defects in Haikou city, Hainan province.
BMC Pediatr 14: 283. http://dx.doi.org/10.1186/sl2887-014-0283-6
Lillis. D: Cruz. CN: Collett. J. Jr: Richards. LW: Pandis. SN. (1999). Production and removal of aerosol in a
polluted fog layer: Model evaluation and fog effect onPM. Atmos Environ 33: 4797-4816.
http://dx.doi.Org/10.1016/S 1352-2310(99)00264-2
Lin. HK: Tsai. JJ: Wea MC: Tsai. MC: Chea CJ: Fu. LS. (2011a). Sodium sulfite aggravated allergic sensitization
and airway inflammation in mite allergen sensitized B ALB/c mice. Hum Exp Toxicol 30: 1682-1689.
http://dx.doi.org/10.1177/0960327111398673
Lin. LC: Takahashi. K. (2016). Will (CH3)(2)COO survive in humid conditions? J Chin Chem Soc 63: 472-479.
http://dx.doi.org/10.1002/iccs.20150Q518
Lin. S: Hwang. SA: Pantea. C: Kielb. C: Fitzgerald. E. (2004). Childhood asthma hospitalizations and ambient air
sulfur dioxide concentrations in Bronx County, New York. Arch Environ Occup Health 59: 266-275.
http://dx.doi.Org/10.3200/AEQH.59.5.266-275
Lin. W: Huang. W: Zhu. T: Hu. M: Brunekreef. B: Zhang. Y: Liu. X: Cheng. H: Gehring. U: Li. C: Tang. X.
(201 lb). Acute respiratory inflammation in children and black carbon in ambient air before and during the
2008 Beijing Olympics. Environ Health Perspect 119: 1507-1512. http://dx.doi.org/10.1289/ehp. 1103461
Lin. W: Zhu. T: Xue. T: Peng. W: Brunekreef. B: Gehring. U: Huang. W: Hu. M: Zhang. Y: Tang. X. (2015).
Association between changes in exposure to air pollution and biomarkers of oxidative stress in children
before and during the Beijing Olympics. Am J Epidemiol 181: 575-583.
http://dx.doi.org/10.1093/aie/kwu327
Lin. YT: Lee. YL: Jung. CR: Jaakkola. JJ: Hwang. BF. (2014). Air pollution and limb defects: A matched-pairs
case-control study in Taiwan. Environ Res 132: 273-280. http://dx.doi.Org/10.1016/i.envres.2014.04.028
Linares. B: Guizar. JM: Amador. N: Garcia. A: Miranda. V: Perez. JR: Chapela. R. (2010). Impact of air pollution
on pulmonary function and respiratory symptoms in children. Longitudinal repeated-measures study. BMC
PulmMed 10: 62. http://dx.doi.org/10.1186/1471-2466-10-62
Link. MS: Luttmann-Gibsoa H: Schwartz. J: Mittlemaa MA: Wessler. B: Gold. PR: Dockerv. DW: Laden. F.
(2013). Acute exposure to air pollution triggers atrial fibrillation. J Am Coll Cardiol 62: 816-825.
http://dx.doi.Org/10.1016/i.iacc.2013.05.043
Lina WS: Avol. EL: Peng. RC: Shamoo. DA: Hackney. JD. (1987). Replicated dose-response study of sulfur
dioxide effects in normal, atopic, and asthmatic volunteers. Am Rev Respir Dis 136:1127-1134.
http://dx.doi.org/10.1164/airccm/136.5.1127
Lina WS: Avol. EL: Shamoo. DA: Peng. RC: Spier. CE: Smith. MN: Hackney. JD. (1988). Effect of
metaproterenol sulfate on mild asthmatics' response to sulfur dioxide exposure and exercise. Arch Environ
Occup Health 43: 399-406. http://dx.doi.org/10.1080/00039896.1988.9935858
Lina WS: Avol. EL: Shamoo. DA: Venet. TG: Anderson. KR: Whvnot. JD: Hackney. JD. (1984a). Asthmatics'
responses to 6-hr sulfur dioxide exposures on two successive days. Arch Environ Health 39:313-319.
http://dx.doi.org/10.1080/00Q39896.1984.10545856
R-32
-------
Linn. WS: Fischer. DA: Shamoo. DA: Spier. CE: Valencia. LM: Anzar. UT: Hackney. JD. (1985a). Controlled
exposures of volunteers with chronic obstructive pulmonary disease to sulfur dioxide. Environ Res 37: 445-
451. http://dx.doi.org/10.1016/0013-9351(85190126-4
Linn. WS: Shamoo. DA: Andersoa KR: Whvnot. JD: Avol. EL: Hackney. JD. (1985b). Effects of heat and humidity
on the responses of exercising asthmatics to sulfur dioxide exposure. Am Rev Respir Dis 131: 221-225.
Linn. WS: Shamoo. DA: Peng. RC: Clark. KW: Avol. EL: Hackney. JD. (1990). Responses to sulfur dioxide and
exercise by medication-dependent asthmatics: Effect of varying medication levels. Arch Environ Occup
Health45: 24-30. http://dx.doi.org/10.1080/00039896.1990.9935920
Linn. WS: Shamoo. DA: Spier. CE: Valencia. LM: Anzar. UT: Venet. TG: Hackney. JD. (1983a). Respiratory
effects of 0.75 ppm sulfur dioxide in exercising asthmatics: Influence of upper-respiratory defenses.
Environ Res 30: 340-348. http://dx.doi.org/10.1016/0013-9351(83)90219-0
Linn. WS: Shamoo. DA: Venet. TG: Bailey. RM: Wightman. LH: Hackney. JD. (1984b). Comparative effects of
sulfur dioxide exposures at 5 degrees C and 22 degrees C in exercising asthmatics. Am Rev Respir Dis 129:
234-239.
Linn. WS: Shamoo. DA: Vinet. TG: Spier. CE: Valencia. LM: Anzar. UT: Hackney. JD. (1984c). Combined effect
of sulfur dioxide and cold in exercising asthmatics. Arch Environ Occup Health 39: 339-346.
http://dx.doi.org/10.1080/00039896.1984.1054586Q
Linn. WS: Venet. TG: Shamoo. DA: Valencia. LM: Anzar. UT: Spier. CE: Hackney. JD. (1983b). Respiratory
effects of sulfur dioxide in heavily exercising asthmatics: A dose-response study. Am Rev Respir Dis 127:
278-283.
Lipfert. FW: Batv. JD: Miller. JP: Wvzga. RE. (2006a). PM2.5 constituents and related air quality variables as
predictors of survival in a cohort of U.S. military veterans. Inhal Toxicol 18: 645-657.
http://dx.doi.org/10.1080/0895837060Q742946
Lipfert. FW: Perry. HM. Jr: Miller. JP: Batv. JD: Wvzga. RE: Carmodv. SE. (2000). The Washington University -
EPRI veterans' cohort mortality study: Preliminary results. Inhal Toxicol 12: 41-73.
http://dx.doi.org/10.1080/713856640
Lipfert. FW: Wvzga. RE. (1996). The effects of exposure error on environmental epidemiology. In RF Phalen; RC
Mannix; MC Tonini (Eds.), The second colloquium on particulate air pollution & human mortality &
morbidity (pp. 4-295-294-302). Sacramento, CA: California Air Resources Board.
Lipfert. FW: Wvzga. RE: Batv. JD: Miller. JP. (2006b). Traffic density as a surrogate measure of environmental
exposures in studies of air pollution health effects: Long-term mortality in a cohort of US veterans. Atmos
Environ 40: 154-169. http://dx.doi.Org/10.1016/i.atmosenv.2005.09.027
Lipfert. FW: Wvzga. RE: Batv. JD: Miller. JP. (2009). Air pollution and survival within the Washington University -
EPRI veterans cohort: Risks based on modeled estimates of ambient levels of hazardous and criteria air
pollutants. J Air Waste Manag Assoc 59: 473-489. http://dx.doi.Org/10.3155/1047-3289.59.4.473
Lipsett. MJ: Ostro. BP: Reynolds. P: Goldberg. D: Hertz. A: Jerrett. M: Smith. DF: Garcia. C: Chang. ET:
Bernstein. L. (2011). Long-term exposure to air pollution and cardiorespiratory disease in the California
teachers study cohort. Am J Respir Crit Care Med 184: 828-835. http://dx.doi.org/10.1164/rccm.201Q12-
2082QC
Liu. CC: Tsai. SS: Chiu. HF: Wu. TN: Chen. CC: Yang. CY. (2009a). Ambient exposure to criteria air pollutants
and risk of death from bladder cancer in Taiwan. Inhal Toxicol 21: 48-54.
http://dx.doi.org/10.1080/089583708022Q7326
Liu. D: Jia H: Tang. C: Du. J. (2010). Sulfur dioxide: A novel gaseous signal in the regulation of cardiovascular
functions [Review], Mini Rev Med Chem 10: 1039-1045. http://dx.doi.org/10.2174/1389557511009011Q39
Liu. F: Zhao. Y: Liu. YO: Liu. Y: Sun. J: Huang. MM: Liu. Y: Dong. GH. (2014a). Asthma and asthma related
symptoms in 23,326 Chinese children in relation to indoor and outdoor environmental factors: The Seven
Northeastern Cities (SNEC) Study. Sci Total Environ 497-498: 10-17.
http://dx.doi.Org/10.1016/i.scitotenv.2014.07.096
R-33
-------
Liu. L. (2013). Email from Dr. Liu to Dr. Patel; Response to data request. Available online at
http://www.regulations.gov/#!documentDetail:D=EPA-HO-ORD-2013-0232-0013
Liu. L: Pooa R: Chen. L: Frescura. AM: Montuschi. P: Ciabattoni. G: Wheeler. A: Dales. R. (2009b). Acute effects
of air pollution on pulmonary function, airway inflammation, and oxidative stress in asthmatic children.
Environ Health Perspect 117: 668-674. http://dx.doi.org/10.1289/ehpll813
Liu. S: Krewski. D: Shi. Y: Chen. Y: Burnett. R. (2007). Association between maternal exposure to ambient air
pollutants during pregnancy and fetal growth restriction. J Expo Sci Environ Epidemiol 17: 426-432.
http://dx.doi.org/10.1038/si.ies.75005Q3
Liu. S: Krewski. D: Shi. Y: Chen. Y: Burnett. RT. (2003). Association between gaseous ambient air pollutants and
adverse pregnancy outcomes in Vancouver, Canada. Environ Health Perspect 111: 1773-1778.
Liu. W: Huang. C: Hu. Y: Fu. O: Zou. Z: Sun. C: Shea L: Wang. X: Cai. J: Pan. J: Huang. Y: Chang. J: Sua Y:
Sundell. J. (2016). Associations of gestational and early life exposures to ambient air pollution with
childhood respiratory diseases in Shanghai, China: A retrospective cohort study. Environ Int 92-93: 284-
293. http://dx.doi.Org/10.1016/i.envint.2016.04.019
Liu. X: Easter. RC: Ghaa SJ: Zaveri. R: Rasch. P: Shi. X: Lamarque. JF: Gettelmaa A: Morrison. H: Vitt. F:
Conlev. A: Park. S: Neale. R: Hannav. C: Ekman. AML: Hess. P: Mahowald. N: Collins. W: Iacono. MJ:
Bretherton. CS: Flanner. MG: Mitchell. D. (2012a). Toward a minimal representation of aerosols in climate
models: Description and evaluation in the Community Atmosphere Model CAM5. Geosci Model Dev 5:
709-739. http://dx.doi.org/10 5194/gmd-5-709-2012
Liu. X: Lessner. L: Carpenter. DO. (2012b). Association between residential proximity to fuel-fired power plants
and hospitalization rate for respiratory diseases. Environ Health Perspect 120: 807-810.
http://dx.doi.org/10.1289/ehp. 1104146
Liu. XH: Penner. JE: Herzog. M. (2005). Global modeling of aerosol dynamics: Model description, evaluation, and
interactions between sulfate and nonsulfate aerosols. J Geophys Res 110: D18206.
http://dx.doi.org/10.1029/2004JD0Q5674
Liu. Y: Baves. KD: Sander. SP. (2014b). Measuring rate constants for reactions of the simplest Criegee intermediate
(CH200) by monitoring the OH radical. J Phys Chem A 118: 741-747.
http://dx.doi.org/10.1021/ip407058b
Llorca. J: Salas. A: Prieto-Salceda. D: Chinchon-Bengoechea. V: Delgado-Rodriguez. M. (2005). Nitrogen dioxide
increases cardiorespiratory admissions in Torrelavega (Spain). J Environ Health 68: 30-35.
Loerting. T: Liedl. KR. (2000). Toward elimination of discrepancies between theory and experiment: The rate
constant of the atmospheric conversion of S03 to H2S04. Proc Natl Acad Sci USA 97: 8874-8878.
http://dx.doi.org/10.1073/pnas.97.16.8874
Long. MS: Keene. WC: Easter. RC: Sander. R: Liu. X: Kerkweg. A: Ericksoa D. (2013). Sensitivity of tropospheric
chemical composition to halogen-radical chemistry using a fully coupled size-resolved multiphase
chemistry/global climate system Part 1: Halogen distributions, aerosol composition, and sensitivity of
climate-relevant gases. Atmos ChemPhys Discuss 13: 6067-6129. http://dx.doi.org/l0.5194/acpd-13-6067-
2013
Longo. BM. (2009). The Kilauea Volcano adult health study. Nurs Res 58: 23-31.
http://dx.doi.org/10.1097/NNR.0b013e318190Qcc5
Longo. BM: Rossignol. A: Greea JB. (2008). Cardiorespiratory health effects associated with sulphurous volcanic
air pollution. Public Health 122: 809-820. http://dx.doi.Org/10.1016/i.puhe.2007.09.017
Longo. BM: Yang. W. (2008). Acute bronchitis and volcanic air pollution: A community-based cohort study at
Kilauea Volcano, Hawai'i,USA. J Toxicol Environ Health A 71: 1565-1571.
http://dx.doi.org/10.1080/152873908Q2414117
Longo. BM: Yang. W: Green. JB: Crosby. FL: Crosby. VL. (2010). Acute health effects associated with exposure to
volcanic air pollution (vog) from increased activity at Kilauea Volcano in 2008. J Toxicol Environ Health
A 73: 1370-1381. http://dx.doi.org/10.1080/15287394.2010.49744Q
R-34
-------
Lopez-Aparicio. S: Smolik. J: Maskova. L: Souckova. M: Grontoft. T: Ondrackova. L: Stankiewicz. J. (2011).
Relationship of indoor and outdoor air pollutants in a naturally ventilated historical building envelope.
Build Environ 46: 1460-1468. http://dx.doi.Org/10.1016/i.buildenv.2011.01.013
Low. RB: Bielorv. L: Oureshi. AI: Dunn. V: Stuhlmiller. DF: Dickey. DA. (2006). The relation of stroke admissions
to recent weather, airborne allergens, air pollution, seasons, upper respiratory infections, and asthma
incidence, September 11, 2001, and day of the week. Stroke 37: 951-957.
http://dx.doi.org/10.1161/01.STR.0000214681.9468Q.66
Lu. C: Deng. O: Yu. CWF: Sundell. J: Ou. C. (2014). Effects of ambient air pollution on the prevalence of
pneumonia in children: Implication for National Ambient Air Quality Standards in China. Indoor Built
Environ 23: 259-269. http://dx.doi.org/10.1177/1420326X13504423
Lu. R: Turco. RP: Jacobson. MZ. (1997a). An integrated air pollution modeling system for urban and regional
scales: 1. Structure and performance. J Geophys Res 102: 6063-6079. http://dx.doi.org/10.1029/96JD03501
Lu. R: Turco. RP: Jacobson. MZ. (1997b). An integrated air pollution modeling system for urban and regional
scales: 2. Simulations for SCAQS 1987. J Geophys Res Atmos 102: 6081-6098.
http://dx.doi.org/10.1029/96JD03502
Luke. WT. (1997). Evaluation of a commercial pulsed fluorescence detector for the measurement of low-level S02
concentrations during the gas-phase sulfur intercomparison experiment. J Geophys Res 102: 16255-16265.
http://dx.doi.org/10.1029/96JD03347
Luo. L: Chea S: Jin. H: Tang. C: Du. J. (2011). Endogenous generation of sulfur dioxide in rat tissues. Biochem
Biophys Res Commun 415: 61-67. http://dx.doi.Org/10.1016/i.bbrc.2011.10.012
Maclntvre. EA: Karr. CJ: Koehoorn. M: Demers. PA: Tamburic. L: Lencar. C: Brauer. M. (2011). Residential air
pollution and otitis media during the first two years of life. Epidemiology 22: 81-89.
http://dx.doi.org/10.1097/EDE.0b013e3181fdb6Qf
Madigan. MT: Martinko. JM: Brock. TP. (2006). Brock biology of microorganisms. In MT Madigan; JM Martinko;
TD Brock (Eds.), Brock biology of microorganisms (11th ed.). Upper Saddle River, NJ: Pearson Prentice
Hall.
Maestrelli. P: Canova. C: Scapellato. ML: Visentia A: Tessari. R: Bartolucci. GB: Simonato. L: Lotti. M. (2011).
Personal exposure to particulate matter is associated with worse health perception in adult asthma. J
Investig Allergol Clin Immunol 21: 120-128.
Maggos. T: Saraga. D: Bairachtari. K: Tzagkaroulaki. I: Pateraki. S: Vasilakos. C: Makarounis. C: Stavdaris. A:
Danias. G: Anagnostopoulos. G: Frountas. V. (2016). Air quality assessment in passenger trains: the impact
of smokestack emissions. Air Qual Atmos Health 9: 391-401. http://dx.doi.org/10.1007/sll869-015-Q348-l
Magnussen. H: Jorres. R: Wagner. HM: von Nieding. G. (1990). Relationship between the airway response to
inhaled sulfur dioxide, isocapnic hyperventilation, and histamine in asthmatic subjects. Int Arch Occup
Environ Health 62: 485-491. http://dx.doi.org/10.1007/BF00381178
Mai. CT: Kirbv. RS: Correa. A: Rosenberg. D: Petros. M: Fagea MC. (2016). Public health practice of population-
based birth defects surveillance programs in the United States. J Public Health Manag Pract 22: E1-E8.
http://dx.doi.org/10.1097/PHH.000000000000Q221
Maier. KL: Wippermana U: Leuschel. L: Jostea M: Pflugmacher. S: Schroder. P: Sandermann. H: Takenaka. S:
Ziesenis. A: Hevder. J. (1999). Xenobiotic-metabolizing enzymes in the canine respiratory tract. Inhal
Toxicol 11: 19-35. http://dx.doi.org/10.1080/089583799197249
Makamure. MT: Reddv. P: Chuturgooa A: Naidoo. RN: Mentz. G: Batterman. S: Robins. TG. (2016a). Interaction
between ambient pollutant exposure, CD14 (-159) polymorphism and respiratory outcomes among children
in Kwazulu-Natal, Durban. Hum Exp Toxicol 36: 238-246. http://dx.doi.org/10.1177/0960327116646620
Makamure. MT: Reddv. P: Chuturgooa A: Naidoo. RN: Mentz. G: Batterman. S: Robins. TG. (2016b). Tumour
necrosis factor a polymorphism (TNF-308a G/A) in association with asthma related phenotypes and air
pollutants among children in KwaZulu-Natal. Asian Pac J Allergy Immunol 34: 217-222.
http://dx.doi.org/10.12932/AP0677
R-35
-------
Mamatsashvili. MI. (1970a). Detrimental effects of carbon monoxide and sulfur dioxide on fertility of female rats.
Hyg Sanit 35: 277-279.
