WASHINGTON D.C. 20460
                                                              OFFICE OF THE ADMINISTRATOR
                                                                SCIENCE ADVISORY BOARD

                                       January 9, 2008


Honorable Stephen L. Johnson
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, NW
Washington, DC 20460

       Subject:       Clean Air Scientific Advisory Committee's (CASAC) Peer Review of
                    EPA's Integrated Science Assessment (ISA) for Sulfur Oxides - Health
                    Criteria (First External Review Draft, September 2007)

Dear Administrator Johnson:

       The Clean Air Scientific Advisory Committee (CASAC or Committee), augmented by
subject-matter-experts to form the CASAC Sulfur Oxides Primary NAAQS Review Panel
(hereafter referred to as the panel) conducted its review of EPA's Integrated Science Assessment
(ISA) for Sulfur Oxides - Health Criteria (First External Review Draft, September 2007) on
December 5-6, 2007. The ISA for sulfur oxides (SOX) was produced for EPA's new process for
reviewing and revising National Ambient Air Quality Standards (NAAQS).  Done properly, the
ISA should be an informative, succinct, and useful summary of the evidence for consideration of
the NAAQS.

       Overall, the panel found that the first draft ISA for SOx is well developed, and with
further revisions, should address CAS AC's concerns. We recognize that an administrative
decision has been made to consider the sulfate particulate matter as part of the PM NAAQS
review process; therefore, sulfates will only be considered here to the degree that they
inform the consideration of the potential health effects of sulfur dioxide (802). Major points
from CASAC panel members are summarized below. Individual recommendations from
CASAC Panel members to strengthen the next draft are appended in Attachment B.

1. To what extent are the atmospheric chemistry and air quality characterizations clearly
communicated, appropriately characterized, and relevant to the review of the primary SOi

       Some improvements need to be made to the discussion of the atmospheric chemistry of
SOx.  For example, the low saturation vapor pressure of sulfuric acid must be mentioned as the
driving force for the gas-to-particle conversion of sulfate. Also, a more comprehensive
discussion of the oxidation of SO2 to sulfate in coal-fired power plant plumes is needed since this
source category is currently the major emission source of 862 in the United States.

       Ship emissions are a major issue near ports and along shipping lanes and as sulfur
emissions from point sources and on-land mobile sources decrease, ship emissions take more
relative importance. Accordingly, it would be useful to discuss ship emissions (and other
relevant SC>2 emission sources), including associated uncertainties.

2. Are the properties of ambient sulfur oxides appropriately characterized, including
policy-relevant background, spatial and temporal patterns, and relationships between
ambient sulfur oxides and human exposure?

       There is a need for a better description/characterization of the urban spatial heterogeneity
of SC>2 concentrations, including contrast of urban/rural concentrations and characterization of
concentrations with respect to current monitor locations (e.g., what is the average concentration
at source-oriented monitors versus at population-oriented monitors and at rural background
monitors).  Quantification of spatial heterogeneity and how this impacts our ability to assess
area-wide ambient concentration patterns, including five minute peaks, is also needed.

       The lack of correlation between SC>2 and sulfate is not surprising considering the slow
conversion of SO2 to sulfate on average.  Correlations should be presented between SO2 and
other air pollutants such as NC>2, PM2.s and PMio-2.5, and specific components of particulate
matter (e.g., elemental carbon and specific metals).

       Additional information is needed to compare the distributions of monitoring data
averaged over the different times considered for future standards (e.g., 5 minute, 15 minute, 1
hour,  1 day and 1 year). It would be useful, for example, to correlate 5 or 15 minute averages
and maxima with the 1-hour averages.

3. Is the information provided on atmospheric sciences and exposure sufficient for the
evaluation of human health effects of sulfur oxides in the ISA?

       A thorough analysis of the Air Quality System (AQS) and other data should be added to
the Annex.  A better understanding of the spatial distribution of population exposure is needed.
The AQS data characterization needs to more directly address the spatio-temporal variation in
SC>2 and the contribution of monitor location characteristics to the summary data because this has
direct bearing on the interpretation of the epidemiological studies.  There is a paucity of ambient
monitors and they may represent near-source impacts rather than average population exposure.
For example, how can we estimate population exposure in areas where there is no SC>2 monitor
or where the SC>2 monitors are located to characterize the potential impacts of SC>2 sources?
Land-use characteristics (e.g., industrial, urban, rural) may help address this issue. Relevant
features should be brought forward in Chapter 2.