Mamatsashvili. MI. (1970b). [Toxic effect of carbon monoxide, sulfur dioxide and their combinations on the fertility
of female rats]. Gig Sanit 35: 100-101.
Mannshardt. E: Sucic. K: Jiao. W: Dominici. F: Frev. HC: Reich. B: Fuentes. M. (2013). Comparing exposure
metrics for the effects of fine particulate matter on emergency hospital admissions. J Expo Sci Environ
Epidemiol 23: 627-636. http://dx.doi.org/10.1038/ies.2013.39
Marshall. JD: Netherv. E: Brauer. M. (2008). Within-urban variability in ambient air pollution: Comparison of
estimation methods. Atmos Environ 42: 1359-1369. http://dx.doi.Org/10.1016/i.atmosenv.2007.08.012
Martins. LC: Pereira. LAA: Lia CA: Santos. UP: Prioli. G: do Carmo Luiz. O: Saldiva. PHN: Braga. ALF. (2006).
The effects of air pollution on cardiovascular diseases: Lag structures. Rev Saude Publica 40: 677-683.
http://dx.doi.org/10.1590/S0034-89102006000500Q18
Mathews. TJ: MacDorman. MF. (2010). Infant mortality statistics from the 2006 period linked birth/infant death
data set. Atlanta, GA: Centers for Disease Control.
http://www.cdc. gov/nchs/data/nvsr/nvsr58/nvsr58 17.pdf
Matsumi. Y: Shigemori. H: Takahashi. K. (2005). Laser-induced fluorescence instrument for measuring atmospheric
S02. Atmos Environ 39: 3177-3185. http://dx.doi.Org/10.1016/i.atmosenv.2005.02.023
Mauldia RL. Ill: Berndt. T: Sipila. M: Paasonea P: Petaia. T: Kim S: Kurten. T: Stratmana F: Kerminea VM:
Kulmala. M. (2012). A new atmospherically relevant oxidant of sulphur dioxide. Nature 488: 193-196.
http://dx.doi.org/10.1038/naturell278
McCormick. BT: Herzog. M: Yang. J: Edmonds. M: Mather. TA: Cam. SA: Hidalgo. S: Langmana B. (2014). A
comparison of satellite- and ground-based measurements of SO 2 emissions from Tungurahua volcano,
Ecuador. J Geophys Res Atmos 119: 42644285. http://dx.doi.org/10.1002/2013JDQ19771
McCurdv. T: Glea G: Smith. L: Lakkadi. Y. (2000). The National Exposure Research Laboratory's consolidated
human activity database. J Expo Anal Environ Epidemiol 10: 566-578.
http://dx.doi.org/10.1038/si.iea.7500114
McLindea CA: Fioletov. V: Boersma. KF: Krotkov. N: Sioris. CE: Veefkind. JP: Yang. K. (2012). Air quality over
the Canadian oil sands: A first assessment using satellite observations. Geophys Res Lett 39: L04804.
http://dx.doi.org/10.1029/2011GL05Q273
Mechtouff. L: Canoui-Poitrine. F: Schott. AM: Nighoghossiaa N: Trouillas. P: Termoz. A: Porthault-Chatard. S:
David. JS: Chasles. V: Derex. L. (2012). Lack of association between air pollutant exposure and short-term
risk of ischaemic stroke in Lyon, France. Int J Stroke 7: 669-674. http://dx.doi.org/10. Ill 1/i. 1747-
4949.2011.00737.x
Medina. PS: Liu. YD: Wang. LM: Zhang. JS. (2011). Detection of sulfur dioxide by cavity ring-down spectroscopy.
Environ Sci Technol 45: 1926-1931. http://dx.doi.org/10.1021/esl03739r
Mehta. S: Ngo. LH: Dzung. DV: Cohen. A: Thach. TO: Dan. VX: Tuaa ND: Giang. LT. (2013). Air pollution and
admissions for acute lower respiratory infections in young children of Ho Chi Minh City. Air Qual Atmos
Health 6: 167-179. http://dx.doi.org/10.1007/sll869-011-0158-z
Melville. GN. (1970). Changes in specific airway conductance in healthy volunteers following nasal and oral
inhalation of S02. West Indian Med J 19: 231-235.
Mendola. P: Wallace. M: Hwang. BS: Liu. D: Robledo. C: Mannisto. T: Sundaram R: Shermaa S: Ying. O: Grantz.
KL. (2016a). Preterm birth and air pollution: Critical windows of exposure for women with asthma. J
Allergy Clin Immunol 138: 432-440.e435. http://dx.doi.Org/10.1016/i.iaci.2015.12.1309
Mendola. P: Wallace. M: Liu. D: Robledo. C: Mannisto. T: Grantz. KL. (2016b). Air pollution exposure and
preeclampsia among US women with and without asthma. Environ Res 148: 248-255.
http://dx.doi.Org/10.1016/i.envres.2016.04.004
R-36
-------
Meng. X: Wang. C: Cao. D: Wong. CM: Kan. H. (2013). Short-term effect of ambient air pollution on COPD
mortality in four Chinese cities. Atmos Environ 77: 149-154.
http://dx.doi.Org/10.1016/i.atmosenv.2013.05.001
Meng. Z: Li. R: Zhang. X. (2005a). Levels of sulfite in three organs from mice exposed to sulfur [corrected] dioxide.
Inhal Toxicol 17: 309-313. http://dx.doi.org/10.1080/0895837059Q922634
Meng. Z: Li. R: Zhang. X. (2005b). Levels of sulfite in three organs from mice exposed to sulfur dioxide [Erratum],
Inhal Toxicol 17: 495. http://dx.doi.org/10.1080/089583705910Q1671
Meng. Z: Zhang. B: Ruan. A: Sang. N: Zhang. J. (2002). Micronuclei induced by sulfur dioxide inhalation in mouse
bone-marrow cells in vivo. Inhal Toxicol 14: 303-309. http://dx.doi.org/10.1080/08958370252809Q77
Metzger. KB: Kleia M: Flanders. WD: Peel. JL: Mulholland. JA: Langberg. JJ: Tolbert. PE. (2007). Ambient air
pollution and cardiac arrhythmias inpatients with implantable defibrillators. Epidemiology 18: 585-592.
http://dx.doi.org/10.1097/EDE.0b013e318124If0e
Metzger. KB: Tolbert. PE: Klein. M: Peel. JL: Flanders. WD: Todd. KH: Mulholland. JA: Ryan. PB: Frumkia H.
(2004). Ambient air pollution and cardiovascular emergency department visits. Epidemiology 15: 46-56.
http://dx.doi.org/10.1097/01.EDE.00001Q1748.28283.97
Michaud. JP: Grove. JS: Krupitskv. D. (2004). Emergency department visits and "vog"-related air quality in Hilo,
Hawai'i. Environ Res 95: 11-19. http://dx.doi.org/10.1016/S0013-9351(03)00122-l
Michikawa. T: Morokuma. S: Fukushima. K: Ueda. K: Takeuchi. A: Kato. K: Nitta. H. (2015). A register-based
study of the association between air pollutants and hypertensive disorders in pregnancy among the Japanese
population. Environ Res 142: 644-650. http://dx.doi.Org/10.1016/i.envres.2015.08.024
Michikawa. T: Morokuma. S: Yamazaki. S: Fukushima. K: Kato. K: Nitta. H. (2016). Exposure to air pollutants
during the early weeks of pregnancy, and placenta praevia and placenta accreta in the western part of Japan.
Environ Int 92-93: 464-470. http://dx.doi.Org/10.1016/i.envint.2016.04.037
Miller. KA: Siscovick. PS: Sheppard. L: Shepherd. K: Sullivan. JH: Anderson. GL: Kaufmaa JD. (2007). Long-
term exposure to air pollution and incidence of cardiovascular events in women. N Engl J Med 356: 447-
458. http://dx.doi.org/10.1056/NEJMoa054409
Miloievic. A: Wilkinson. P: Armstrong. B: Bhaskaran. K: Smeeth. L: Haiat. S. (2014). Short-term effects of air
pollution on a range of cardiovascular events in England and Wales: Case-crossover analysis of the MINAP
database, hospital admissions and mortality. Heart 100: 1093-1098. http://dx.doi.org/10.1136/heartinl-
2013-304963
Min. JY: Mia KB: Cho. SI: Paek. D. (2008a). Combined effects of cigarette smoking and sulfur dioxide on lung
function in Koreans. JToxicol Environ Health A71: 301-303.
http://dx.doi.org/10.1080/152873907Q1738475
Min. JY: Mia KB: Cho. SI: Paek. D. (2009). Combined effect of cigarette smoking and sulfur dioxide on heart rate
variability [Letter], Int J Cardiol 133: 119-121. http://dx.doi.Org/10.1016/i.iicard.2007.08.139
Min. KB: Min. JY: Cho. SI: Paek. D. (2008b). The relationship between air pollutants and heart-rate variability
among community residents in Korea. Inhal Toxicol 20: 435-444.
http://dx.doi.org/10.1080/089583708019Q3834
Mochizuki. H: Shigeta. M: Kato. M: Maeda. S: Shimizu. T: Mirokawa. A. (1995). Age-related changes in bronchial
hyperreactivity to methacholine in asthmatic children. Am J Respir Crit Care Med 152: 906-910.
http://dx.doi.Org/10.1164/airccm.152.3.7663803
Moolgavkar. SH. (2000). Air pollution and hospital admissions for diseases of the circulatory system in three U.S.
metropolitan areas. J Air Waste Manag Assoc 50: 1199-1206.
http://dx.doi.org/10.1080/10473289.200Q.10464162
Moolgavkar. SH. (2003). Air pollution and daily deaths and hospital admissions in Los Angeles and Cook counties.
In Revised Analyses of the National Morbidity, Mortality, and Air Pollution Study, Part II (pp. 183-198).
Boston, MA: Health Effects Institute.
R-37
-------
Moolgavkar. SH: McClellaa RQ: Dewanii. A: Turim. J: Luebeck. EG: Edwards. M. (2013). Time-series analyses of
air pollution and mortality in the United States: A subsampling approach. Environ Health Perspect 121:73-
78. http://dx.doi.org/10.1289/ehp. 1104507
Moon. JS: Kim. YS: Kim. JH: Soa BS: Kim. PS: Yang. W. (2009). Respiratory health effects among
schoolchildren and their relationship to air pollutants in Korea. Int J Environ Health Res 19: 31-48.
http://dx.doi.org/10.1080/096031208Q2272201
Morello-Frosch. R: Jesdale. BM: Sadd. JL: Pastor. M. (2010). Ambient air pollution exposure and full-term birth
weight in California. Environ Health 9: 44. http://dx.doi.org/10.1186/1476-069X-9-44
Moridi. M: Ziaei. S: Kazemneiad. A. (2014). Exposure to ambient air pollutants and spontaneous abortion. J Obstet
Gynaecol Res 40: 743-748. http://dx.doi.org/10.1111/iog. 12231
Morikawa. A: Mochizuki. H: Shigeta. M: Tokuvama. K: Kuroume. T. (1994). Age-related changes in bronchial
hyperreactivity during the adolescent period. J Asthma 31: 445-451.
http://dx.doi.org/10.3109/02770909409Q89486
Morris. RD: Naumova. EN: Munasinghe. RL. (1995). Ambient air pollution and hospitalization for congestive heart
failure among elderly people in seven large US cities. Am J Public Health 85: 1361-1365.
http://dx.doi.org/10.2105/aiph.85.10.1361
Mortimer. KM: Neas. LM: Dockerv. DW: Redline. S: Tager. IB. (2002). The effect of air pollution on inner-city
children with asthma. Eur Respir J 19: 699-705. http://dx.doi.org/10.1183/09031936.02.002471Q2
Mueller. SF: Mao. O: Mallard. JW. (2011). Modeling natural emissions in the Community Multiscale Air Quality
(CMAQ) model Part 2: Modifications for simulating natural emissions. Atmos ChemPhys 11: 293-320.
http://dx.doi.org/10.5194/acp-l 1-293-2011
Muhaiarine. N: Mustard. C: Roos. LL: Young. TK: Gelskev. DE. (1997). Comparison of survey and physician
claims data for detecting hypertension. J Clin Epidemiol 50: 711-718. http://dx.doi.org/10.1016/SQ895-
4356(97)00019-X
Murgia. N: Brismaa J: Claesson. A: Muzi. G: Olia AC: Toren. K. (2014). Validity of a questionnaire-based
diagnosis of chronic obstructive pulmonary disease in a general population-based study. BMC Pulm Med
14: 49. http://dx.doi.org/10.1186/1471-2466-14-49
Myers. DJ: Bigbv. BG: Boushev. HA. (1986a). The inhibition of sulfur dioxide-induced bronchoconstriction in
asthmatic subjects by cromolyn is dose dependent. Am Rev Respir Dis 133: 1150-1153.
Myers. DJ: Bigbv. BG: Calvavrac. P: Sheppard. D: Boushev. HA. (1986b). Interaction of cromolyn and a
muscarinic antagonist in inhibiting bronchial reactivity to sulfur dioxide and to eucapnic hyperpnea alone.
Am Rev Respir Dis 133: 1154-1158.
Nadel. JA: Salem. H: Tamplin. B: Tokiwa. Y. (1965a). Mechanism of bronchoconstriction during inhalation of
sulfur dioxide. J Appl Physiol (1985) 20: 164-167.
Nadel. JA: Salem. H: Tamplin. B: Tokiwa. Y. (1965b). Mechanism of bronchoconstriction: During inhalation of
sulfur dioxide; reflex involving vagus nerves. Arch Environ Health 10: 175-178.
http://dx.doi.org/10.1080/00Q39896.1965.10663979
Nadzieiko. C: Fang. K: Narciso. S: Zhong. M: Su. WC: Gordoa T: Nadas. A: Chea LC. (2004). Effect of
particulate and gaseous pollutants on spontaneous arrhythmias in aged rats. Inhal Toxicol 16: 373-380.
http://dx.doi.org/10.1080/0895837049Q439533
NAEPP (National Asthma Education and Prevention Program). (2007). Expert panel report 3 (EPR-3): Guidelines
for the diagnosis and management of asthmasummary report 2007. J Allergy Clin Immunol 120: S94-S138.
http://dx.doi.Org/10.1016/i.iaci.2007.09.029
Nafstad. P: Haheim. LL: Wisloff. T: Gram. F: Oftedal. B: Holme. I: Hiermana I: Lerea P. (2004). Urban air
pollution and mortality in a cohort of Norwegian men. Environ Health Perspect 112: 610-615.
Nahidi. F: Gholami. R: Rashidi. Y: Maid. HA. (2014). Relationship between air pollution and pre-eclampsia in
pregnant women: a case-control study. East Mediterr Health J 19: S60-S66.
R-38
-------
NASA (National Aeronautics and Space Administration). (2008a). Okmok Volcano in Alaskas Aleutian Islands.
Available online at http://earthobservatorv.nasa. gov/IOTD/view.php?id=8939
NASA (National Aeronautics and Space Administration). (2008b). Sulfur dioxide plume fromKilauea. Available
online at http://earthobservatorv.nasa.gov/IOTD/view.php?id=8614&eocn=image&eoci=related image
Neas. LM: Dockerv. DW: Koutrakis. P: Tollerud. DJ: Speizer. FE. (1995). The association of ambient air pollution
with twice daily peak expiratory flow rate measurements in children. Am J Epidemiol 141: 111-122.
Neophvtou. AM: White. MJ: Oh. SS: Thakur. N: Galanter. JM: Nishimura. KK: Pino-Yanes. M: Torgerson. DG:
Gignoux. CR: Eng. C: Nguvea EA: Hu. D: Mak. AC: Kumar. R: Seibold. MA: Davis. A: Farber. HJ:
Meade. K: Avila. PC: Serebriskv. D: Lenoir. MA: Brigino-Buenaventura. E: Rodriguez-Cintroq W:
Bibbins-Domingo. K: Thvne. SM: Williams. LK: Sea S: Gilliland. FD: Gauderman. WJ: Rodriguez-
Santana. JR: Lurmana F: Balmes. JR: Eisen. EA: Burchard. EG. (2016). Air pollution and lung function in
minority youth with asthma in the GALA II (Genesenvironments and admixture in Latino Americans) and
SAGE II (Study of African Americans, asthma, genes, and environments) studies. Am J Respir Crit Care
Med 193: 1271-1280. http://dx.doi.org/10.1164/rccm.201508-1706QC
Netherv. E: Brauer. M: Janssen. P. (2009). Time-activity patterns of pregnant women and changes during the course
of pregnancy. J Expo Sci Environ Epidemiol 19: 317-324. http://dx.doi.org/10.1038/ies.2008.24
Neukirch. F: Segala. C: Le Moullec. Y: Korobaeff. M: Aubier. M. (1998). Short-term effects of low-level winter
pollution on respiratory health of asthmatic adults. Arch Environ Occup Health 53: 320-328.
http://dx.doi.org/10.1080/000398998096Q5716
Neupane. B: Jerrett. M: Burnett. RT: Marrie. T: Araia A: Loeb. M. (2010). Long-term exposure to ambient air
pollution and risk of hospitalization with community-acquired pneumonia in older adults. Am J Respir Crit
Care Med 181: 47-53. http://dx.doi.org/10.1164/rccm.200901-0160QC
NHLBINAEPP (National Heart, Lung, and Blood Institute, National Asthma Education and Prevention Program).
(2007). Expert panel report 3: Guidelines for the diagnosis and management of asthma. (Report No: 07-
4051). Bethesda, MD: U.S. Department of Health and Human Services, National Institutes of Health.
http://www.nhlbi.nih.gov/files/docs/guidelines/asthgdln.pdf
Niinimaa. V: Cole. P: Mintz. S: Shephard. RJ. (1981). Oronasal distribution of respiratory airflow. Respir Physiol
43: 69-75. http://dx.doi.org/10.1016/0034-5687(8D90089-X
Nishimura. KK: Galanter. JM: Roth. LA: Oh. SS: Thakur. N: Nguvea EA: Thvne. S: Farber. HJ: Serebriskv. D:
Kumar. R: Brigino-Buenaventura. E: Davis. A: Lenoir. MA: Meade. K: Rodriguez-Cintron. W: Avila. PC:
Borrell. LN: Bibbins-Domingo. K: Rodriguez-Santana. J. R.: Sea S: Lurmann. F: Balmes. J. R.: Burchard.
EG. (2013). Early-life air pollution and asthma risk in minority children: The GALA II and SAGE II
studies. Am J Respir Crit Care Med 188: 309-318. http://dx.doi.org/10.1164/rccm.201302-0264QC
NOAA (National Oceanic and Atmospheric Administration). (2014). HYSPLIT - hybrid single particle lagrangian
integrated trajectory model. Available online at http://readv.arl.noaa.gov/HYSPLIT.php
Nordling. E: Berglind. N: Melea E: Emenius. G: Hallberg. J: Nvberg. F: Pershagen. G: Svartengrea M: Wickman.
M: Bellander. T. (2008). Traffic-related air pollution and childhood respiratory symptoms, function and
allergies. Epidemiology 19: 401-408. http://dx.doi.org/10.1097/EDE.0b013e31816alce3
Nowak. D: Jorres. R: Berger. J: Claussen. M: Magnussen. H. (1997). Airway responsiveness to sulfur dioxide in an
adult population sample. Am J Respir Crit Care Med 156: 1151-1156.
http://dx.doi.Org/10.1164/airccm.156.4.9607025
Nowlaa CR: Liu. X: Chance. K: Cai. Z: Kurosu. TP: Lee. C: Martia RV. (2011). Retrievals of sulfur dioxide from
the Global Ozone Monitoring Experiment 2 (GOME-2) using an optimal estimation approach: Algorithm
and initial validation. J Geophys Res 116: D18301. http://dx.doi.org/10.1029/2011JD015808
NRC (National Research Council). (2012). Exposure science in the 21st century: a vision and a strategy.
Washington, DC: The National Academies Press, http://www.nap.edu/catalog/13507/exposure-science-in-
the-21 st-centurv-a-vision-and-a
R-39
-------
O'Bvrne. PM: Gauvreau. GM: Brannan. JD. (2009). Provoked models of asthma: What have we learnt? [Review].
Clin Exp Allergy 39: 181-192. http://dx.doi.org/10.1111/i.1365-2222.2008.03172.x
O'Connor. GT: Neas. L: Vaughn. B: Kattan. M: Mitchell. H: Craia EF: Evans. R. Ill: Gruchalla. R: Morgaa W:
Stout. J: Adams. GK: Lippmann. M. (2008). Acute respiratory health effects of air pollution on children
with asthma in US inner cities. J Allergy Clin Immunol 121: 1133-1139.el 131.
http://dx.doi.Org/10.1016/i.iaci.2008.02.020
Office of Transportation and Air Quality. (2010). Designation of North American emission control area to reduce
emissions from ships [Fact Sheet]. (EPA-420-F-10-015). Washington, D.C.: U.S. Environmental Protection
Agency. https://nepis.epa.gOv/Exe/ZyPDF.cgi/P 100AU0I.PDF?Dockey=P 100AU0I.PDF
Qgawa & Co (Ogawa & Company). (2007). Ambient air passive sampler for NO-N02, NOx, S02, 03, NH3.
Available online at http://ogawausa.com/passive-sampler/
Ohvama. K: Ito. T: Kanisawa. M. (1999). The roles of diesel exhaust particle extracts and the promotive effects of
N02 and/or S02 exposure on rat lung tumorigenesis. Cancer Lett 139: 189-197.
http://dx.doi.org/10.1016/S0304-3835(99)00040-3
Oiamo. TH: Luginaah. IN. (2013). Extricating sex and gender in air pollution research: a community-based study on
cardinal symptoms of exposure. Int J Environ Res Public Health 10: 3801-3817.
http://dx.doi.org/10.3390/iierphlQ093801
Oiamo. TH: Luginaah. IN: Atari. DO: Gorev. KM. (2011). Air pollution and general practitioner access and
utilization: a population based study in Sarnia, 'Chemical Valley,' Ontario. Environ Health 10: 71.
http://dx.doi.org/10.1186/1476-069X-10-71
Olesen. HR: Lofstrom. P: Berkowicz. R: Jensen. AB. (1992). An improved dispersion model for regulatory use -
The OML model. In H van Dop; G Kallos (Eds.), Air pollution modelling and its application IX (pp. 29-
38). New York, New York: Plenum Press, http://dx.doi.org/10.1007/978-l-4615-3052-7 3
Ong. CB: Kumagai. K: Brooks. PT: Brandenberger. C: Lewandowski. RP: Jackson-Humbles. DN: Nault. R:
Zacharewski. TR: Wagner. JG: Harkema. JR. (2016). Ozone-induced type 2 immunity in nasal airways.
Development and lymphoid cell dependence in mice. Am J Respir Cell Mol Biol 54: 331-340.
http://dx.doi.org/10.1165/rcmb.2015-0165QC
Orazzo. F: Nespoli. L: Ito. K: Tassinari. D: Giardina. D: Funis. M: Cecchi. A: Trapani. C: Forgeschi. G: Vignini. M:
Nosetti. L: Pigna. S: Zanobetti. A. (2009). Air pollution, aeroallergens, and emergency room visits for
acute respiratory diseases and gastroenteric disorders among young children in six Italian cities. Environ
Health Perspect 117: 1780-1785. http://dx.doi.org/10.1289/ehp.0900599
Ott. W: Klepeis. N: Switzer. P. (2008). Air change rates of motor vehicles and in-vehicle pollutant concentrations
from secondhand smoke. J Expo Sci Environ Epidemiol 18: 312-325.
http://dx.doi.org/10.1038/si.ies.750Q601
Paciorek. CJ. (2010). The importance of scale for spatial-confounding bias and precision of spatial regression
estimators. Stat Sci 25: 107-125. http://dx.doi.org/10.1214/10-STS326
Paine. R: Samani. O: Kaplan. M: Knipping. E: Kumar. N. (2015). Evaluation of low wind modeling approaches for
two tall-stack databases. J Air Waste Manag Assoc 65: 1341-1353.
http://dx.doi.org/10.1080/10962247.2Q15.1085924
Pan. G: Zhang. S: Feng. Y: Takahashi. K: Kagawa. J: Yu. L: Wang. P: Liu. M: Liu. O: Hou. S: Pan. B: Li. J. (2010).
Air pollution and children's respiratory symptoms in six cities of Northern China. Respir Med 104: 1903-
1911. http://dx.doi.Org/10.1016/i.rmed.2010.07.018
Panasevich. S: Leander. K: Liungman. P: Bellander. T: de Faire. U: Pershagen. G: Nvberg. F. (2013). Interaction
between air pollution exposure and genes in relation to levels of inflammatory markers and risk of
myocardial infarction. BMJ Open 3: e003058. http://dx.doi.org/10.1136/bmiopen-2013-003058
R-40
-------
Panasevich. S: Leander. K: Rosenlund. M: Liungman. P: Bellander. T: de Faire. U: Pershagea G: Nvberg. F.
(2009). Associations of long- and short-term air pollution exposure with markers of inflammation and
coagulation in a population sample. Occup Environ Med 66: 747-753.
http://dx.doi.org/10.1136/oem.2008.Q43471
Pandis. SN: Seinfeld. JH. (1989). Mathematical modeling of acid deposition due to radiation fog. J Geophys Res 94:
1291112923. http://dx.doi.org/10.1029/JD094iD10pl2911
Panter. R: Penzhorn. RD. (1980). Alkyl sulfonic acids in the atmosphere. Atmos Environ 14: 149-152.
http://dx.doi.org/10.1016/0004-6981(80)90119-5
Park. JK: Kim YK: Lee. SR: Cho. SH: Min. KU: Kim. YY. (2001). Repeated exposure to low levels of sulfur
dioxide (S02) enhances the development of ovalbumin-induced asthmatic reactions in guinea pigs. Ann
Allergy Asthma Immunol 86: 62-67. http://dx.doi.Org/10.1016/S 1081-1206(10)62358-7
Park. JW: Lim YH: Kvung. SY: An. CH: Lee. SP: Jeong. SH: Ju. YS. (2005). Effects of ambient particulate matter
on peak expiratory flow rates and respiratory symptoms of asthmatics during Asian dust periods in Korea.
Respirology 10: 470-476. http://dx.doi.Org/10.llll/i.1440-1843.2005.00728.x
Parker. JD: Akinbami. LJ: Woodruff. TJ. (2009). Air pollution and childhood respiratory allergies in the United
States. Environ Health Perspect 117: 140-147. http://dx.doi.org/10.1289/ehp. 11497
Pascal. L: Pascal. M: Stempfelet. M: Goria. S: Declerca. C. (2013). Ecological study on hospitalizations for cancer,
cardiovascular, and respiratory diseases in the industrial area of Etang-de-Berre in the South of France. J
EnvironPublic Health2013: 328737. http://dx.doi.org/10.1155/2013/328737
Pasauill. F: Smith. FB. (1983). Atmospheric diffusion (3rd ed.). New York, NY: John Wiley & Sons.
Patterson. E: Eatough. DJ. (2000). Indoor/outdoor relationships for ambient PM2.5 and associated pollutants:
epidemiological implications in Lindon, Utah. J Air Waste Manag Assoc 50: 103-110.
http://dx.doi.org/10.1080/10473289.200Q.10463986
Pauluhn. J: Thvssea J: Althoff. J: Kimmerle. G: Mohr. U. (1985). Long-term inhalation study with benzo(a)pyrene
and S02 in Syrian golden hamsters [Abstract]. Exp Pathol 28: 31. http://dx.doi.org/10.1016/SQ232-
1513(85)80029-3
Peacock. JL: Andersoa HR: Bremner. SA: Marston. L: Seemungal. TA: Strachan. DP: Wedzicha. JA. (2011).
Outdoor air pollution and respiratory health in patients with COPD. Thorax 66: 591-596.
http://dx.doi.org/10.1136/thx.2010.155358
Peacock. PR: Spence. JB. (1967). Incidence of lung tumours in LX mice exposed to (1) free radicals; (2) S02. Br J
Cancer 21: 606-618. http://dx.doi.org/10.1038/bic.1967.71
Pearce. JL: Waller. LA: Mulholland. JA: Sarnat. SE: Strickland. MJ: Chang. HH: Tolbert. PE. (2015). Exploring
associations between multipollutant day types and asthma morbidity: Epidemiologic applications of self-
organizing map ambient air quality classifications. Environ Health 14:55.
http://dx.doi.org/10.1186/sl2940-015-0Q41-8
Peel. JL: Kleia M: Flanders. WD: Mulholland. JA: Freed. G: Tolbert. PE. (2011). Ambient air pollution and apnea
and bradycardia in high-risk infants on home monitors. Environ Health Perspect 119: 1321-1327.
http://dx.doi.org/10.1289/ehp. 1002739
Peel. JL: Metzger. KB: Klein. M: Flanders. WD: Mulholland. JA: Tolbert. PE. (2007). Ambient air pollution and
cardiovascular emergency department visits in potentially sensitive groups. Am J Epidemiol 165: 625-633.
http://dx.doi.org/10.1093/aie/kwk051
Peel. JL: Tolbert. PE: Klein. M: Metzger. KB: Flanders. WD: Todd. K: Mulholland. JA: Rvaa PB: Frumkin. H.
(2005). Ambient air pollution and respiratory emergency department visits. Epidemiology 16: 164-174.
http://dx.doi.org/10.1097/01.ede.00001529Q5.42113.db
Penard-Morand. C: Raherisoa C: Charpia D: Kopferschmitt. C: Lavaud. F: Caillaud. D: Annesi-Maesano. I.
(2010). Long-term exposure to proximity air pollution and asthma and allergies in urban children. Eur
Respir J 36: 33-40. http://dx.doi.org/10.1183/09031936.00116109
R-41
-------
Pereira. LAA: Loomis. D: Conceicao. GMS: Braga. ALF: Areas. RM: Kishi. HS: Singer. JM: Bohm. GM: Saldiva.
PHN. (1998). Association between air pollution and intrauterine mortality in Sao Paulo, Brazil. Environ
Health Perspect 106: 325-329. http://dx.doi.org/10.2307/3434038
Perry. SG: Cimorelli. AJ: Paine. RJ: Brode. RW: Weil. JC: Venkatram. A: Wilsoa RB: Lee. RF: Peters. WD.
(2005). AERMOD: A dispersion model for industrial source applications. Part II: Model performance
against 17 field study databases. J Appl Meteorol 44: 694-708. http://dx.doi.org/10.1175/JAM2228.1
Peters. A: Goldsteia IF: Bever. U: Franke. K: Heinrich. J: Dockerv. DW: Spengler. JD: Wichmana HE. (1996a).
Acute health effects of exposure to high levels of air pollution in eastern Europe. Am J Epidemiol 144:
570-581. http://dx.doi.org/10.1093/oxfordiournals.aie.a008967
Peters. J: Hedlev. AJ: Wong. CM: Lam TH: Ong. SG: Liu. J: Spiegelhalter. DJ. (1996b). Effects of an ambient air
pollution intervention and environmental tobacco smoke on children's respiratory health in Hong Kong. Int
J Epidemiol 25: 821-828. http://dx.doi.Org/10.1093/iie/25.4.821
Petroeschevskv. A: Simpson. RW: Thalib. L: Rutherford. S. (2001). Associations between outdoor air pollution and
hospital admissions in Brisbane, Australia. Arch Environ Occup Health 56: 37-52.
http://dx.doi.org/10.1080/00039890109604Q53
Phalea RF: Oldham. MJ: Beaucage. CB: Crocker. TT: Mortensea JD. (1985). Postnatal enlargement of human
tracheobronchial airways and implications for particle deposition. AnatRec 212: 368-380.
http://dx.doi.org/10.1002/ar. 1092120408
Pham. M: Muller. JF: Brasseur. GP: Granier. C: Megie. G. (1996). A 3D model study of the global sulphur cycle:
Contributions of anthropogenic and biogenic sources. Atmos Environ 30: 1815-1822.
http://dx.doi.org/10.1016/1352-2310(95)00390-8
Poloniecki. JD: Atkinson. RW: Ponce de Leon. A: Anderson. HR. (1997). Daily time series for cardiovascular
hospital admissions and previous day's air pollution in London, UK. Occup Environ Med 54: 535-540.
http://dx.doi.org/10.1136/oem.54.8.535
Pool. BL: Brendler. S: Klein. RG: Monarca. S: Pasauini. R: Schmezer. P: Zeller. WJ. (1988a). Effects of S02 or
NOx on toxic and genotoxic properties of chemical carcinogens. II. Short term in vivo studies.
Carcinogenesis 9: 1247-1252. http://dx.doi.org/10.1093/carcin/9.7.1247
Pool. BL: Janowskv. I: Kleia P: Kleia RG: Schmezer. P: Vogt-Leucht. G: Zeller. WJ. (1988b). Effects of S02 or
NOx on toxic and genotoxic properties of chemical carcinogens. I. In vitro studies. Carcinogenesis 9: 1237-
1245. http://dx.doi.Org/10.1093/carcin/9.7.1237
Pope. CA. Ill: Burnett. RT: Thun. MJ: Calle. EE: Krewski. D: Ito. K: Thurston. GD. (2002). Lung cancer,
cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. JAMA 287: 1132-
1141. http://dx.doi.org/10. lOOl/iama.287.9.1132
Pope. CA. Ill: Thua MJ: Namboodiri. MM: Dockerv. DW: Evans. JS: Speizer. FE: Heath. CW. 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. http://dx.doi.org/10.1164/airccm/15L3 Pt 1.669
Portnov. BA: Reiser. B: Karkabi. K: Cohen-Kastel. O: Dubnov. J. (2012). High prevalence of childhood asthma in
Northern Israel is linked to air pollution by particulate matter: evidence from GIS analysis and Bayesian
Model Averaging. Int J Environ Health Res 22: 249-269. http://dx.doi.org/10.1080/096Q3123.2011.634387
Pouliot. G: van Per Goa HAC. D: Kuenen. J: Zhang. J: Mo ran. MP: Makar. PA. (2015). Analysis of the emission
inventories and model-ready emission datasets of Europe and North America for phase 2 of the AQMEII
project. Atmos Environ 115: 345-360. http://dx.doi.Org/10.1016/i.atmosenv.2014.10.061
Poursafa. P: Baradaran-Mahdavi. S: Moradi. B: Haghioov Javanmard. S: Taiadini. M: Mehrabiaa F: Kelishadi. R.
(2016). The relationship of exposure to air pollutants in pregnancy with surrogate markers of endothelial
dysfunction in umbilical cord. Environ Res 146: 154-160. http://dx.doi.Org/10.1016/i.envres.2015.12.018
Power. J. (2013). Alaska volcanoes. Available online at
http://woodshole.er.usgs.gov/operations/obs/rmobs pub/html/alaska.html
R-42
-------
Prata. AJ: Gangale. G: Clarisse. L: Karagulian. F. (2010). Ash and sulfur dioxide in the 2008 eruptions of Okmok
and Kasatochi: Insights from high spectral resolution satellite measurements. J Geophys Res Atmos 115:
D00L18. http://dx.doi.org/10.1029/2009JDQ13556
Oian. Z: Liang. S: Yang. S: Trevathan. E: Huang. Z: Yang. R: Wang. J: Hu. K: Zhang. Y: Vaughn. M: Shea L: Liu.
W: Li. P: Ward. P: Yang. L: Zhang. W: Chen. W: Dong. G: Zheng. T: Xu. S: Zhang. B. (2015). Ambient
air pollution and preterm birth: A prospective birth cohort study in Wuhan, China. Int J Hyg Environ
Health219: 195-203. http://dx.doi.Org/10.1016/i.iiheh.2015.ll.003
Oiaa Z: Lia HM: Chinchilli. VM: Lehmaa EB: Duan. Y: Craig. TJ: Wilson. WE: Liao. D: Lazarus. SC: Bascom
R. (2009a). Interaction of ambient air pollution with asthma medication on exhaled nitric oxide among
asthmatics. Arch Environ Occup Health 64: 168-176. http://dx.doi.org/10.1080/1933824090324Q616
Oiaa Z: Lia HM: Chinchilli. VM: Lehmaa EB: Stewart. WF: Shah. N: Duaa Y: Craig. TJ: Wilsoa WE: Liao. D:
Lazarus. SC: Bascom. R. (2009b). Associations between air pollution and peak expiratory flow among
patients with persistent asthma. J Toxicol Environ Health A 72: 39-46.
http://dx.doi.org/10.1080/152873908Q2445517
Oia G: Meng. Z. (2006). The expressions of protooncogenes and CYP1A in lungs of rats exposed to sulfur dioxide
and benzo(a)pyrene. Regul Toxicol Pharmacol 45: 36-43. http://dx.doi.Org/10.1016/i.vrtph.2006.02.006
Oia G: Wang. J: Huo. Y: Yan. H: Jiang. C: Zhou. J: Wang. X: Sang. N. (2012). Sulfur dioxide inhalation stimulated
mitochondrial biogenesis in rat brains. Toxicology 300: 67-74. http://dx.doi.Org/10.1016/i.tox.2012.05.026
Oin. G: Wu. M: Wang. J: Xu. Z: Xia. J: Sang. N. (2016). Sulfur dioxide contributes to the cardiac and mitochondrial
dysfunction in rats. Toxicol Sci 151: 334-346.
Oia XD: Oiaa Z: Vaugha MG: Trevathaa E: Emo. B: Paul. G: Ren. WH: Hao. YT: Dong. GH. (2015). Gender-
specific differences of interaction between obesity and air pollution on stroke and cardiovascular diseases
in Chinese adults from a high pollution range area: A large population based cross sectional study. Sci
Total Environ 529: 243-248. http://dx.doi.org/10.1016/i.scitotenv.2015.05.041
Oiu. H: Yu. IT: Wang. X: Tian. L: Tse. LA: Wong. TW. (2013a). Cool and dry weather enhances the effects of air
pollution on emergency IHD hospital admissions. Int J Cardiol 168: 500-505.
http://dx.doi.Org/10.1016/i.iicard.2012.09.199
Oiu. H: Yu. ITS: Wang. X: Tiaa L: Tse. LA: Wong. TW. (2013b). Season and humidity dependence of the effects
of air pollution on COPD hospitalizations in Hong Kong. Atmos Environ 76: 74-80.
http://dx.doi.Org/10.1016/i.atmosenv.2012.07.026
Radwan. M: Jurewicz. J: Polanska. K: Sobala. W: Radwaa P: Bochenek. M: Hanke. W. (2015). Exposure to
ambient air pollution-does it affect semen quality and the level of reproductive hormones? Ann Hum Biol
43: 50-56. http://dx.doi.org/10.3109/03014460.2Q15.1013986
Rage. E: Siroux. V: Ktinzli. N: Pin. I: Kauffmann. F. (2009). Air pollution and asthma severity in adults. Occup
EnvironMed66: 182-188. http://dx.doi.org/10.1136/oem.2007.038349
Ranguelova. K: Chatteriee. S: Ehrenshaft. M: Ramirez. DC: Summers. FA: Kadiiska. MB: Mason. RP. (2010).