       The draft ISA needs to better address how accurately we can predict personal and
population exposure using ambient monitoring.  In addition to understanding the spatial
distribution of population exposure, a better understanding of the relationships between outdoor,
indoor and personal exposure is needed; it will be a useful background for the subsequent use of
APEX, EPA's exposure assessment model, for population exposure calculations.

       A discussion of the bias and uncertainty associated with the SC>2 monitoring network is
needed because of the implications for the assessment of population exposure to SO2. In
particular, the error is likely to be large for population exposure to high 862 concentrations.

       CAS AC asks for a description of the database that exists for 5-minute peak
concentrations. A figure should be added to depict an example of the probability distribution of
5-minute average 862 concentrations.  Plots of similar form (e.g., cumulative distributions)
should be provided for annual, 24-hour average, 1-hour average and 5-minute average of SC>2
concentrations. Correlations between peak 5-minute averages, peak 1-hour averages and 24-
hour average concentrations should be provided. These correlations could facilitate
understanding of the relationships between findings of human clinical studies and
epidemiological studies.

       A deeper discussion of the physical and chemical interactions between co-pollutants with
SC>2 would be useful.  In particular, interactions of 862 with particles should be discussed.

4.     To what extent are the discussion and integration of evidence on the health effects
of sulfur oxides from the animal toxicological, human clinical, and epidemiological studies,
technically sound, appropriately balanced, and clearly communicated?

       The presentation of the results of the animal toxicological, controlled human exposure,
and epidemiological studies that have been reviewed is generally technically sound.  However,
the criteria for selection of specific studies presented in each of the three categories should  be
clearly stated. In addition, the criteria for judging the strength of findings from specific studies,
determining consistency of results, and assessment of aggregate findings of studies on relevant
research questions, should also be clearly  stated. The existence of publication bias and its
consequences, both positive and negative, should be assessed.  It should be noted that several
relevant studies have not been included in the draft; some of them are recent publications and
others did not emphasize SC>2 in their conclusions. These studies are listed in individual
comments. The assessment of the aggregate findings needs to be reviewed in light of the
approach to the evaluation of evidence of causality to be described in a revised ISA.

       The epidemiological data are relatively consistent and coherent with regard to the
association of short-term exposure to 862 and emergency department visits/hospitalizations for
asthma and all respiratory diseases, particularly among children.  There is not sufficient
discussion of the fact that a substantial fraction (15-30%) of the population may be driving  the
effects noted in non-asthmatic subjects. This requires additional discussion, since null results
across larger (non-asthmatic)  population groups may obscure the identification of actual
susceptible sub-groups within. Additionally, positive results may be driven by this susceptible

non-asthmatic sub-group, possibly resulting in over-interpretation and assignment of a general
population effect.

       Interpretation of all epidemiological studies of SO2 effects (both long-term and short-
term) needs to include the context of exposure assessment, particularly with respect to local
source monitor siting and how evenly distributed SC>2 is in space. There needs to be a
standardized protocol for summarizing relevant studies for the ISA and in selecting which
information to report.  Characterization of epidemiological study results from multi-pollutant
models (such as those involving analytical partitioning between  862, PM, and other pollutant
effects) should be re-visited to clarify the current presentation and interpretation.  Thoughtfully
interpreted discussions in the context of (a) interactive effects and effect modification; (b)
mediated effects; (c) confounding possibilities; and (d) independent effects need to be expanded
in the analysis of epidemiological studies involving two-pollutant and multi-pollutant models.

       The controlled human exposure data presented in the current version of the health effects
chapter indicate that asthmatic individuals are especially sensitive to SO2 exposure in terms of
respiratory symptoms and bronchoconstriction. While these data do provide some plausibility
for the epidemiological studies reporting associations between ambient SC>2 and emergency
department visits or total hospitalizations for asthma, they are not informative as to how SC>2
exposure might induce respiratory hospitalizations in non-asthmatic individuals.  Plausible
mechanisms for other respiratory associations should be considered as well.