Protein radical formation resulting from eosinophil peroxidase-catalyzed oxidation of sulfite. J Biol Chem
285: 24195-24205. http://dx.doi.org/10.1074/ibc.M109.069Q54
Ranguelova. K: Rice. AB: Khaio. A: Triquigneaux. M: Garantziotis. S: Magliozzo. RS: Mason. RP. (2012).
Formation of reactive sulfite-derived free radicals by the activation of human neutrophils: an ESR study.
Free Radic Biol Med 52: 1264-1271. http://dx.doi.org/10.1016/i.freeradbiomed.2012.01.016
Ranguelova. K: Rice. AB: Lardinois. OM: Triquigneaux. M: Steinckwich. N: Deterding. LJ: Garantziotis. S: Mason.
RP. (2013). Sulfite-mediated oxidation of myeloperoxidase to a free radical: immuno-spin trapping
detection in human neutrophils. Free Radic Biol Med 60: 98-106.
http://dx.doi.Org/10.1016/i.freeradbiomed.2013.01.022
Rankia J: Chadwick. T: Nataraian. M: Howel. D: Pearce. MS: Pless-Mulloli. T. (2009). Maternal exposure to
ambient air pollutants and risk of congenital anomalies. Environ Res 109: 181-187.
http://dx.doi.Org/10.1016/i.envres.2008.ll.007
R-43
-------
Rao. KS. (2005). Uncertainty analysis in atmospheric dispersion modeling. Pure Appl Geophys 162: 1893-1917.
http://dx.doi.org/10.1007/s00024-0Q5-2697-4
Rao. ST: Galmarini. S: Puckett. K. (2011). Air Quality Model Evaluation International Initiative (AQMEII):
Advancing the state of the science in regional photochemical modeling and its applications. Bull Am
Meteorol Soc 92: 23-30. http://dx.doi.org/10.1175/2010BAMS3069.1
Raulf-Heimsoth. M: Hoffmever. F: van Thriel. C: Blaszkewicz. M: Btinger. J: Briining. T. (2010). Assessment of
low dose effects of acute sulphur dioxide exposure on the airways using non-invasive methods. Arch
Toxicol 84: 121-127. http://dx.doi.org/10.1007/s00204-009-048Q-5
Reddel. HK. (2009). Characterizing asthma phenotypes: Predictors and outcomes at the extremes of asthma severity
[Editorial]. Respirology 14: 778-780. http://dx.doi.Org/10.llll/i.1440-1843.2009.01593.x
Reddel. HK: Taylor. PR: Bateman. ED: Boulet. LP: Boushev. HA: Busse. WW: Casale. TB: Chanez. P: Enright.
PL: Gibsoa PG: de Jongste. JC: Kerstiens. HA: Lazarus. SC: Lew. ML: O'Bvrne. PM: Partridge. MR:
Pavord. ID: Sears. MR: Sterk. PJ: Stoloff. SW: Sullivaa SD: Szefler. SJ: Thomas. MP: Wenzel. SE.
(2009). An official American Thoracic Society/European Respiratory Society statement: Asthma control
and exacerbations: Standardizing endpoints for clinical asthma trials and clinical practice. Am J Respir Crit
Care Med 180: 59-99. http://dx.doi.org/10.1164/rccm.200801-06QST
Reddv. P: Naidoo. RN: Robins. TG: Mentz. G: Li. H: Londoa SJ: Batterman. S. (2012). GSTM1 and GSTP1 gene
variants and the effect of air pollutants on lung function measures in South African children. Am J Ind Med
55: 1078-1086. http://dx.doi.org/10.1002/aiim.22012
Reeves. GK: Cox. PR: Darby. SC: Whitley. E. (1998). Some aspects of measurement error in explanatory variables
for continuous and binary regression models. Stat Med 17: 2157-2177.
http://dx.doi.org/10.1002/(SICI) 1097-0258(19981015) 17:19<2157: :AIP-SIM916>3.0.CQ:2-F
Rehbein. PJG: Kennedy. MG: Cotsmaa PJ: Campeau. MA: Greenfield. MM: Annett. MA: Lepage. MF. (2014).
Combined analysis of modeled and monitored S02 concentrations at a complex smelting facility. J Air
Waste Manag Assoc 64: 272-279. http://dx.doi.org/10.1080/10962247.2Q13.856817
Reiss. J: Bonin. M: Schwegler. H: Sass. JO: Garattini. E: Wagner. S: Lee. HJ: Engel. W: Riess. O: Schwarz. G.
(2005). The pathogenesis of molybdenum cofactor deficiency, its delay by maternal clearance, and its
expression pattern in microarray analysis. Mol Genet Metab 85: 12-20.
http://dx.doi.Org/10.1016/i.vmgme.2005.01.008
Rich. PO: Pemissie. K: Lu. SE: Kamat. L: Wallenberg. P: Rhoads. GG. (2009). Ambient air pollutant
concentrations during pregnancy and the risk of fetal growth restriction. J Epidemiol Community Health
63: 488-496. http://dx.doi.org/10.1136/iech.2008.082792
Rich. PO: Kipea HM: Huang. W: Wang. G: Wang. Y: Zhu. P: Ohman-Strickland. P: Hu. M: Philipp. C: Piehl. SR:
Lu. SE: Tong. J: Gong. J: Thomas. P: Zhu. T: Zhang. JJ. (2012). Association between changes in air
pollution levels during the Beijing Olympics and biomarkers of inflammation and thrombosis in healthy
young adults. JAMA 307: 2068-2078. http://dx.doi.org/10.1001/iama.2012.3488
Rich. PO: Kipen. HM: Zhang. J: Kamat. L: Wilsoa AC: Kostis. JB. (2010). Triggering of transmural infarctions,
but not nontransmural infarctions, by ambient fine particles. Environ Health Perspect 118: 1229-1234.
http://dx.doi.org/10.1289/ehp.0901624
Rich. PO: Liu. K: Zhang. J: Thurston. SW: Stevens. TP: Pan. Y: Kane. C: Weinberger. B: Ohman-Strickland. P:
Woodruff. TJ: Puaa X: Assibev-Mensah. V: Zhang. J. (2015). Pifferences in birth weight associated with
the 2008 Beijing Olympic air pollution reduction: results from a natural experiment. Environ Health
Perspect 123: 880-887. http://dx.doi.org/10.1289/ehp.1408795
Riedel. F: Kramer. M: Scheibenbogea C: Rieger. CHL. (1988). Effects of S02 exposure on allergic sensitization in
the guinea pig. J Allergy Clin Immunol 82: 527-534. http://dx.doi.org/10.1016/0091-6749(88)90961-X
Ritz. B: Wilhelm. M. (2008). Ambient air pollution and adverse birth outcomes: Methodologic issues in an
emerging field [Review]. Basic Clin Pharmacol Toxicol 102: 182-190. http://dx.doi.org/10. Ill 1/i. 1742-
7843.2007.00161.x
R-44
-------
Rivera-Gonzalez. LO: Zhang. Z: Sanchez. BN: Zhang. K: Browa DG: Roias-Bracho. L: Osornio-Vargas. A:
Vadillo-Ortega. F: O'Neill. MS. (2015). An assessment of air pollutant exposure methods in Mexico City,
Mexico. J Air Waste Manag Assoc 65: 581-591. http://dx.doi.org/10.1080/10962247.2015.102Q974
Robledo. CA: Mendola. P: Yeung. E: Mannisto. T: Sundaram. R: Liu. D: Ying. O: Shermaa S: Grantz. KL. (2015).
Preconception and early pregnancy air pollution exposures and risk of gestational diabetes mellitus.
Environ Res 137: 316-322. http://dx.doi.Org/10.1016/i.envres.2014.12.020
Roger. LJ: Kehrl. HR: Hazucha. M: Horstmaa PH. (1985). Bronchoconstriction in asthmatics exposed to sulfur
dioxide during repeated exercise. J Appl Physiol (1985) 59: 784-791.
Romieu. I: Meneses. F: Ruiz. S: Sienra. JJ: Huerta. J: White. MC: Etzel. RA. (1996). Effects of air pollution on the
respiratory health of asthmatic children living in Mexico City. Am J Respir Crit Care Med 154: 300-307.
http://dx.doi.Org/10.1164/airccm.154.2.8756798
Rood. AS. (2014). Performance evaluation of AERMOD, CALPUFF, and legacy air dispersion models using the
Winter Validation Tracer Study dataset. Atmos Environ 89: 707-720.
http://dx.doi.Org/10.1016/i.atmosenv.2014.02.054
Roonev. AA: Bovles. AL: Wolfe. MS: Bucher. JR: Thaver. KA. (2014). Systematic review and evidence integration
for literature-based environmental health science assessments. Environ Health Perspect 122: 711-718.
http://dx.doi.org/10.1289/ehp. 1307972
Rosenlund. M: Berglind. N: Pershagea G: Hallavist. J: Jonsoa T: Bellander. T. (2006). Long-term exposure to
urban air pollution and myocardial infarction. Epidemiology 17: 383-390.
http://dx.doi.org/10.1097/01.ede.000Q219722.25569.0f
Rosenthal. FS: Kuisma. M: Lanki. T: Hussein. T: Bovd. J: Halonea JI: Pekkanen. J. (2013). Association of ozone
and particulate air pollution with out-of-hospital cardiac arrest in Helsinki, Finland: evidence for two
different etiologies. J Expo Sci Environ Epidemiol 23: 281-288. http://dx.doi.org/10.1038/ies.2012.121
Rosenthal. RA: Kavic. SM. (2004). Assessment and management of the geriatric patient [Review]. Crit Care Med
32: S92-105. http://dx.doi.org/10.1097/01.CCM.0000122Q69.56161.97
Rothman. KJ: Greenland. S. (1998). Modern epidemiology. In Modern Epidemiology (2 ed.). Philadelphia, PA:
Lippincott-Raven.
Routledge. HC: Mannev. S: Harrisoa RM: Avres. JG: Townend. JN. (2006). Effect of inhaled sulphur dioxide and
carbon particles on heart rate variability and markers of inflammation and coagulation in human subjects.
Heart 92: 220-227. http://dx.doi.org/10.1136/hrt.2004.051672
Roy. A: Gong. J: Thomas. DC: Zhang. J: Kipen. HM: Rich. DO: Zhu. T: Huang. W: Hu. M: Wang. G: Wang. Y:
Zhu. P: Lu. SE: Ohman-Strickland. P: Diehl. SR: Eckel. SP. (2014). The cardiopulmonary effects of
ambient air pollution and mechanistic pathways: a comparative hierarchical pathway analysis. PLoS ONE
9: ell4913. http://dx.doi.org/10.1371/iournal.pone.0114913
Ruaa A: Mia H: Meng. Z: Lu. Z. (2003). Protective effects of seabuckthorn seed oil on mouse injury induced by
sulfur dioxide inhalation. Inhal Toxicol 15: 1053-1058. http://dx.doi.org/10.1080/0895837039Q226558
Rubinstein. I: Bigbv. BG: Reiss. TF: Boushev. HA. Jr. (1990). Short-term exposure to 0.3 ppm nitrogen dioxide
does not potentiate airway responsiveness to sulfur dioxide in asthmatic subjects. Am Rev Respir Dis 141:
381-385. http://dx.doi.Org/10.1164/airccm/141.2.381
Rusznak. C: Devalia. JL: Davies. RJ. (1996). Airway response of asthmatic subjects to inhaled allergen after
exposure to pollutants. Thorax 51: 1105-1108. http://dx.doi.org/10.1136/thx.51.11.1105
Sacks. J. (2015). Email exchange between Matthew Strickland, Emory University and Jason Sacks, U.S. EPA. RE:
Response to Question about 2010 Am J Respir Crit Care Med Paper. Available online at
http://www.re gulations. gov/#! documentDetail:D=EPA-HO-ORD-2Q 13 -03 57-0003
Sacks. JD: Ito. K: Wilsoa WE: Neas. LM. (2012). Impact of covariate models on the assessment of the air
pollution-mortality association in a single- and multipollutant context. Am J Epidemiol 176: 622-634.
http://dx.doi.org/10.1093/aie/kwsl35
R-45
-------
Sagiv. SK: Mendola. P: Loomis. D: Herring. AH: Neas. LM: Savitz. DA: Poole. C. (2005). A time-series analysis of
air pollution and preterm birth in Pennsylvania, 1997-2001. Environ Health Perspect 113: 602-606.
http://dx.doi.org/10.1289/ehp.7646
Sahsuvaroglu. T: Jerrett. M: Sears. MR: Mcconnell. R: Finkelsteia N: Arain. A: Newbold. B: Burnett. R. (2009).
Spatial analysis of air pollution and childhood asthma in Hamilton, Canada: comparing exposure methods
in sensitive subgroups. Environ Health 8: Z. http://dx.doi.org/10.1186/1476-069X-8-14
Saigal. S: Dovle. LW. (2008). An overview of mortality and sequelae of preterm birth from infancy to adulthood.
Lancet 371: 261-269. http://dx.doi.org/10.1016/S0140-6736(08)60136-1
Samoli. E: Nastos. PT: Paliatsos. AG: Katsouvanni. K: Priftis. KN. (2011). Acute effects of air pollution on
pediatric asthma exacerbation: Evidence of association and effect modification. Environ Res 111: 418-424.
http://dx.doi.Org/10.1016/i.envres.2011.01.014
San Tam. W: Wong. T: Wong. AHS. (2015). Association between air pollution and daily mortality and hospital
admission due to ischaemic heart diseases in Hong Kong. Atmos Environ 120: 360-368.
http://dx.doi.Org/10.1016/i.atmosenv.2015.08.068
Sander. SP: Abbatt. JPD: Barker. JR: Burkholder. JB: Friedl. RR: Goldea DM: Huie. RE: Kolb. CE: Kurvlo. MJ:
Moortgat. GK: Orkin. VL: Wine. PH. (2011). Chemical kinetics and photochemical data for use in
atmospheric studies: Evaluation number 17. (JPL Publication 10-6). Pasadena, CA: Jet Propulsion
Laboratory, http://ipldataeval.ipl.nasa. gov/pdf/JPL%2010-6%20Final%2015June2011 .pdf
Sang. N: Yua Y: Li. H: Hou. L: Han. M: Li. G. (2010). S02 inhalation contributes to the development and
progression of ischemic stroke in the brain. Toxicol Sci 114: 226-236.
Sarnat. JA: Browa KW: Schwartz. J: Coull. BA: Koutrakis. P. (2005). Ambient gas concentrations and personal
particulate matter exposures: Implications for studying the health effects of particles. Epidemiology 16:
385-395. http://dx.doi.org/10.1097/01.ede.0000155505.Q4775.33
Sarnat. JA: Koutrakis. P: Suh. HH. (2000). Assessing the relationship between personal particulate and gaseous
exposures of senior citizens living in Baltimore, MD. J Air Waste Manag Assoc 50: 1184-1198.
http://dx.doi.org/10.1080/10473289.200Q.10464165
Sarnat. JA: Schwartz. J: Catalano. PJ: Suh. HH. (2001). Gaseous pollutants in particulate matter epidemiology:
Confounders or surrogates? Environ Health Perspect 109: 1053-1061.
Sarnat. SE. (2016). Response to data question regarding: Ambient air pollution and emergency department visits for
asthma: A multi-city assessment of effect modification by age. Sarnat, SE.
Sarnat. SE: Coull. BA: Schwartz. J: Gold. PR: Suh. HH. (2006). Factors affecting the association between ambient
concentrations and personal exposures to particles and gases. Environ Health Perspect 114: 649-654.
http://dx.doi.org/10.1289/ehp.8422
Sarnat. SE: Klein. M: Sarnat. JA: Flanders. WD: Waller. LA: Mulholland. JA: Russell. AG: Tolbert. PE. (2010). An
examination of exposure measurement error from air pollutant spatial variability in time-series studies. J
Expo Sci Environ Epidemiol 20: 135-146. http://dx.doi.org/10.1038/ies.2009.10
Schachter. EN: Witek. TJ. Jr: Beck. GJ: Hosein. HR: Colice. G: Leaderer. BP: Cain. W. (1984). Airway effects of
low concentrations of sulfur dioxide: dose-response characteristics. Arch Environ Occup Health 39: 34-42.
Schauberger. G: Piringer. M: Schmitzer. R: Kamp. M: Sowa. A: Koch. R: Eckhof. W: Grimm E: Kypke. J:
Hartung. E. (2012). Concept to assess the human perception of odour by estimating short-time peak
concentrations from one-hour mean values. Reply to a comment by Janicke et al. Atmos Environ 54: 624-
628. http://dx.doi.Org/10.1016/i.atmosenv.2012.02.017
Schildcrout. JS: Sheppard. L: Lumlev. T: Slaughter. JC: Koenig. JO: Shapiro. GG. (2006). Ambient air pollution
and asthma exacerbations in children: An eight-city analysis. Am J Epidemiol 164: 505-517.
http://dx.doi.org/10.1093/aie/kwi225
Schiller. JS: Lucas. JW: Ward. BW: Peregov. JA. (2012). Summary health statistics for U.S. adults: National Health
Interview Survey, 2010. (DHHS Publication No. (PHS) 20121580). Hyattsville, MD: National Center for
Health Statistics, http://www.cdc.gov/nchs/data/series/sr 10/srlQ 252.pdf
R-46
-------
Schlesinger. WH. (1997). Biogeochemistry: An analysis of global change. In Biogeochemistry: An Analysis of
Global Change (2 ed.). San Diego: Academic Press.
Schwab. M: Mcdermott. A: Spengler. JD. (1992). USING LONGITUDINAL DATA TO UNDERSTAND
CHILDRENS ACTIVITY PATTERNS IN AN EXPOSURE CONTEXT - DATA FROM THE
KANAWHA COUNTY HEALTH STUDY. Environ Int 18: 173-189. http://dx.doi.org/10.1016/0160-
4120(92)90005-0
Schwab. RJ: Kim. C: Bagchi. S: Keenan. BT: Comva FL: Wang. S: Tapia. IE: Huang. S: Travlor. J: Torigian. DA:
Bradford. RM: Marcus. CL. (2015). Understanding the anatomic basis for obstructive sleep apnea
syndrome in adolescents. Am J Respir Crit Care Med 191: 1295-1309.
http://dx.doi.org/10.1164/rccm.201501-0169QC
Schwartz. J. (1989). Lung function and chronic exposure to air pollution: a cross-sectional analysis of NHANES II.
Environ Res 50: 309-321.
Schwartz. J. (1995). Short term fluctuations in air pollution and hospital admissions of the elderly for respiratory
disease. Thorax 50: 531-538. http://dx.doi.Org/10.1136/thx.50.5.531
Schwartz. J. (1997). Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology 8:
371-377. http://dx.doi.org/10.1097/00001648-199707000-000Q4
Schwartz. J: Morris. R. (1995). Air pollution and hospital admissions for cardiovascular disease in Detroit,
Michigan. Am J Epidemiol 142: 23-35.
Schwartz. J: Spix. C: Touloumi. G: Bacharova. L: Barumamdzadeh. T: le Tertre. A: Piekarksi. T: Ponce de Leoa A:
Ponka. A: Rossi. G: Saez. M: Schouten. JP. (1996). Methodological issues in studies of air pollution and
daily counts of deaths or hospital admissions. J Epidemiol Community Health 1: S3-S11.
Segala. C: Fauroux. B: Just. J: Pascual. L: Grimfeld. A: Neukirch. F. (1998). Short-term effect of winter air
pollution on respiratory health of asthmatic children in Paris. Eur Respir J 11: 677-685.
Segala. C: Poizeau. D: Mesbah. M: Willems. S: Maidenberg. M. (2008). Winter air pollution and infant bronchiolitis
in Paris. Environ Res 106: 96-100. http://dx.doi.Org/10.1016/i.envres.2007.05.003
Seinfeld. JH: Pandis. S. (2006). Atmospheric Chemistry and Physics. Hoboken, NJ: John Wiley.
Shadwick. PS: Sickles. IJE. (2004). Sample size for seasonal mean concentration, deposition velocity and
deposition: A resampling study. Atmos Environ 38: 477-489.
Shah. AS: Lee. KK: Mcallister. DA: Hunter. A: Nair. H: Whitelev. W: Langrish. JP: Newbv. DE: Mills. NL. (2015).
Short term exposure to air pollution and stroke: systematic review and meta-analysis [Review]. B M J
(Online) 350: hl295. http://dx.doi.org/10.1136/bmi.hl295
Shallcross. DE: Taaties. CA: Percival. CJ. (2014). Criegee intermediates in the indoor environment: new insights.
Indoor Air 24: 495-502. http://dx.doi.org/10. Ill 1/ina. 12102
Shen. X: Wu. H: Zhao. Y. ue: Huang. D. ao: Huang. L: Chen. Z. (2016). Heterogeneous reactions of glyoxal on
mineral particles: A new avenue for oligomers and organosulfate formation. Atmos Environ 131: 133-140.
http://dx.doi.Org/10.1016/i.atmosenv.2016.01.048
Sheppard. D: Epstein. J: Bethel. RA: Nadel. JA: Boushev. HA. (1983). Tolerance to sulfur dioxide-induced
bronchoconstriction in subjects with asthma. Environ Res 30: 412-419. http://dx.doi.org/10.1016/0Q13-
9351(83)90227-X
Sheppard. D: Wong. WS: Uehara. CF: Nadel. JA: Boushev. HA. (1980). Lower threshold and greater bronchomotor
responsiveness of asthmatic subjects to sulfur dioxide. Am Rev Respir Dis 122: 873-878.
Sheppard. L. (2003). Ambient air pollution and nonelderly asthma hospital admissions in Seattle, Washington, 1987-
1994. In Revised analyses of time-series studies of air pollution and health (pp. 227-230). Boston, MA:
Health Effects Institute. http://pubs.healtheffects.org/view.php?id=4
Sheppard. L. (2005). Acute air pollution effects: Consequences of exposure distribution and measurements. J
Toxicol Environ Health A 68: 1127-1135. http://dx.doi.org/10.1080/1528739059Q935987
R-47
-------
Sheppard. L: Lew. D: Norris. G: Larson. TV: Koenig. JO. (1999). Effects of ambient air pollution on nonelderly
asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
http://dx.doi.org/10.1097/00001648-199901000-000Q6
Sheppard. L: Slaughter. JC: Schildcrout. J: Liu. JS: Lumlev. T. (2005). Exposure and measurement contributions to
estimates of acute air pollution effects. J Expo Anal Environ Epidemiol 15: 366-376.
http://dx.doi.org/10.1038/si.iea.7500413
Shim. SR: Kim. JH: Song. YS: Lee. WJ. (2015). Association between air pollution and benign prostatic hyperplasia:
an ecological study. Arch Environ Occup Health 71:0. http://dx.doi.org/10.1080/19338244.2Q15.1093458
Sickles. JE: Shadwick. PS. (2007). Seasonal and regional air quality and atmospheric deposition in the eastern
United States. J Geophys Res 112: D17302. http://dx.doi.org/10.1029/2006JD0Q8356
Silvermaa RA: Ito. K: Freese. J: Kaufman. BJ: De Claro. D: Braua J: Prezant. DJ. (2010). Association of ambient
fine particles with out-of-hospital cardiac arrests in New York City. Am J Epidemiol 172: 917-923.
http://dx.doi.org/10.1093/aie/kwa217
Sim. VM: Pattle. RE. (1957). Effect of possible smog irritants on human subjects. JAMA 165: 1908-1913.
Simeonssoa JB: Matta. A: Boddeti. R. (2012). Direct measurement of S02 in air by laser induced fluorescence
spectrometry using a nontunable laser source. Anal Lett 45: 894-906.
http://dx.doi.org/10.1080/00032719.2Q12.655681
Simpson. D: Winiwarter. W: Boriesson. G: Cinderbv. S: Ferreiro. A: Guenther. A: Hewitt. CN: Jansoa R: Khalil.
MAK: Owea S: Pierce. TE: Puxbaum. H: Shearer. M: Skiba. U: Steinbrecher. R: Tarrasoa L: Qquist. MG.
(1999). Inventorying emissions from nature in Europe. J Geophys Res Atmos 104: 8113-8152.
http://dx.doi.org/10.1029/98JDQ2747
Sinclair. AH: Edgerton. ES: Wvzga. R: Tolsma. D. (2010). A two-time-period comparison of the effects of ambient
air pollution on outpatient visits for acute respiratory illnesses. J Air Waste Manag Assoc 60: 163-175.
http://dx.doi.Org/10.3155/1047-3289.60.2.163
Sinclair. AH: Tolsma. D. (2004). Associations and lags between air pollution and acute respiratory visits in an
ambulatory care setting: 25-month results from the Aerosol Research and Inhalation Epidemiological
Study. J Air Waste Manag Assoc 54: 1212-1218. http://dx.doi.org/10.1080/10473289.2004.10470979
Singer. BC: Hodgsoa AT: Hotchi. T: Kim JJ. (2004). Passive measurement of nitrogen oxides to assess traffic-
related pollutant exposure for the East Bay Children's Respiratory Health Study. Atmos Environ 38: 393-
403. http://dx.doi.Org/10.1016/i.atmosenv.2003.10.005
Slade. D. (1968a). Meteorological instruments for use in the atomic energy industry. InDH Slade (Ed.),
Meteorology and atomic energy (pp. 257-300). Washington, DC: U.S. Atomic Energy Commission.
Slade. PH. (1968b). Meteorology and atomic energy 1968. InDH Slade (Ed.). Washington, DC: U.S. Atomic
Energy Commission. https://ntrl.ntis.gov/NTRL/dashboard/searchResults.xhtml?searchQuery=TID24190
Slama. R: Bottagisi. S: Solanskv. I: Lepeule. J: Giorgis-Allemand. L: Sram R. (2013). Short-term impact of
atmospheric pollution on fecundability. Epidemiology 24: 871-879.
http://dx.doi.org/10.1097/EDE.0b013e3182a7Q2c5
Slama. R: Darrow. L: Parker. J: Woodruff. TJ: Strickland. M: Nieuwenhuiisea M: Glinianaia. S: Hoggatt. KJ:
Kannaa S: Hurley. F: Kalinka. J: Sram R: Brauer. M: Wilhelm. M: Heinrich. J: Ritz. B. (2008). Meeting
report: Atmospheric pollution and human reproduction. Environ Health Perspect 116: 791-798.
http://dx.doi.org/10.1289/ehp. 11074
Smargiassi. A: Kosatskv. T: Hicks. J: Plante. C: Armstrong. B. en: Villeneuve. PJ: Goudreau. S. (2009). Risk of
asthmatic episodes in children exposed to sulfur dioxide stack emissions from a refinery point source in
Montreal, Canada. Environ Health Perspect 117: 653-659. http://dx.doi.org/10.1289/ehp.080001Q
Smith. E. (1993). [Subject data supplied by the researchers for the recent controlled human studies analyzed in the
staff paper supplement and accompanying memorandum. Memorandum to docket no. A-84-25, item IV-B-
5], Research Triangle Park, NC: U.S. Environmental Protection Agency.
http://www.re gulations. gov/#! documentDetail:D=EPA-HO-OAR-2007-Q3 52-1368
R-48
-------
Smith. GS: VanDenEedea SK: Garcia. C: Shaa J: Baxter. R: Herring. AH: Richardsoa DB: VanRie. A: Emch.
M: Gammoa MP. (2016). Air pollution and pulmonary tuberculosis: A nested case-control study among
members of a northern California health plan. Environ Health Perspect 124: 761-768.
http://dx.doi.org/10.1289/ehp. 1408166
Smith. LG: Busch. RH: Buschbom. RL: Cannon. WC: Loscutoff. SM: Morris. JE. (1989). Effects of sulfur dioxide
or ammonium sulfate exposure, alone or combined, for 4 or 8 months on normal and elastase-impaired rats.
Environ Res 49: 60-78. http://dx.doi.org/10.1016/S0013-9351(89)80022-2
Smith. TJ: Eatough. DJ: Hansen. LP: Mangelsen. NF. (1976). The chemistry of sulfur and arsenic in airborne
copper smelter particulates. Bull Environ Contam Toxicol 15: 651-659.
http://dx.doi.org/10.1007/BF01685613
Snashall. PP: Baldwin. C. (1982). Mechanisms of sulphur dioxide induced bronchoconstriction in normal and
asthmatic man. Thorax 37: 118-123. http://dx.doi.Org/10.1136/thx.37.2.118
Snell. RE: Luchsinger. PC. (1969). Effects of sulfur dioxide on expiratory flow rates and total respiratory resistance
in normal human subjects. Arch Environ Occup Health 18: 693-698.
http://dx.doi.org/10.1080/00Q39896.1969.10665472
Soa JY: Bell. ML: Lee. JT. (2010). Individual exposure to air pollution and lung function in Korea: Spatial analysis
using multiple exposure approaches. Environ Res 110: 739-749.
http://dx.doi.Org/10.1016/i.envres.2010.08.003
Soa JY: Cho. YS: Lee. JT. (2008). Effects of air pollution on postneonatal infant mortality among firstborn infants
in Seoul, Korea: Case-crossover and time-series analyses. Arch Environ Occup Health 63: 108-113.
http://dx.doi.Org/10.3200/AEQH.63.3.108-113
Soa JY: Lee. JT: Park. YH: Bell. ML. (2013). Short-term effects of air pollution on hospital admissions in Korea.
Epidemiology 24: 545-554. http://dx.doi.org/10.1097/EPE.0b013e3182953244
Song. A: Liao. O: Li. J: Lia F: Liu. E: Jiang. X: Peng. L. (2012). Chronic exposure to sulfur dioxide enhances
airway hyperresponsiveness only in ovalbumin-sensitized rats. Toxicol Lett 214: 320-327.
http://dx.doi.Org/10.1016/i.toxlet.2012.09.010
Soto-Ramos. M: Castro-Rodriguez. JA: Hinoios-Gallardo. LC: Hernandez-Saldafia. R: Cisneros-Castolo. M:
Carrillo-Rodriguez. V. (2013). Fractional exhaled nitric oxide has a good correlation with asthma control
and lung function in Latino children with asthma. J Asthma 50: 590-594.
http://dx.doi.org/10.3109/02770903.2Q13.792349
Souto. JA: Moral. C: Rodriguez. A: Saavedra. S: Casares. JJ. (2014). Simulation of plume dispersion using different
stack configurations and meteorological inputs. Int J Environ Pollut 55: 139-147.
http://dx.doi.org/10.1504/IJEP.2014.065917
Sovseth. V: Kongerud. J: Broea P: Lilleng. P: Boe. J. (1995). Bronchial responsiveness, eosinophilic and short
term exposure to air pollution. ArchPis Child 73: 418-422. http://dx.doi.Org/10.1136/adc.73.5.418
Spalt. EW: Curl. CL: Allea RW: Cohen. M: Williams. K: Hirsch. JA: Adar. SP: Kaufman. JP. (2015). Factors
influencing time-location patterns and their impact on estimates of exposure: the Multi-Ethnic Study of
Atherosclerosis and Air Pollution (MESA Air). J Expo Sci Environ Epidemiol 26: 341-348.
http://dx.doi.org/10.1038/ies.2015.26
Speizer. FE: Frank. NR. (1966). The uptake and release of S02 by the human nose. Arch Environ Occup Health 12:
725-728. http://dx.doi.org/10.1080/00039896.1966.10664471
Spiegelmaa P. (2013). Regression calibration in air pollution epidemiology with exposure estimated by spatio-
temporal modeling. Environmetrics 24: 521-524. http://dx.doi.org/10.10Q2/env.2249
Spira-Cohen. A. (2013). Email fromPr. Spira-Cohen to Pr. Patel; Response to data request. Available online at
https://www.regulations.gov/document?P=EPA-HO-ORP-2013-0232-0Q15
Spira-Cohen. A: Chen. LC: Kendall. M: Lall. R: Thurston. GP. (2011). Personal exposures to traffic-related air
pollution and acute respiratory health among Bronx schoolchildren with asthma. Environ Health Perspect
119: 559-565. http://dx.doi.org/10.1289/ehp. 1002653
R-49
-------
Stanoievic. S: Wade. A: Stocks. J: Hankinson. J: Coates. AL: Paa H: Rosenthal. M: Corey. M: Lebecaue. P: Cole.
TJ. (2008). Reference ranges for spirometry across all ages: A new approach. Am J Respir Crit Care Med
177: 253-260. http://dx.doi.org/10.1164/rccm.200708-1248QC
Steinvil. A: Fireman. E: Kordova-Biezuner. L: Cohen. M: Shapira. I: Berliner. S: Rogowski. O. (2009).
Environmental air pollution has decremental effects on pulmonary function test parameters up to one week
after exposure. Am J Med Sci 338: 273-279. http://dx.doi.org/10.1097/MAJ.ObO 13e3181 adb3ed
Steinvil. A: Kordova-Biezuner. L: Shapira. I: Berliner. S: Rogowski. O. (2008). Short-term exposure to air pollution
and inflammation-sensitive biomarkers. Environ Res 106: 51-61.
http://dx.doi.Org/10.1016/i.envres.2007.08.006
Stieb. DM: Beveridge. RC: Brook. JR: Smith-Doiroa M: Burnett. RT: Dales. RE: Beaulieu. S: Judek. S: Mamedov.
A. (2000). Air pollution, aeroallergens and cardiorespiratory emergency department visits in Saint John,
Canada. J Expo Anal Environ Epidemiol 10: 461-477. http://dx.doi.org/10.1038/si.iea.7500112
Stieb. DM: Judek. S: Burnett. RT. (2003). Meta-analysis of time-series studies of air pollution and mortality: Update
in relation to the use of generalized additive models. J Air Waste Manag Assoc 53: 258-261.
http://dx.doi.org/10.1080/10473289.20Q3.10466149
Stieb. DM: Szvszkowicz. M: Rowe. BH: Leech. JA. (2009). Air pollution and emergency department visits for
cardiac and respiratory conditions: A multi-city time-series analysis. Environ Health 8: 25.
http://dx.doi.org/10.1186/1476-069X-8-25
Stingone. JA: Luben. TJ: Daniels. JL: Fuentes. M: Richardson. DB: Avlsworth. AS: Herring. AH: Anderka. M:
Botto. L: Correa. A: Gilboa. SM: Langlois. PH: Moslev. B: Shaw. GM: Siffel. C: Olshaa AF. (2014).
Maternal exposure to criteria air pollutants and congenital heart defects in offspring: Results from the
national birth defects prevention study. Environ Health Perspect 122: 863-872.
http://dx.doi.org/10.1289/ehp. 1307289
Stranev. L: Fina J: Dennekamp. M: Bremner. A: Tonkia A: Jacobs. I. (2014). Evaluating the impact of air pollution
on the incidence of out-of-hospital cardiac arrest in the Perth Metropolitan Region: 2000-2010. J Epidemiol
Community Health 68: 6-12. http://dx.doi.org/10.1136/iech-2013-2Q2955
Streets. DG: de Fov. B: Duncan. BN: Lamsal. LN: Li. C: Lu. Z. (2014). Using satellite observations to measure
power plant emissions and their trends [Magazine]. EM: Environmental Manager, February, 16-21.
Strickland. MJ: Darrow. LA: Klein. M: Flanders. WD: Sarnat. JA: Waller. LA: Sarnat. SE: Mulholland. JA: Tolbert.
PE. (2010). Short-term associations between ambient air pollutants and pediatric asthma emergency
department visits. Am J Respir Crit Care Med 182: 307-316. http://dx.doi.org/10.1164/rccm.2009Q8-
1201QC
Strickland. MJ: Darrow. LA: Mulholland. JA: Kleia M: Flanders. WD: Winquist. A: Tolbert. PE. (2011).
Implications of different approaches for characterizing ambient air pollutant concentrations within the
urban airshed for time-series studies and health benefits analyses. Environ Health 10: 36.
http://dx.doi.org/10.1186/1476-069X-10-36
Strickland. MJ: Kleia M: Correa. A: Reller. MP: Mahle. WT: Riehle-Colarusso. TJ: Botto. LP: Flanders. WD:
Mulholland. JA: Siffel. C: Marcus. M: Tolbert. PE. (2009). Ambient air pollution and cardiovascular
malformations in Atlanta, Georgia, 1986-2003. Am J Epidemiol 169: 1004-1014.
http://dx.doi.org/10.1093/aie/kwpQ 11
Sua Y: Song. X: Haa Y: Ji. Y: Gao. S: Shang. Y: Lu. SE: Zhu. T: Huang. W. (2015). Size-fractioned ultrafine
particles and black carbon associated with autonomic dysfunction in subjects with diabetes or impaired
glucose tolerance in Shanghai, China. Part Fibre Toxicol 12: 8. http://dx.doi.org/10.1186/sl2989-015-0Q84-
6
Sunver. J: Ballester. F: Le Tertre. A: Atkinsoa R: Avres. JG: Forastiere. F: Forsberg. B: Vonk. JM: Bisanti. L:
Tenias. JM: Medina. S: Schwartz. J: Katsouvanni. K. (2003). The association of daily sulfur dioxide air
pollution levels with hospital admissions for cardiovascular diseases in Europe (The Aphea-II study). Eur
Heart J 24: 752-760. http://dx.doi.org/10.1016/S0195-668X(02)00808-4
R-50
-------
Swaans. W: Goelen. E: De Fre. R: Damea E: Van Avermaet. P: Roekens. E: Keppens. V. (2007). Laboratory and
field validation of a combined N02-S02 Radiello passive sampler. J Environ Monit 9: 1231-1240.
http://dx.doi.org/10.1039/b7Q8925b
Svkes. RI: Parker. S: Henn. D: Chowdhurv. B. (2007). SCIPUFF version 2.3 technical documentation. Princeton,
New Jersey: L-3 Titan Corp.
Szpiro. AA: Paciorek. CJ. (2013). Measurement error in two-stage analyses, with application to air pollution
epidemiology. Environmetrics 24: 501-517. http://dx.doi.org/10.1002/env.2233
Szpiro. AA: Paciorek. CJ: Sheppard. L. (2011). Does more accurate exposure prediction necessarily improve health
effect estimates? Epidemiology 22: 680-685. http://dx.doi.org/10.1097/EDE.0b013e3182254cc6
Szvszkowicz. M. (2008). Ambient air pollution and daily emergency department visits for ischemic stroke in
Edmonton, Canada. Int J Occup Med Environ Health 21: 295-300. http://dx.doi.org/10.2478/vlQ001-008-
0029-5
Szvszkowicz. M: Porada. E: Tremblav. N: Grafstein. E. (2012a). Sulfur dioxide and emergency department visits for
stroke and seizure. Stroke Res Treat 2012: 824724. http://dx.doi.org/10.1155/2012/824724
Szvszkowicz. M: Rowe. BH: Brook. RD. (2012b). Even low levels of ambient air pollutants are associated with
increased emergency department visits for hypertension. Can J Cardiol 28: 360-366.
http://dx.doi.Org/10.1016/i.cica.2011.06.011
Taggart. SCO: Custovic. A: Francis. HC: Faragher. EB: Yates. CJ: Higgins. BG: Woodcock. A. (1996). Asthmatic
bronchial hyperresponsiveness varies with ambient levels of summertime air pollution. Eur Respir J 9:
1146-1154. http://dx.doi.org/10.1183/09031936.96.09061146
Tam E: Miike. R: Labrenz. S: Sutton. AJ: Elias. T: Davis. J: Chen. YL: Tantisira. K: Dockerv. D: Avol. E. (2016).