       More attention needs to be given to the results from the older human clinical studies to
leverage new insights from previous studies. This approach may be informative as to who is
responsive and under what conditions (levels, length of exposure, co-exposures, severity of
health status, etc.) that responsiveness is present.  This information needs to be summarized in
the document and integrated with the epidemiological results. In addition, recently published
additional studies need to be identified and included, as appropriate, in the review.

       There are places in the text that need to be tighter, less redundant, and more thematically
organized (i.e., each section should have a story line). In particular, the summary/integration
subsections should provide an overview of the quantity and quality of the  evidence for the health
outcome(s) of interest as well as evaluation of how well the toxicological and clinical data
support the epidemiologic findings.  Although there is substantial information for a number of
different health outcomes, the focus of the integrated summary of the findings, the strength of
those findings, and confidence in the findings are not clearly presented.  A summary table listing
the various outcomes, a determination of the level of confidence in the findings, and the
implications of the findings would help guide the reader to a cumulative sense of the current
aggregate status of SC>2 knowledge.

       Much of the material in Chapters 4 and 5 might be brought forward to Chapter 3 to better
inform the discussion on susceptibility and vulnerability (two separate issues currently presented
in multiple chapters that also should be collected  and discussed in one chapter location).

       A section of the document needs to be added to summarize current levels of uncertainty
and to discuss key areas of research needed to reduce those uncertainty levels.

5.     To what extent does the integration of health evidence focus on the most policy-
relevant studies or health findings?

       The document does a commendable job of selecting out the key documents since the last
review, but EPA staff and consultants should more carefully re-evaluate whether there are any
key studies prior to the last review that provide especially important insights, or that can be
reconsidered in the light of new evidence, in order to aid in the present evaluation of the health
effects of SOx. This was especially noted as being important in the review of the toxicological
and human clinical studies done on sulfur oxides and co-pollutants.

6.     What are the views of the Panel on the conclusions drawn in the draft ISA
regarding the strength, consistency, coherence and plausibility of health effects of sulfur

       The EPA staff is to be commended for producing a summary of findings on SOx that is a
significant improvement over the same section of the NOx document. Chapter 5 properly
summarizes the conclusions from earlier chapters, and the conclusions that are presented are for
the most part clearly relevant to the various risk assessment and policy questions that will arise in
later activities related to establishing a NAAQS for SOx. Still, there is room for improvement in
several areas:

   •   Chapter 1  contains a series of six policy questions that the ISA is intended to inform.  It
       would be useful to structure Chapter 5 to specifically address these six  questions, and
       then to phrase the final  concluding statement - about whether effects are occurring at
       existing ambient levels  - as a natural consequence of this structured assessment of the
       information. At present, it is not clear how the concluding statement is related to the
       other findings in the chapter.

   •   The hierarchy of causal claims used in Chapter 5 is appropriate, but the criteria used to
       satisfy each of the categories of causal strength are  not well specified and in some cases
       do not comport with best scientific practice.  This aspect of the chapter can be improved,
       especially with respect to criteria of coherence of evidence and robustness of conclusions.
       A complete description of the approach to causal inference should be provided in a
       revised ISA.

   •   The conclusion that evidence of adverse effects at ambient exposures is robust (p. 5-16)
       appears to be too strong a statement, given the evidence presented in this draft. We
       recommend a slightly less emphatic conclusion given problems associated with biological
       plausibility in the face of heterogeneous information from the various categories of data
       (epidemiological, clinical and animal toxicological). The strongest conclusions should be
       reserved for those responses with the greatest coherence across the available evidence.

   •   The chapter should better summarize the limitations in conclusions about causal
       connections at ambient  levels that are introduced by problems of confounding and

       weakness in exposure assessment underlying epidemiological studies. In general, more
       attention should be given to how the strengths and limitations of the existing information,
       from concentration to exposure to effects, affect the ability to form definitive answers to
       the risk and policy questions in Chapter 1.

With these improvements, Chapter 5 should provide a good template for the way in which this
chapter should be written for future NAAQS ISAs.

7      What are the views of the Panel on the appropriateness of public health impact and
the characterization of groups likely to be susceptible or vulnerable to sulfur oxides?

       This is a good first draft that has correctly identified and discussed data showing which
respiratory effects (hospital admissions, ER visits, and acute bronchoconstriction) are the adverse
health effects of concern for SC>2.  Because of the overlap of Chapter 4 with Chapter 3, it is
recommended that the discussions of susceptible subpopulations be moved to Chapter 3 and that
Chapter 4 focus on discussions of concentration-response relationships, vulnerable populations,
and the potential size of the population  at risk. The nature of the dose-response relationship of
SC>2 concentration with adverse health effects at current ambient levels  is critical to the
estimation of the burden of disease imposed on the population. Chapter 4 addresses dose-
response relationships as addressed in individual studies, both the human clinical studies and
selected epidemiological studies. Toxicological findings that might be relevant are not
integrated into this discussion although  the majority of animal toxicology data are at SC>2
concentrations that are much higher than ambient levels.  Additionally,  Chapter 4 should more
clearly address the general difficulties of exploring the existence of thresholds and the form of
the dose-response relationship at ambient levels and possible influences of co-pollutants in
epidemiological studies. Uncertainty issues and variable sensitivity in the populations (i.e.,
presence of responders) might be addressed more systematically as well.

       The discussion of the extrapolation of the clinical studies, involving generally healthy or
asthmatic volunteers, to ambient exposure levels needs expansion with  a deepening of the
discussion of underlying mechanisms and their potential implications for the dose-response
relationship.  The data summarized in the first three chapters of the ISA allow estimates of (a)
the distribution of outdoor air levels for various averaging times, and (b) the distribution  of
susceptibilities to specific respiratory responses within susceptible subgroups (e.g.,  exercising
asthmatics). Based on these two components, when refined, an integrative analysis is possible
that would inform the agency about the likely overall incidence of effects for current exposures,
and the potential benefits of reducing exposures further.  This kind of preliminary analysis would
be of benefit for inclusion in a revised Chapter 4. Finally, the next version of Chapter 4 should
state more explicitly whether effects are seen (or not) at current annual  and 24 hr time frames
and potentially at 5 minute time frames. The 862 level at which adverse health effects occur is a
key question (stated in Chapter 1 of the ISA); it would be appropriate to directly answer this
question in this Chapter. Public health impact should be addressed from a regulatory  point of
view and the health evidence should be  discussed from this perspective.

8.     What are the Panel's views on the adequacy of this first external review draft ISA to
provide support for future risk, exposure and policy assessments?

       In summary, this document represents a commendable first draft.  A great deal of useful
and policy-relevant information is presented in the carefully crafted summary statements of key
findings and conclusions in Chapter 5. These kinds of summary statements provide a useful
foundation for future risk, exposure and policy assessments. However, there are some areas that
the committee noted that could be improved to make these summary statements even more useful
for policy decisions.

       With regard to exposure assessment, it is important to include cumulative exposure and
spatial exposure plots of the peak 5-minute concentrations (short-term) recently collected by the
states to better inform the extent to which such peaks presently occur in the U.S.  It is also
important to better clarify the relationships among personal, population average, and central-site
SC>2 monitoring data.

       In terms of collectively evaluating all considered health studies, the framework for
making the evaluations of the studies needs further development, including explicitly  stating the
process for reviewing individual studies, the classification of the level of evidence for causality,
the criteria for drawing a causal inference from the body of information in the document, and the
approach to evaluating estimation uncertainty. The criteria for the inclusion of older studies
(e.g., when they provide useful insights to the SOx-health relationship) need to be stated and
consistently followed, as well.

       With regard to deriving risk estimates from human clinical and animal toxicological
studies, it is important to consider how exposure covariates might modify any SO2-health effect
relationship.  For example, does the co-presence of particles or exercise alter the concentration
benchmarks used in policy analysis?

       With regard to epidemiological studies, it is important to better and more critically
evaluate the weaknesses of multi-pollutant models.