Volcanic air pollution over the Island of Hawai'i: Emissions, dispersal, and composition. Association with
respiratory symptoms and lung function in Hawai'i Island school children. Environ Int 92-93: 543-552.
http://dx.doi.Org/10.1016/i.envint.2016.03.025
Taa WC: Cripps. E: Douglas. N: Sudlow. MF. (1982). Protective effect of drugs on bronchoconstriction induced by
sulphur dioxide. Thorax 37: 671-676. http://dx.doi.Org/10.1136/thx.37.9.671
TFESC and NASPE (Task Force of the European Society of Cardiology and the North American Society of Pacing
and Electrophysiology). (1996). Heart rate variability: Standards of measurement, physiological
interpretation and clinical use. Circulation 93: 1043-1065. http://dx.doi.Org/10.1161/01.CIR.93.5.1043
Thach. TO: Wong. CM: Chan. KP: Chau. YK: Thomas. GN: Ou. CO: Yang. L: Peiris. JSM: Lam. TH: Hedlev. AJ.
(2010). Air pollutants and health outcomes: Assessment of confounding by influenza. Atmos Environ 44:
1437-1442. http://dx.doi.Org/10.1016/i.atmosenv.2010.01.036
Thompsoa AM: Zanobetti. A: Silvermaa F: Schwartz. J: Coull. B: Urch. B: Speck. M: Brook. JR: Manno. M:
Gold. DR. (2010). Baseline repeated measures from controlled human exposure studies: Associations
between ambient air pollution exposure and the systemic inflammatory biomarkers IL-6 and fibrinogen.
Environ Health Perspect 118: 120-124. http://dx.doi.org/10.1289/ehp.090055Q
Thurlbeck. WM. (1982). Postnatal human lung growth. Thorax 37: 564-571. http://dx.doi.org/10.1136/thx.37.8.564
Tolbert. PE: Kleia M: Peel. JL: Sarnat. SE: Sarnat. JA. (2007). Multipollutaat modeling issues in a study of ambient
air quality and emergency department visits in Atlanta. J Expo Sci Environ Epidemiol 17: S29-S35.
http://dx.doi.org/10.1038/si.ies.750Q625
Toren. K: Brismaa J: Jarvholm. B. (1993). Asthma and asthma-like symptoms in adults assessed by questionnaires:
A literature review [Review]. Chest 104: 600-608. http://dx.doi.Org/10.1378/chest.104.2.600
Trasande. L: Wong. K: Roy. A: Savitz. DA: Thurston. G. (2013). Exploring prenatal outdoor air pollution, birth
outcomes and neonatal health care utilization in a nationally representative sample. J Expo Sci Environ
Epidemiol 23: 315-321. http://dx.doi.org/10.1038/ies.2012.124
Trenga. CA: Koenig. JO: Williams. PV. (1999). Sulphur dioxide sensitivity and plasma antioxidants in adult
subjects with asthma. Occup Environ Med 56: 544-547.
R-51
-------
Trenga. CA: Koenig. JO: Williams. PV. (2001). Dietary antioxidants and ozone-induced bronchial
hyperresponsiveness in adults with asthma. Arch Environ Occup Health 56: 242-249.
http://dx.doi.org/10.1080/000398901096Q4448
Triche. EW: Belanger. K: Brackea MB: Beckett. WS: Holford. TR: Gent. JF: Mcsharrv. JE: Leaderer. BP. (2005).
Indoor heating sources and respiratory symptoms in nonsmoking women. Epidemiology 16: 377-384.
http://dx.doi.org/10.1097/01.ede.000Q158225.44414.85
Tsai. SS: Chea PS: Yang. YH: Liou. SH: Wu. TN: Sung. FC: Yang. CY. (2012). Air pollution and hospital
admissions for myocardial infarction: Are there potentially sensitive groups? J Toxicol Environ Health A
75: 242-251. http://dx.doi.org/10.1080/15287394.2012.6412Q2
Tsai. SS: Chiu. HF: Wu. TN: Yang. CY. (2009). Air pollution and emergency room visits for cardiac arrhythmia in a
subtropical city: Taipei, Taiwan. Inhal Toxicol 21: 1113-1118.
http://dx.doi.org/10.3109/089583709Q2758939
Tseng. CY: Huang. YC: Su. SY: Huang. JY: Lai. CH: Lung. CC: Ho. CC: Liaw. YP. (2012). Cell type specificity of
female lung cancer associated with sulfur dioxide from air pollutants in Taiwan: An ecological study. BMC
Public Health 12: 4. http://dx.doi.org/10.1186/1471-2458-12-4
Tsuii. H: Larson. MG: Venditti. FJ. Jr: Manders. ES: Evans. JC: Feldmaa CL: Lew. D. (1996). Impact of reduced
heart rate variability on risk for cardiac events. The Framingham Heart Study. Circulation 94: 2850-2855.
http://dx.doi.org/10.1161/01.CIR.94.ll.2850
Tsuii. H: Venditti. FJ: Manders. ES: Evans. JC: Larsoa MG: Feldmaa. CL: Lew. D. (1994). Reduced heart rate
variability and mortality risk in an elderly cohort: The Framingham Heart Study. Circulation 90: 878-883.
http://dx.doi.Org/10.1161/01.CIR.90.2.878
Tsuiino. I: Kawakami. Y: Kaneko. A. (2005). Comparative simulation of gas transport in airway models of rat, dog,
and human. Inhal Toxicol 17: 475-485. http://dx.doi.org/10.1080/0895837059Q964476
Tunnicliffe. WS: Harrison. RM: Kelly. FJ: Punster. C: Avres. JG. (2003). The effect of sulphurous air pollutant
exposures on symptoms, lung function, exhaled nitric oxide, and nasal epithelial lining fluid antioxidant
concentrations in normal and asthmatic adults. Occup Environ Med 60: 1-7.
http://dx.doi.org/10.1136/oem.60.1 l.e!5
Tunnicliffe. WS: Hilton. MF: Harrison. RM: Avres. JG. (2001). The effect of sulphur dioxide exposure on indices of
heart rate variability in normal and asthmatic adults. Eur Respir J 17: 604-608.
Turia TC: Kita. Y: Rumana. N: Nakamura. Y: Ueda. K: Takashima. N: Sugihara. H: Morita. Y: Ichikawa. M:
Hirose. K: Nitta. H: Okavama. A: Miura. K: Ueshima. H. (2012). Short-term exposure to air pollution and
incidence of stroke and acute myocardial infarction in a Japanese population. Neuroepidemiology 38: 84-
92. http://dx.doi.org/10.1159/000335654
Turner. B. (1964). A diffusion model for an urban area. J ApplMeteorol 3: 83-91. http://dx.doi.org/10.1175/152Q-
0450(1964)003<0083 :ADMFAU>2.0.CQ:2
Turner. DB. (1970). Workbook of atmospheric dispersion estimates. (999-AP-26). Cincinnati, OH: U.S. Department
of Health, Education, and Welfare, Public Health Service.
U.S. Census Bureau. (2011). 2011 American community survey [Database]. Washington, DC: U.S. Department of
Commerce, U.S. Census Bureau. Retrieved from http://www.census.gov/programs-survevs/acs/news/data-
releases.2011.html
U.S. EPA (U.S. Environmental Protection Agency). (1971). National primary and secondary ambient air quality
standards. Fed Reg 36: 8186-8201.
U.S. EPA (U.S. Environmental Protection Agency). (1982a). Air quality criteria for particulate matter and sulfur
oxides (final, 1982) [EPA Report]. (EPA 600/8-82/029a). Washington, DC: Environmental Criteria and
Assessment Office, http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=46205
U.S. EPA (U.S. Environmental Protection Agency). (1982b). Air quality criteria for particulate matter and sulfur
oxides, volume I addendum [EPA Report]. (EPA-600/8-82-029a). Research Triangle Park. NC:
Environmental Criteria and Assessment Office.
R-52
-------
U.S. EPA (U.S. Environmental Protection Agency). (1986). Second addendum to air quality criteria for particulate
matter and sulfur oxides (1982): Assessment of newly available health effects information [EPA Report].
(EPA/600/8-86/020F). Research Triangle Park, NC: Environmental Criteria and Assessment Office.
http://nepis.epa. gov/exe/ZvPURL.cgi?Dockev=30001FM5.txt
U.S. EPA (U.S. Environmental Protection Agency). (1991). Guidelines for developmental toxicity risk assessment
(pp. 1-71). (EPA/600/FR-91/001). Washington, DC: U.S. Environmental Protection Agency, Risk
Assessment Forum, http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=23162
U.S. EPA (U.S. Environmental Protection Agency). (1992). Protocol For Determining The Best Performing Model.
(454R92025). http://nepis.epa. gov/exe/ZvPURL.cgi?Dockev=2000DE5J.txt
U.S. EPA (U.S. Environmental Protection Agency). (1994). Supplement to the second addendum (1986) to air
quality criteria for particulate matter and sulfur oxides (1982): Assessment of new findings on sulfur
dioxide acute exposure health effects in asthmatic individuals [EPA Report]. (EPA/600/FP-93/002).
Research Triangle Park, NC: Environmental Criteria and Assessment Office.
http://nepis.epa.gov/Exe/ZvPURL.cgi?Dockev=30002382.txt
U.S. EPA (U.S. Environmental Protection Agency). (1996a). Guidelines for reproductive toxicity risk assessment
(pp. 1-143). (EPA/630/R-96/009). Washington, DC: U.S. Environmental Protection Agency, Risk
Assessment Forum.
U.S. EPA (U.S. Environmental Protection Agency). (1996b). National ambient air quality standards for sulfur
oxides (sulfur dioxide)-final decision. Fed Reg 61: 25566-25580.
U.S. EPA (U.S. Environmental Protection Agency). (1998). Guidelines for neurotoxicity risk assessment [EPA
Report] (pp. 1-89). (EPA/630/R-95/001F). Washington, DC: U.S. Environmental Protection Agency, Risk
Assessment Forum, http://www.epa.gov/risk/guidelines-neurotoxicitv-risk-assessment
U.S. EPA (U.S. Environmental Protection Agency). (2005a). Guidelines for carcinogen risk assessment [EPA
Report] (pp. 1-166). (EPA/630/P-03/001F). Washington, DC: U.S. Environmental Protection Agency, Risk
Assessment Forum, http://www2.epa.gov/osa/guidelines-carcinogen-risk-assessment
U.S. EPA (U.S. Environmental Protection Agency). (2005b). Revision to the guildline on air quality models:
adoption of a preferred general purpose (flat and complext terrain) dispersion model and other revisions;
Final rule (pp. 216). (40 CFR Part 51). http://www.epa.gov/ttn/scram/guidance/guide/appw 05.pdf
U.S. EPA (U.S. Environmental Protection Agency). (2006). Estimating contributions of outdoor fine particles to
indoor concentrations and personal exposures: Effects of household characteristics and personal activities
[EPA Report]. (EPA/600/R-06/023). Research Triangle Park, NC: National Exposures Research
Laboratory.
https://nepis.epa.gOv/Exe/ZyNET.exe/9100CCPM.TXT?ZyActionD=ZyDocument&Client=EPA&Index=2
006+Thru+2010&Docs=&Query=&Time=&EndTime=&SearchMethod=l&TocRestrict=n&Toc=&TocEnt
ry=&QField=&QFieldYear=&QFieldMonth=&QFieldDay=&IntQFieldOp=0&ExtQFieldOp=0&XmlQuer
y=&File=D%3A%5Czyfiles%5CIndex%20Data%5C06thrul0%5CTxt%5C00000017%5C9100CCPM.txt
&User=ANONYMOUS&Password=anonymous&SortMethod=h%7C-
&MaximumDocuments=l&FuzzyDegree=0&ImageQuality=r75g8/r75g8/xl50yl50gl6/i425&Display=hpf
r&DefSeekPage=x&SearchBack=ZyActionL&Back=ZyActionS&BackDesc=Results%20page&Maximum
Pages=l&ZyEntry=l&SeekPage=x&ZyPURL
U.S. EPA (U.S. Environmental Protection Agency). (2007a). Integrated plan for review of the primary national
ambient air quality standards for sulfur oxides [EPA Report]. Washington, DC.
http://www.epa.gov/ttn/naaas/standards/so2/data/so2 review plan final 10-09-07.pdf
U.S. EPA (U.S. Environmental Protection Agency). (2007b). Sulfur dioxide health assessment plan: Scope and
methods for exposure and risk assessment draft [EPA Report]. Washington, DC: Office of Air Quality
Planning and Standards.
http://www.epa.gov/ttn/naaas/standards/so2/data/20071113 healthassessmentplan.pdf
R-53
-------
U.S. EPA (U.S. Environmental Protection Agency). (2008a). Integrated science assessment for oxides of nitrogen -
Health criteria: Annexes [EPA Report]. (EPA/600/R-08/072). Research Triangle Park, NC: U.S.
Environmental Protection Agency, National Center for Environmental Assessment.
http://cfpub.epa. gov/ncea/cfm/recordisplav.cfm?deid= 194645
U.S. EPA (U.S. Environmental Protection Agency). (2008b). Integrated science assessment for oxides of nitrogen
and sulfur: Ecological criteria [EPA Report]. (EPA/600/R-08/082F). Research Triangle Park, NC: U.S.
Environmental Protection Agency, Office of Research and Development, National Center for
Environmental Assessment- RTP Division, http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=201485
U.S. EPA (U.S. Environmental Protection Agency). (2008c). Integrated science assessment for oxides of nitrogen
Health criteria (Final report, 2008) [EPA Report]. (EPA/600/R-08/071). Research Triangle Park, NC: U.S.
Environmental Protection Agency, Office of Research and Development, National Center for
Environmental Assessment- RTP Division, http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid= 194645
U.S. EPA (U.S. Environmental Protection Agency). (2008d). Integrated science assessment for sulfur oxides: Health
criteria [EPA Report]. (EPA/600/R-08/047F). Research Triangle Park, NC: U.S. Environmental Protection
Agency, Office of Research and Development, National Center for Environmental Assessment- RTP.
http://cfpub.epa. gov/ncea/cfm/recordisplav.cfm?deid= 198843
U.S. EPA (U.S. Environmental Protection Agency). (2009a). Integrated science assessment for particulate matter
[EPA Report]. (EPA/600/R-08/139F). Research Triangle Park, NC: U.S. Environmental Protection
Agency, Office of Research and Development, National Center for Environmental Assessment- RTP
Division. http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=216546
U.S. EPA (U.S. Environmental Protection Agency). (2009b). Metabolically derived human ventilation rates: A
revised approach based upon oxygen consumption rates (final report) [EPA Report]. (EPA/600/R-06/129F).
Washington, DC. http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=202543
U.S. EPA (U.S. Environmental Protection Agency). (2009c). Risk and exposure assessment to support the review of
the S02 primary national ambient air quality standards: Final report [EPA Report]. (EPA-452/R-09-007).
Washington, DC: Office of Air Quality Planning and Standards.
http://www.epa.gov/ttn/naaas/standards/so2/data/200908S02REAFinalReport.pdf
U.S. EPA (U.S. Environmental Protection Agency). (2009d). Risk assessment guidance for superfund volume I:
Human health evaluation manual (Part F, supplemental guidance for inhalation risk assessment): Final
[EPA Report]. (EPA/540/-R-070/002). Washington, DC. https://www.epa.gov/risk/risk-assessment-
guidance-superfund-rags-part-f
U.S. EPA (U.S. Environmental Protection Agency). (2009e). S02 network review and background.
http://www.re gulations.gov/#!documentDetail:D=EPA-HO-OAR-2007-0352-0037
U.S. EPA (U.S. Environmental Protection Agency). (2010a). Dispersion modeling. Available online at
http://www.epa.gov/ttn/scram/dispersionindex.htm
U.S. EPA (U.S. Environmental Protection Agency). (2010b). Integrated science assessment for carbon monoxide
[EPA Report] (pp. 1-593). (EPA/600/R-09/019F). Research Triangle Park, NC: U.S. Environmental
Protection Agency, Office of Research and Development, National Center for Environmental Assessment.
http://cfpub.epa. gov/ncea/cfm/recordisplav.cfm?deid=218686
U.S. EPA (U.S. Environmental Protection Agency). (2011). Exposure factors handbook: 2011 edition (final) [EPA
Report]. (EPA/600/R-090/052F). Washington, DC: U.S. Environmental Protection Agency, Office of
Research and Development, National Center for Environmental Assessment.
http://cfpub.epa. gov/ncea/cfm/recordisplav.cfm?deid=236252
U.S. EPA (U.S. Environmental Protection Agency). (2012a). 10th modeling conference. The 10th Conference on
Air Quality Modeling, March 13-15, 2012, Research Triangle Park, NC.
U.S. EPA (U.S. Environmental Protection Agency). (2012b). Air trends: Sulfur dioxide. Available online at
https://www.epa.gov/air-trends/sulIur-dioxide-trends
R-54
-------
U.S. EPA (U.S. Environmental Protection Agency). (2012c). Total risk integrated methodology (TRIM) air
pollutants exposure model documentation (TRIM.Expo / APEX, Version 4) Volume 1: User's Guide.
(EPA-452/B-12-001a). Research Triangle Park, North Carolina: Office of Air Quality Planning and
Standards, Health and Environmental Impacts Division, htto ://www2. epa. gov/sites/production/files/2013-
08/documents/apex45 usersguide voll aug2012 O.pdf
U.S. EPA (U.S. Environmental Protection Agency). (2013a). Integrated science assessment for lead [EPA Report].
(EPA/600/R-10/075F). Research Triangle Park, NC: U.S. Environmental Protection Agency, National
Center for Environmental Assessment. http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=255721
U.S. EPA (U.S. Environmental Protection Agency). (2013b). Integrated science assessment for ozone and related
photochemical oxidants [EPA Report]. (EPA/600/R-10/076F). Research Triangle Park, NC: U.S.
Environmental Protection Agency, Office of Research and Development, National Center for
Environmental Assessment-RTP Division. http://cfpub.epa.gov/ncea/isa/recordisplav.cfm?deid=247492
U.S. EPA (U.S. Environmental Protection Agency). (2013c). Notice of workshop and call for information on
integrated science assessment for oxides of nitrogen and oxides of sulfur. Fed Reg 78: 53452-53454.