       Overall, the draft ISA covers most of the relevant evidence needed, but needs to more
explicitly answer the key question "Does 862 cause human health effects at ambient levels?",
including a more fully integrated explanation of the conclusion across the relevant lines of
evidence. This will provide a more solid foundation for risk assessment and policy formulation.
Specifically, the ISA needs to describe the nature of the concentration-response (C-R)
relationships  of the health effects with 862 concentration. The uncertainties of the attendant C-R
response curve estimation also need to be systematically described for both the clinical studies
and epidemiological studies, including a listing of their respective contributions to estimating

       Finally, the CASAC is pleased to provide the Agency with advice and recommendations
in the development of this ISA, which is a fundamental new part of the NAAQS review process.
We look forward to reviewing the revised version in the coming year.

                                 Rogene Henderson, Chair
                                 Clean Air Scientific Advisory Committee
Attachment A: Roster of CASAC Sulfur Oxides Primary NAAQS Review Panel
Attachment B: Compilation of Individual Panel Member Comments on EPA's Integrated Science
Assessment for Sulfur Oxides - Health Criteria, First External Review Draft (September 2007)

                                                                       Attachment A
                     U.S. Environmental Protection Agency
               Clean Air Scientific Advisory Committee (CASAC)
                  Sulfur Oxides Primary NAAQS Review Panel
Dr. Rogene Henderson (Chair), Scientist Emeritus, Lovelace Respiratory Research Institute,
Albuquerque, NM

Dr. Ellis B. Cowling, Emeritus Professor,, Colleges of Natural Resources and Agriculture and
Life Sciences, North Carolina State University, Raleigh, NC

Dr. James Crapo, Professor of Medicine, Department of Medicine , National Jewish Medical
and Research Center, Denver, CO

Dr. Douglas Crawford-Brown, Professor and Director, Department of Environmental Sciences
and Engineering, Carolina Environmental Program, University of North Carolina at Chapel Hill,
Chapel Hill, NC

Dr. Donna Kenski, Director of Data Analysis, Lake Michigan Air Directors Consortium, Des
Plaines, IL

Dr. Armistead (Ted) Russell, Professor, Department of Civil and Environmental Engineering ,
Georgia Institute of Technology, Atlanta, GA

Dr. Jonathan M. Samet, Professor and Chair of the Department of Epidemiology, Bloomberg
School of Public Health, Johns Hopkins University, Baltimore, MD

Mr. Ed Avol, Professor, Preventive Medicine, Keck School of Medicine, University of Southern
California, Los Angeles, CA

Dr. John R. Balmes, Professor, Department of Medicine, Division of Occupational and
Environmental Medicine,  University of California, San Francisco, CA

Dr. Terry Gordon, Professor, Environmental Medicine, NYU School of Medicine, Tuxedo, NY

Dr. Dale Hattis, Research Professor, Center for Technology, Environment,  and Development,
George Perkins Marsh Institute, Clark University, Worcester, MA

Dr. Patrick Kinney, Associate Professor, Department of Environmental Health Sciences,
Mailman School of Public Health , Columbia University, New York, NY

Dr. Steven Kleeberger, Professor, Lab Chief, Laboratory of Respiratory Biology, National
Institute of Environmental Health Sciences, National Institutes of Health, Research Triangle

Park, NC

Dr. Timothy V. Larson, Professor, Department of Civil and Environmental Engineering,
University of Washington, Seattle, WA

Dr. Kent Pinkerton, Professor, Regents of the University of California, Center for Health and
the Environment, University of California, Davis, CA

Dr. Edward Postlethwait, Professor and Chair, Department of Environmental Health Sciences,
School of Public Health, University of Alabama at Birmingham, Birmingham, AL

Dr. Richard Schlesinger, Associate Dean, Department of Biology, Dyson College, Pace
University, New York, NY

Dr. Christian Seigneur, Vice President, Atmospheric & Environmental Research, Inc., San
Ramon, CA

Dr. Elizabeth A. (Lianne) Sheppard, Research Professor, Biostatistics and Environmental &
Occupational Health Sciences,  Public Health and Community Medicine, University of
Washington, Seattle, WA

Dr. Frank Speizer, Edward Kass Professor of Medicine, Channing Laboratory, Harvard
Medical School, Boston, MA

Dr. George Thurston, Associate Professor, Environmental Medicine, NYU School of Medicine,
New York University, Tuxedo, NY

Dr. James Ultman, Professor,  Chemical Engineering, Bioengineering Program, Pennsylvania
State University, University Park, PA

Dr. Ronald Wyzga, Technical Executive, Air Quality Health and Risk, Electric Power
Research Institute, Palo Alto, CA
Dr. Holly Stallworth, Designated Federal Officer, Science Advisory Board Staff Office,
Washington, D.C.