U.S. EPA (U.S. Environmental Protection Agency). (2013d). Toxicological review of trimethylbenzenes (CASRN
25551-13-7, 95-63-6, 526-73-8, and 108-67-8) in support of summary information on the Integrated Risk
Information System (IRIS): revised external review draft [EPA Report]. (EPA/635/R13/171a). Washington,
D.C.: U.S. Environmental Protection Agency, National Center for Environmental Assessment.
http://vosemite.epa.gov/sab/SABPRODUCT.NSF/b5d8alce9b07293485257375007Q12b7/eele280e77586d
e985257b65005d37e7!QpenDocument
U.S. EPA (U.S. Environmental Protection Agency). (2014a). Health Risk and Exposure Assessment for Ozone Final
Report. (452/R-14/004A). Research Triangle Park, NC: Office of Air and Radiation.
http://nepis.epa.gov/exe/ZvPURL.cgi?Dockev=P100KBUF.txt
U.S. EPA (U.S. Environmental Protection Agency). (2014b). Integrated review plan for the primary national
ambient air quality standard for sulfur dioxide [EPA Report]. (EPA-452/R-14-007). Research Triangle
Park, NC: U.S. Environmental Protection Agency, National Center for Environmental Assessment.
http://www.epa.gov/ttn/naaas/standards/so2/data/20141Q28so2reviewplan.pdf
U.S. EPA (U.S. Environmental Protection Agency). (2014c). Integrated review plan for the primary national
ambient air quality standards for nitrogen dioxide [EPA Report]. (EPA-452/R-14/003). Research Triangle
Park, NC: U.S. Environmental Protection Agency, National Center for Environmental Assessment.
http://www.epa.gov/ttn/naaas/standards/nox/data/2014Q6finalirpprimarvno2.pdf
U.S. EPA (U.S. Environmental Protection Agency). (2015a). Evaluation of Prognostic Meteorological Data in
AERMOD Applications. (454R15004). http://nepis.epa.gov/exe/ZvPURL.cgi?Dockev=P 100QIH4.txt
U.S. EPA (U.S. Environmental Protection Agency). (2015b). Preamble to the integrated science assessments [EPA
Report]. (EPA/600/R-15/067). Research Triangle Park, NC: U.S. Environmental Protection Agency, Office
of Research and Development, National Center for Environmental Assessment, RTP Division.
https://cfpub.epa. gov/ncea/isa/recordisplay.cfm?deid=310244
U.S. EPA (U.S. Environmental Protection Agency). (2015c). Table 5S-2. Study specific details of experimental
studies of S02 exposure and other morbidity effects (i.e., hematological and nervous system effects).
U.S. EPA (U.S. Environmental Protection Agency). (2016a). 11th modeling conference presentations. Available
online at https://www3.epa.gov/ttn/scram/llthmodconfpres.htm
U.S. EPA (U.S. Environmental Protection Agency). (2016b). 2014 National Emissions Inventory (NEI) data.
Washington, DC. Retrieved from https://www.epa.gov/air-emissions-inventories/2014-national-emissions-
inventorv-nei-data
U.S. EPA (U.S. Environmental Protection Agency). (2016c). Integrated review plan for the national ambient air
quality standards for particulate matter [EPA Report]. (EPA-452/R-16-005). Research Triangle Park, NC.
https://yosemite.epa. gov/sab/sabproduct.nsf/0/eb862b233fbd0cde85257dda004fcb8c!OpenDocument&Tabl
eRow=2.0
R-55
-------
U.S. EPA (U.S. Environmental Protection Agency). (2016d). Integrated science assessment for oxides of nitrogen-
Health Criteria (final report) [EPA Report]. (EPA/600/R-15/068). Research Triangle Park, NC: U.S.
Environmental Protection Agency, Office of Research and Development, National Center for
Environmental Assessment, http://ofmpub.epa.gov/eims/eimscomm.getfile7p download id=526855
U.S. EPA (U.S. Environmental Protection Agency). (2016e). Memo: List of designated reference and equivalence
methods. Available online at https://www3.epa.gov/ttnamtil/files/ambient/criteria/reference-equivalent-
methods-listpdf
U.S. EPA (U.S. Environmental Protection Agency). (2016f). Table 5S-1. Summary of epidemiologic studies of S02
exposure and other morbidity effects (i.e., sensory, nervous and gastrointestinal and other effects).
U.S. EPA (U.S. Environmental Protection Agency). (2017a). Integrated review plan for the secondary national
ambient air quality standards for ecological effects of oxides of nitrogen, oxides of sulfur and particulate
matter [EPA Report]. (EPA-452/R-17-002). Research Triangle Park, NC.
U.S. EPA (U.S. Environmental Protection Agency). (2017b). Section 3.4.3.1 references. Available online
U.S. EPA (U.S. Environmental Protection Agency). (2017c). Supplemental material for chapter 5 of the integrated
science assessment for sulfur oxides [EPA Report]. Washington, DC.
Ulirsch. GV: Ball. LM: Kave. W: Shy. CM: Lee. CV: Crawford-Brown. D: Svmons. M: Hollowav. T. (2007). Effect
of particulate matter air pollution on hospital admissions and medical visits for lung and heart disease in
two southeast Idaho cities. J Expo Sci Environ Epidemiol 17: 478-487.
http://dx.doi.org/10.1038/si.ies.7500542
University of Michigan. (2016). The panel study of income dynamics. Available online at
http://psidonline.isr.umich.edu/Studies.aspx
USGS (U.S. Geological Survey). (1999). Potentially active volcanoes of Western United States. Available online at
http://www.washingtonstatesearch.com/Washington maps/Potentially Active Volcanoes Western United
States map.html
van Aalst. R: Edwards. L: Pulles. T: De Saeger. E: Tombrou. M: Tonnesen. D. (1998). Guidance report on
preliminary assessment under EC air quality directives: 5. Modelling (pp. 34-45). (Technical report no. 11).
Copenhagen, Denmark: European Environment Agency.
http://www.eea.europa.eu/publications/TECl la/pageOl 1.html
van Thriel. C: Schaper. M: Kleinbeck. S: Kiesswetter. E: Blaszkewicz. M: Golka. K: Nies. E: Raulf-Heimsoth. M:
Briining. T. (2010). Sensory and pulmonary effects of acute exposure to sulfur dioxide (S02). Toxicol Lett
196: 42-50. http://dx.doi.Org/10.1016/i.toxlet.2010.03.013
Vandeviivere. S: Temme. E: Andielkovic. M: De Wil. M: Vinkx. C: Goevens. L: Van Loco. J. (2010). Estimate of
intake of sulfites in the Belgian adult population. Food Addit Contam Part A Chem Anal Control Expo
Risk Assess 27: 1072-1083. http://dx.doi.org/10.1080/194400410037545Q6
Vanos. JK: Cakmak. S: Bristow. C: Brion. V: Tremblav. N: Martin. SL: Sheridaa SS. (2013). Synoptic weather
typing applied to air pollution mortality among the elderly in 10 Canadian cities. Environ Res 126: 66-75.
http://dx.doi.Org/10.1016/i.envres.2013.08.003
Velicka. H: Puklova. V: Keder. J: Brabec. M: Malv. M: Bobak. M: Kotlik. B: Jink. V: Janout. V: Kazmarova. H.
(2015). Asthma exacerbations and symptom variability in children due to short-term ambient air pollution
changes in Ostrava, Czech Republic. Cent Eur J Public Health 23: 292-298.
http://dx.doi.org/10.21101/ceiph.a4548
Venkatram. A. (2002). Accounting for averaging time in air pollution modeling. Atmos Environ 36: 2165-2170.
http://dx.doi.Org/10.1016/S 1352-2310(02)00200-5
Venkatram A: Brode. R: Cimorelli. A: Lee. R: Paine. R: Perry. S: Peters. W: Weil. J: Wilson. R. (2001). A complex
terrain dispersion model for regulatory applications. Atmos Environ 35: 4211-4221.
http://dx.doi.Org/10.1016/S 1352-2310(01)00186-8
Vig. PS: Zaiac. DJ. (1993). Age and gender effects on nasal respiratory function in normal subjects. Cleft Palate
Craniofac J 30: 279-284. http://dx.doi.org/10.1597/1545-1569(1993)030<0279:AAGEON>2.3.CQ:2
R-56
-------
Villeneuve. PJ: Chea L: Rowe. BH: Coates. F. (2007). Outdoor air pollution and emergency department visits for
asthma among children and adults: A case-crossover study in northern Alberta, Canada. Environ Health 6:
40. http://dx.doi.org/10.1186/1476-069X-6-4Q
Villeneuve. PJ: Chea L: Stieb. D: Rowe. BH. (2006a). Associations between outdoor air pollution and emergency
department visits for stroke in Edmonton, Canada. Eur J Epidemiol 21: 689-700.
http://dx.doi.org/10.1007/sl0654-006-905Q-9
Villeneuve. PJ: Doiroa MS: Stieb. D: Dales. R: Burnett. RT: Dugandzic. R. (2006b). Is outdoor air pollution
associated with physician visits for allergic rhinitis among the elderly in Toronto, Canada? Allergy 61: 750-
758. http://dx.doi.org/10.1111/i. 1398-9995.2006.01070.X
Villeneuve. PJ: Johnsoa JY: Pasichnvk. D: Lowes. J: Kirkland. S: Rowe. BH. (2012). Short-term effects of ambient
air pollution on stroke: Who is most vulnerable? Sci Total Environ 430: 193-201.
http://dx.doi.Org/10.1016/i.scitotenv.2012.05.002
Vincent. GK: Velkoff. VA. (2010). The next four decades: The older population in the United States: 2010 to 2050.
(P25-1138). Washington, DC: U.S. Department of Commerce, Economics and Statistics Administration,
U.S. Census Bureau, http://www.census.gov/prod/2010pubs/p25-1138.pdf
Vinikoor-Imler. LC: Owens. EO: Nichols. JL: Ross. M: Brown. JS: Sacks. JD. (2014). Evaluating potential
response-modifying factors for associations between ozone and health outcomes: A weight-of-evidence
approach [Review]. Environ Health Perspect 122: 1166-1176. http://dx.doi.org/10.1289/ehp.1307541
von Elm. E: Altmaa DG: Egger. M: Pocock. SJ: Gotzsche. PC: Vandenbroucke. JP. (2007). The Strengthening the
Reporting of Observational Studies in Epidemiology (STROBE) statement: Guidelines for reporting
observational studies [Review]. PLoS Med 4: e296. http://dx.doi.org/10.1371/iournal.pmed.0040296
Wagner. U: Staats. P: Fehmana HC: Fischer. A: Welte. T: Groneberg. DA. (2006). Analysis of airway secretions in
a model of sulfur dioxide induced chronic obstructive pulmonary disease (COPD). J Occup Med Toxicol 1:
12. http://dx.doi.org/10.1186/1745-6673-l-12
Wallace. ME: Grantz. KL: Liu. D: Zhu. Y: Kim. SS: Mendola. P. (2016). Exposure to ambient air pollution and
premature rupture of membranes. Am J Epidemiol 183: 1114-1121. http://dx.doi.org/10.1093/aie/kwv284
Wang. AL: Blackford. TL: Lee. LY. (1996). Vagal bronchopulmonary C-fibers and acute ventilatory response to
inhaled irritants. Respir Physiol Neurobiol 104: 231-239. http://dx.doi.org/10.1016/0034-5687(96)00014-X
Wang. C: Corbett. JJ: Firestone. J. (2007). Modeling energy use and emissions from North American shipping:
application of the ship traffic, energy, and environment model. Environ Sci Technol 41: 3226-3232.
http://dx.doi.org/10.1021/es060752e
Wang. L: Wei. B: Li. Y: Li. H: Zhang. F: Rosenberg. M: Yang. L: Huang. J: Krafft. T: Wang. W. (2014a). A study
of air pollutants influencing life expectancy and longevity from spatial perspective in China. Sci Total
Environ487: 57-64. http://dx.doi.Org/10.1016/i.scitotenv.2014.03.142
Wang. T: Hendrick. F: Wang. P: Tang. G: Clemer. K: Yu. H: Favt. C: Hermans. C: Gielea C: Pinardi. G: Thevs. N:
Brenot. H: Van Roozendael. M. (2014b). Evaluation of tropospheric S02 retrieved from MAX-DO AS
measurements in Xianghe, China. Atmos Chem Phys Discuss 14: 11149-11164.
http://dx.doi.org/10.5194/acpd-14-6501-2014
Wang. XY: Hu. W: Tong. S. (2009). Long-term exposure to gaseous air pollutants and cardio-respiratory mortality
in Brisbane, Australia. Geospat Health 3: 257-263.
Wannberg. VE: Williams. G: Sawyer. P: Venedam R. (2010). An experimental field dataset with buoyant, neutral,
and dense gas atmospheric releases and model comparisons in lowwind speed (Diffusion) conditions. J
Appl Meteor Climatol 49: 1805-1817. http://dx.doi.org/10.1175/2010JAMC2383.1
Ware. JH: Ferris. BG. Jr: Dockerv. DW: Spengler. JD: Stram. DO: Speizer. FE. (1986). Effects of ambient sulfur
oxides and suspended particles on respiratory health of preadolescent children. Am Rev Respir Dis 133:
834-842.
R-57
-------
Weakley. J: Webber. MP: Ye. F: Zeig-Owens. R: Cohea HW: Hall. C'B: Kelly. K: Prezant. DJ. (2013). Agreement
between obstructive airways disease diagnoses from self-report questionnaires and medical records. Prev
Med 57: 38-42. http://dx.doi.Org/10.1016/i.ypmed.2013.04.001
Wei. Y: Davis. J: Bina. WF. (2012). Ambient air pollution is associated with the increased incidence of breast
cancer in US. Int J Environ Health Res 22: 12-21. http://dx.doi.org/10.1080/096Q3123.2011.588321
Weil. JC. (1992). Updating the ISC model through AERMIC. 85th Annual Meeting of Air and Waste Management
Association, Kansas City, MO.
Weil. JC: Corio. LA: Brower. RP. (1997). A PDF dispersion model for buoyant plumes in the convective boundary
layer. J Appl Meteorol 36: 982-1003. http://dx.doi.org/10.1175/1520-
0450(1997)036<0982:APDMFB>2.0.CQ:2
Wellenius. GA: Batesoa TF: Mittlemaa MA: Schwartz. J. (2005a). Particulate air pollution and the rate of
hospitalization for congestive heart failure among medicare beneficiaries in Pittsburgh, Pennsylvania. Am J
Epidemiol 161: 1030-1036. http://dx.doi.org/10.1093/aie/kwil35
Wellenius. GA: Schwartz. J: Mittleman. MA. (2005b). Air pollution and hospital admissions for ischemic and
hemorrhagic stroke among medicare beneficiaries. Stroke 36: 2549-2553.
http://dx.doi.org/10.1161/01.STR.0000189687.7876Q.47
Wellenius. GA: Yeh. GY: Coull. BA: Suh. HH: Phillips. RS: Mittlemaa MA. (2007). Effects of ambient air
pollution on functional status in patients with chronic congestive heart failure: A repeated-measures study.
Environ Health 6. http://dx.doi.org/10.1186/1476-069X-6-26
Welz. O: Savee. JD: Osbora PL: Vasu. SS: Percival. CJ: Shallcross. DE: Taaties. CA. (2012). Direct kinetic
measurements of Criegee intermediate (CHOO) formed by reaction of CHI with O. Science 335: 204-207.
http://dx.doi.org/10.1126/science.1213229
Weng. CH: Hu. CC: Yen. TH: Huang. WH. (2015). Association between environmental particulate matter and
arterial stiffness inpatients undergoing hemodialysis. BMC Cardiovasc Disord 15: 115.
http://dx.doi.org/10.1186/sl2872-015-01Q7-0
Wheeler. AJ: Smith-Doiroa M: Xu. X: Gilbert. NL: Brook. JR. (2008). Intra-urban variability of air pollution in
Windsor, Ontario - Measurement and modeling for human exposure assessment. Environ Res 106: 7-16.
http://dx.doi.Org/10.1016/i.envres.2007.09.004
WHO (World Health Organization). (1948). Preamble to the Constitution of the World Health Organization as
adopted by the International Health Conference, New York, 19-22 June, 1946. In Constitution of the World
Health Organization (pp. 2). Geneva, Switzerland.
http://whalibdoc.who.int/hist/official records/constitutioapdf
WHO (World Health Organization). (2006). WHO air quality guidelines for particulate matter, ozone, nitrogen
dioxide and sulfur dioxide - global update 2005 - summary of risk assessment. Copenhagen, Denmark.
http://www.who.int/phe/health topics/outdoorair/outdoorair aqg/en/
Widdicombe. J: Lee. LY. (2001). Airway reflexes, autonomic function, and cardiovascular responses [Review].
Environ Health Perspect 4: 579-584.
Widdicombe. JG. (2003). Overview of neural pathways in allergy and asthma [Review]. Pulm Pharmacol Ther 16:
23-30. http://dx.doi.org/10.1016/S1094-5539(02)00178-5
Wiedinmver. C: Ouavle. B: Geroa C: Belote. A: Mckenzie. D. on: Zhang. X: O'Neill. S: Wynne. KK. (2006).
Estimating emissions from fires in North America for air quality modeling. Atmos Environ 40: 3419-3432.
http://dx.doi.Org/10.1016/i.atmosenv.2006.02.010
Williams. M: Barrowcliffe. R: Laxea D: Monks. P. (2011). Review of air quality modelling in Defra. London,
England: King's College, School of Biomedical and Health Sciences. http://uk-
air.defra.gov.uk/reports/cat20/1106290858 DefraModellingReviewFinalReport.pdf
R-58
-------
Williams. R: Rea. A: Vette. A: Croghaa C: Whitaker. D: Stevens. C: Mcdow. S: Fortmann. R: Sheldon. L: Wilsoa
H: Thornburg. J: Phillips. M: Lawless. P: Rodes. C: Daughtrev. H. (2009). The design and field
implementation of the Detroit exposure and aerosol research study. J Expo Sci Environ Epidemiol 19: 643-
659. http://dx.doi.org/10.1038/ies.20Q8.61
Wilson. AM: Wake. CP: Kelly. T: Sallowav. JC. (2005). Air pollution, weather, and respiratory emergency room
visits in two northern New England cities: an ecological time-series study. Environ Res 97: 312-321.
http://dx.doi.Org/10.1016/i.envres.2004.07.010
Wilson. WE: Mage. DT: Grant. LP. (2000). Estimating separately personal exposure to ambient and nonambient
particulate matter for epidemiology and risk assessment: Why and how. J Air Waste Manag Assoc 50:
1167-1183. http://dx.doi.org/10.1080/10473289.200Q.10464164
Wilson. WE: Suh. HH. (1997). Fine particles and coarse particles: Concentration relationships relevant to
epidemiologic studies. J Air Waste Manag Assoc 47: 1238-1249.
http://dx.doi.org/10.1080/10473289.1997.10464Q74
Winquist. A: Kirrane. E: Kleia M: Strickland. M: Darrow. LA: Sarnat. SE: Gass. K: Mulholland. J: Russell. A:
Tolbert. P. (2014). Joint effects of ambient air pollutants on pediatric asthma emergency department visits
in Atlanta, 1998-2004. Epidemiology 25: 666-673. http://dx.doi.org/10.1097/EDE.000000000000Q146
Wintertoa PL: Kaufman. J: Keener. CV: Ouiglev. S: Farin. FM: Williams. PV: Koenig. JO. (2001). Genetic
polymorphisms as biomarkers of sensitivity to inhaled sulfur dioxide in subjects with asthma. Ann Allergy
Asthma Immunol 86: 232-238. http://dx.doi.org/10.1016/S1081-1206(10)62697-X
Wiwatanadate. P: Liwsrisakun. C. (2011). Acute effects of air pollution on peak expiratory flow rates and symptoms
among asthmatic patients in Chiang Mai, Thailand. Int J Hyg Environ Health 214: 251-257.
http://dx.doi.Org/10.1016/i.iiheh.2011.03.003
Wiwatanadate. P: Trakultivakora M. (2010). Air pollution-related peak expiratory flow rates among asthmatic
children in Chiang Mai, Thailand. Inhal Toxicol 22: 301-308.
http://dx.doi.org/10.3109/0895837090330Q327
Woermaa AL: Mendelowitz. D. (2013a). Perinatal sulfur dioxide exposure alters brainstem parasympathetic control
of heart rate. Cardiovasc Res 99: 16-23. http://dx.doi.org/10.1093/cvr/cvt057
Woermaa AL: Mendelowitz. D. (2013b). Postnatal sulfur dioxide exposure reversibly alters parasympathetic
regulation of heart rate. Hypertension 62: 274-280.
http://dx.doi.org/10.1161/HYPERTENSIQNAHA.113.01552
Wollmana HA. (1998). Intrauterine growth restriction: Definition and etiology. Horm Res 49: 1-6.
http://dx.doi.org/10.1159/000053Q79
Wong. CM: Lam. TH: Peters. J: Hedlev. AJ: Ong. SG: Tam AYC: Liu. J: Spiegelhalter. DJ. (1998). Comparison
between two districts of the effects of an air pollution intervention on bronchial responsiveness in primary
school children in Hong Kong. J Epidemiol Community Health 52: 571-578.
http://dx.doi.Org/10.1136/iech.52.9.571
Wong. CM: Ou. CO: Chan. KP: Chau. YK: Thach. TO: Yang. L: Chung. RY: Thomas. GN: Peiris. JS: Wong. TW:
Hedlev. AJ: Lam. TH. (2008a). The effects of air pollution on mortality in socially deprived urban areas in
Hong Kong, China. Environ Health Perspect 116: 1189-1194. http://dx.doi.org/10.1289/ehp. 10850
Wong. CM: Vichit-Vadakan. N: Kaa H: Oiaa Z. (2008b). Public Health and Air Pollution in Asia (PAPA): A
multicity study of short-term effects of air pollution on mortality. Environ Health Perspect 116: 1195-1202.
http://dx.doi.org/10.1289/ehp. 11257
R-59
-------
Wong. CM: Vichit-Vadakan. N: Vaianapoom. N: Ostro. B: Thach. TO: Chau. PY: Chaa EK: Chung. RY: Ou. CO:
Yang. L: Peiris. JS: Thomas. GN: Lam. TH: Wong. TW: Hedlev. AJ: Kan. H: Chea B: Zhao. N: London.
SJ: Song. G: Chea G: Zhang. Y: Jiang. L: Oian. Z: He. O: Lia HM: Kong. L: Zhou. D: Liang. S: Zhu. Z:
Liao. D: Liu. W: Bentlev. CM: Daa J: Wang. B: Yang. N: Xu. S: Gong. J: Wei. H: Sua H: Oin. Z. (2010).
Part 5. Public health and air pollution in Asia (PAPA): A combined analysis of four studies of air pollution
and mortality. In Public Health and Air Pollution in Asia (PAPA): Coordinated studies of short-term
exposure to air pollution and daily mortality in four cities (pp. 377-418). (ISSN 1041-5505). Boston, MA:
Health Effects Institute. https://www.healtheffects.org/sYStem/files/PAPA 154 Part5 Combined.pdf
Wong. CM: Yang. L: Thach. TO: Chau. PY: Chan. KP: Thomas. GN: Lam. TH: Wong. TW: Hedlev. AJ: Peiris. JS.
(2009). Modification by influenza on health effects of air pollution in Hong Kong. Environ Health Perspect
117: 248-253. http://dx.doi.org/10.1289/ehp. 11605
Wong. DC: Pleim. J: Mathur. R: Binkowski. F: Otte. T: Gilliam. R: Pouliot. G: Xiu. A: Young. JO: Kang. D.
(2012). WRF-CMAQ two-way coupled system with aerosol feedback: Software development and
preliminary results. Geosci Model Dev 5: 299-312. http://dx.doi.org/10.5194/gmd-5-299-2012
Wong. TW: Lau. TS: Yu. TS: Neller. A: Wong. SL: Tam. W: Pang. SW. (1999). Air pollution and hospital
admissions for respiratory and cardiovascular diseases in Hong Kong. Occup Environ Med 56: 679-683.
http://dx.doi.org/10.1136/oem.56.10.679
Woodruff. TJ: Darrow. LA: Parker. JD. (2008). Air pollution and postneonatal infant mortality in the United States,
1999-2002. Environ Health Perspect 116: 110-115. http://dx.doi.org/10.1289/ehp. 10370
Woodruff. TJ: Parker. JD: Darrow. LA: Slama. R: Bell. ML: Choi. H: Glinianaia. S: Hoggatt. KJ: Karr. CJ: Lobdell.
DT: Wilhelm. M. (2009). Methodological issues in studies of air pollution and reproductive health. Environ
Res 109: 311-320. http://dx.doi.Org/10.1016/i.envres.2008.12.012
Wu. J: Jiang. C: Houstoa D: Baker. D: Delfino. R. (201 la). Automated time activity classification based on global
positioning system (GPS) tracking data. Environ Health 10: 101. http://dx.doi.org/10.1186/1476-Q69X-10-
101
Wu. J: Jiang. C: Liu. Z: Houston. D: Jaimes. G: McConnell. R. (2010). Performances of different global positioning
system devices for time-location tracking in air pollution epidemiological studies. Environ Health Insights
4: 93-108. http://dx.doi.org/10.4137/EHI.S6246
Wu. S: Ni. Y: Li. H: Pan. L: Yang. D: Baccarelli. AA: Deng. F: Chen. Y: Shima. M: Guo. X. (2016). Short-term
exposure to high ambient air pollution increases airway inflammation and respiratory symptoms in chronic
obstructive pulmonary disease patients in Beijing, China. Environ Int 94: 76-82.
http://dx.doi.Org/10.1016/i.envint.2016.05.004
Wu. X. (; Bennett, DH; Lee, K; Cassady, DL; Ritz, B; Hertz-Picciotto, I. (2012). Feasibility of using web surveys to
collect time-activity data. J Expo Sci Environ Epidemiol 22: 116-125. http://dx.doi.org/10.1038/ies.2011.23
Wu. X: Bennett. DH: Lee. K: Cassadv. DL: Ritz. B: Hertz-Picciotto. I. (201 lb). Longitudinal variability of time-
location/activity patterns of population at different ages: A longitudinal study in California. Environ Health
10. http://dx.doi.org/10.1186/1476-069X-10-80
Xie. J: Li. R: Fan. R: Meng. Z. (2009). Effects of sulfur dioxide on expressions of p53, bax and bcl-2 in lungs of
asthmatic rats. Inhal Toxicol 21: 952-957. http://dx.doi.org/10.1080/089583708026296Q2
Xu. X: Ding. H: Wang. X. (1995). Acute effects of total suspended particles and sulfur dioxides on preterm delivery:
A community-based cohort study. Arch Environ Occup Health 50: 407-415.
http://dx.doi.org/10.1080/00Q39896.1995.9935976
Xu. X: Dockerv. DW: Wang. L. (1991). Effects of air pollution on adult pulmonary function. Arch Environ Occup
Health 46: 198-206. http://dx.doi.org/10.1080/00Q39896.1991.9937448
Xu. X: Hu. H: Ha. S: Roth. J. (2014). Ambient air pollution and hypertensive disorder of pregnancy. J Epidemiol
Community Health 68: 13-20. http://dx.doi.org/10.1136/iech-2013-202902
R-60
-------
Xue. J: McCurdv. T: Spengler. J: Ozkavnak. H. (2004). Understanding variability in time spent in selected locations
for 7-12-year old children. J Expo Anal Environ Epidemiol 14: 222-233.
http://dx.doi.org/10.1038/si.iea.7500319
Yamamoto. N: Shendell. D: Winer. A: Zhang. J. (2010). Residential air exchange rates in three major US
metropolitan areas: results from the Relationship Among Indoor, Outdoor, and Personal Air Study
19992001. Indoor Air 20: 85-90. http://dx.doi.org/10.1111/i. 1600-0668.2009.00622.X
Yang. C: Chea A: Chen. R: Oi. Y: Ye. J: Li. S: Li. W: Liang. Z: Liang. O: Guo. D: Kan. H: Chen. X. (2014a).
Acute effect of ambient air pollution on heart failure in Guangzhou, China. Int J Cardiol 177: 436-441.
http://dx.doi.Org/10.1016/i.iicard.2014.09.003
Yang. CL: To. T: Fotv. RG: Stieb. DM: Dell. SD. (2011). Verifying a questionnaire diagnosis of asthma in children
using health claims data. BMC Pulm Med 11. http://dx.doi.org/10.1186/1471-2466-11-52
Yang. CY. (2008). Air pollution and hospital admissions for congestive heart failure in a subtropical city: Taipei,
Taiwan. J Toxicol Environ Health A 71: 1085-1090. http://dx.doi.org/10.1080/152873908Q2114428
Yang. O: Chea Y: Shi. Y: Burnett. RT: McGrail. KM: Krewski. D. (2003). Association between ozone and
respiratory admissions among children and the elderly in Vancouver, Canada. Inhal Toxicol 15: 1297-1308.
http://dx.doi.org/10.1080/0895837039Q241768
Yang. WS: Wang. X: Deng. O: Faa WY: Wang. WY. (2014b). An evidence-based appraisal of global association
between air pollution and risk of stroke. Int J Cardiol 175: 307-313.
http://dx.doi.Org/10.1016/i.iicard.2014.05.044
Yao. G: Yue. H: Yua Y: Sang. N. (2015). Chronic S02 inhalation above environmental standard impairs neuronal
behavior and represses glutamate receptor gene expression and memory-related kinase activation via
neuroinflammation in rats. Environ Res 137: 85-93. http://dx.doi.Org/10.1016/i.envres.2014.ll.012
Yao. G: Yun. Y: Sang. N. (2014). Differential effects between one week and four weeks exposure to same mass of
S02 on synaptic plasticity in rat hippocampus. Environ Toxicol 31: 820-829.
http://dx.doi.org/10.1002/tox.22Q93
Yildirim. Z: Kilic. T: Koksal. N: Kotuk. M. (2005). Protective effect of ipratropium bromide on bronchoconstriction
induced by sulfur dioxide exposure during apricot sufurization processes that causes asthma-like syndrome
in agricultural environment. Pharmacol Res 51: 479-482. http://dx.doi.Org/10.1016/i.phrs.2004.12.004
Yogev-Baggio. T: Bibi. H: Dubnov. J: Or-Hea K: Carel. R: Portnov. BA. (2010). Who is affected more by air
pollution-sick or healthy? Some evidence from a health survey of schoolchildren living in the vicinity of a
coal-fired power plant in Northern Israel. Health Place 16: 399-408.
http://dx.doi.Org/10.1016/i.healthplace.2009.ll.013
Yokovama. E: Yoder. RE: Frank. NR. (1971). Distribution of 35S in the blood and its excretion in urine of dogs
exposed to 35S02. Arch Environ Health 22: 389-395. http://dx.doi.org/10.1080/00Q39896.1971.10665861
Yorifuii. T: Kashima. S: Doi. H. (2015a). Outdoor air pollution and term low birth weight in Japan. Environ Int 74:
106-111. http://dx.doi.Org/10.1016/i.envint.2014.09.003
Yorifuii. T: Kashima. S: Higa Diez. M: Kado. Y: Sanada. S: Doi. H. (2015b). Prenatal Exposure to Traffic-related
Air Pollution and Child Behavioral Development Milestone Delays in Japan. Epidemiology 27: 57-65.
http://dx.doi.org/10.1097/EDE.000000000000Q361
Yun. Y: Yao. G: Yue. H: Guo. L. in: Oia G: Li. G: Sang. N. an. (2013). S02 inhalation causes synaptic injury in rat
hippocampus via its derivatives in vivo. Chemosphere 93: 2426-2432.
http://dx.doi.Org/10.1016/i.chemosphere.2013.08.063
Zartariaa V: Bahadori. T: McKone. T. (2005). Adoption of an official ISEA glossary. J Expo Anal Environ
Epidemiol 15: 1-5.
Zeger. SL: Thomas. D: Dominici. F: Samet. JM: Schwartz. J: Dockerv. D: Cohen. A. (2000). Exposure
measurement error in time-series studies of air pollution: Concepts and consequences. Environ Health
Perspect 108: 419-426. http://dx.doi.org/10.1289/ehp.00108419
R-61
-------
Zeman. KL: Bennett. WD. (2006). Growth of the small airways and alveoli from childhood to the adult lung
measured by aerosol-derived airway morphometry. J Appl Physiol (1985) 100: 965-971.
http://dx.doi.org/10.1152/iapplphvsiol.00409.2005
Zemek. R: Szvszkowicz. M: Rowe. BH. (2010). Air pollution and emergency department visits for otitis media: js^A
case-crossover study in Edmonton, Canada. Environ Health Perspect 118: 1631-1636.
http://dx.doi.org/10.1289/ehp.0901675
Zhang. J: Zhu. T: Kipen. H: Wang. G: Huang. W: Rich. D: Zhu. P: Wang. Y: Lu. SE: Ohman-Strickland. P: Diehl.
S: Hu. M: Tong. J: Gong. J: Thomas. D. (2013). Cardiorespiratory biomarker responses in healthy young
adults to drastic air quality changes surrounding the 2008 Beijing Olympics. Res Rep Health Eff Inst 174:
5-174.
Zhang. JZ: Millero. FJ. (1991). The rate of sulfite oxidation in seawater. Geochim Cosmo Act 55: 677-685.
http://dx.doi.org/10.1016/0016-7037(91)90333-Z
Zhang. N: Yu. L: Wan. J: Wang. Y. (2015a). Influences of meteorological conditions on air pollution in peninsula-
based city. In 4th International conference on energy and environmental protection (ICEEP 2015).
Lancaster, PA: DEStech Publication, Inc. http://www.destechpub.com/product/4th-international-
conference-energy-environmental-protection-iceep-2015/
Zhang. PF: Dong. GH: Sun. BJ: Zhang. LW: Chea X: Ma. NN: Yu. F: Guo. HM: Huang. H: Lee. YL: Tang. NJ:
Chea J. (2011). Long-term exposure to ambient air pollution and mortality due to cardiovascular disease
and cerebrovascular disease in Shenyang, China. PLoS ONE 6.
http://dx.doi.org/10.1371/iournal.pone.0020827
Zhang. Y: Wang. SG: Ma. YX: Shang. KZ: Cheng. YF: Li. X: Ning. GC: Zhao. WJ: Li. NR. (2015b). Association
between ambient air pollution and hospital emergency admissions for respiratory and cardiovascular
diseases inBeijing: A time series study. Biomed Environ Sci 28: 352-363.
http://dx.doi.org/10.3967/bes2015.049
Zhao. A. ng: Chen. R: Kuang. X: Kan. H. (2014). Ambient air pollution and daily outpatient visits for cardiac
arrhythmia in Shanghai, China. J Epidemiol 24: 321-326. http://dx.doi.org/10.2188/iea.JE2014003Q
Zhao. OG: Liang. ZJ: Tao. SJ: Zhu. JA: Du. YK. (2011). Effects of air pollution on neonatal prematurity in
Guangzhou of China: A time-series study. Environ Health 10. http://dx.doi.org/10.1186/1476-Q69X-10-2
Zhao. Y: Frev. HC. (2004). Quantification of variability and uncertainty for censored data sets and application to air
toxic emission factors. Risk Anal 24: 1019-1034. http://dx.doi.Org/10.llll/i.0272-4332.2004.00504.x
Zhao. Y: Oiaa Z: Wang. J: Vaughn. MG: Liu. Y: Rea W: Dong. G. (2013). Does obesity amplify the association
between ambient air pollution and increased blood pressure and hypertension in adults? Findings from the
33 Communities Chinese Health Study [Letter], Int J Cardiol 168: E148-E150.
http://dx.doi.Org/10.1016/i.iicard.2013.08.071
Zhao. YC: Frev. HC. (2006). Uncertainty for data with non-detects: Air toxic emissions from combustion. Hum Ecol
Risk Assess 12: 1171-1191. http://dx.doi.org/10.1080/1080703060Q977178
Zhao. Z: Zhang. Z: Wang. Z: Ferm M: Liang. Y: Norback. D. (2008). Asthmatic symptoms among pupils in
relation to winter indoor and outdoor air pollution in schools in Taiyuan, China. Environ Health Perspect
116: 90-97. http://dx.doi.org/10.1289/ehp. 10576
Zheng. S: Wang. M: Wang. S: Tao. Y: Shang. K. (2013). Short-term effects of gaseous pollutants and particulate
matter on daily hospital admissions for cardio-cerebrovascular disease in Lanzhou: Evidence from a
heavily polluted city in China. Int J Environ Res Public Health 10: 462-477.
http://dx.doi.org/10.3390/iierphlQ020462
Zheng. XY: Ding. H: Jiang. LN: Chen. SW: Zheng. JP: Oiu. M: Zhou. YX: Chea O: Guan. WJ. (2015). Association
between air pollutants and asthma emergency room visits and hospital admissions in time series studies: A
systematic review and meta-analysis [Review]. PLoS ONE 10: 1-24.
http://dx.doi.org/10.1371/iournal.pone.0138146
R-62
-------
Zhou. N: Cui. Z: Yang. S: Haa X: Chen. G: Zhou. Z: Zhai. C: Ma. M: Li. L: Cai. M: Li. Y: Ao. L: Shu. W: Liu. J:
Cao. J. (2014). Air pollution and decreased semen quality: A comparative study of Chongqing urban and
rural areas. EnvironPollut 187: 145-152. http://dx.doi.Org/10.1016/i.envpol.2013.12.030
Zhu. Y: Zhang. C: Liu. D: Grantz. KL: Wallace. M: Mendola. P. (2015). Maternal ambient air pollution exposure
preconception and during early gestation and offspring congenital orofacial defects. Environ Res 140: 714-
720. http://dx.doi.Org/10.1016/i.envres.2015.06.002
Ziemana C: Hansen. T: Pohlmana G: Farrar. D: Pohlenz-Michel. C: Tillmann. T: Mangelsdorf. I. (2010).
Genotoxicity testing of sulfur dioxide (S02) in a mouse bone marrow micronucleus test complemented
with hematological endpoints. Mutat Res Genet Toxicol Environ Mutagen 697: 38-46.
http://dx.doi.Org/10.1016/i.mrgentox.2010.02.002
Zmirou. D: Schwartz. J: Saez. M: Zanobetti. A: Woitvniak. B: Touloumi. G: Spix. C: Ponce de Leoa A: Le
Moullec. Y: Bacharova. L: Schouten. J: Ponka. A: Katsouvanni. K. (1998). Time-series analysis of air
pollution and cause-specific mortality. Epidemiology 9: 495-503.
Zou. B: Wilson. JG: Zhaa FB: Zeng. Y. (2009a). Air pollution exposure assessment methods utilized in
epidemiological studies [Review]. J Environ Monit 11: 475-490. http://dx.doi.org/10.1039/b813889c
Zou. B: Wilson. JG: Zhaa FB: Zeng. Y. (2009b). An emission-weighted proximity model for air pollution exposure
assessment. Sci Total Environ 407: 4939-4945. http://dx.doi.Org/10.1016/i.scitotenv.2009.05.014
Zou. B: Wilson. JG: Zhaa FB: Zeng. YN. (2009c). Spatially differentiated and source-specific population exposure
to ambient urban air pollution. Atmos Environ 43: 3981-3988.
http://dx.doi.Org/10.1016/i.atmosenv.2009.05.022
Zou. B: Zhaa FB: Zeng. Y: Yorke. C. h: Liu. X. (2011). Performance of kriging and EWPM for relative air
pollution exposure risk assessment. International Journal of Environmental Research 5: 769-778.
R-63
-------
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT NO. G-35
United States
Environmental Protection
Agency
National Center for Environmental Assessment
Office of Research and Development
Washington, DC 20460
Official Business
Penalty for Private Use
$300
Printed with vegetable-based ink on paper that
W contains a minimum of 50% post-consumer fiber
content and processed chlorine free.
------- |