                                                                 Attachment B
Comments from CASAC Sulfur Oxides Primary NAAQS Review Panel on EPA's
Integrated Science Assessment for Sulfur Oxides - Health Criteria (First External Review
Draft, September, 2007)
Comments from Mr. Ed Avol	2
Comments from Dr. Cowling	7
Comments from Dr. Crawford-Brown	12
Comments from Dr. Kenski	16
Comments from Dr. Kinney	19
Comments from Dr. Larson	24
Comments from Dr. Russell	26
Comments from Dr. Schlesinger	29
Comments from Dr. Seigneur	32
Comments from Dr. Speizer	35
Comments from Dr. Wyzga	39
Comments from Dr. Postlethwait	40
Comments from Dr. Gordon	41
Comments from Dr. Thurston	46
Comments from Dr. Hattis	51
Comments from Dr. Samet	59
Comments from Dr. Ultman	63
Comments from Dr. Balmes	65
Comments from Dr. Pinkerton	70
Comments from Dr. Sheppard	72

Comments from Mr. Ed Avol
General Responses to Charge Questions:
(Charge Questions 1 & 2) To what extent are the atmospheric chemistry and air quality
characterizations clearly communicated, appropriately characterized, and relevant to the review
of the primary SO2 NAAQS? Are the properties of ambient sulfur oxides appropriately
characterized, including policy-relevant background, spatial and temporal patterns, and
relationships between ambient sulfur oxides and human exposure?

The chemistry, characterizations, and properties are generally well-presented, and most of the
presentation is relevant (see specific comments below for some identified exceptions). An
integrated and focused summary of the current state of air quality characterization for SO2
(including what is known, what is not known, and what potential gaps need to be filled) would
have been a useful addition to tie together these sections.

3. Is the information provided on atmospheric sciences and exposure sufficient for the evaluation
of human health effects of sulfur oxides in the ISA?

The provided information appears to be sufficient, but additional information about intra-
community variability of sulfur oxides would be helpful.  It is interesting to note that although
there was a previous determination that a short-term (e.g.  five-minute) standard was not needed
to protect the public health, there is significant effort, discussion, and monitoring based on five-
minute maxima - apparently sometimes to the  exclusion of monitoring for hourly or longer time-
based metrics. A more productive approach might have been to monitor hourly concentrations,
but retain the five-minute readings,  which would allow direct comparison of multiple shorter-
term metrics of interest; this should be considered for recommendation to those areas continuing
to monitor ambient SOx levels.

4. To what extent are the discussion and integration of evidence on the health effects of sulfur
oxides from the animal, lexicological, human clinical, and epidemiological studies, technically
sound, appropriately balanced, and clearly communicated?

I was struck by the heavy reliance on (especially) clinical  and (to a lesser extent) other studies
from 20+ years ago, since this was supposed to be a consideration of information since the last
review (ca. 1995).  If little additional work has been performed but is needed, this "need" could
be identified in a summary section.
Several figures compiled multiple studies into one coherent presentation, and these were
especially useful.  Unlike the NOx document, the SOx ISA appropriately presents informative
peer-reviewed research from studies across the globe, without over-emphasizing the special
status of US studies (and the ISA authors should be  congratulated for this enlightened approach).

That said, Chapter Three's presentation still leaves the reader with a sense that there is a lot of
information for a number of different health outcomes, but the focused integrated  summary of
findings, the strength of those findings, and the confidence of conclusions to be drawn from

those findings, remain elusive. Perhaps a summary table listing the various outcomes, a
determination of level of confidence in the findings, and the implications of the findings would
help guide the reader to a cumulative sense of the current aggregate state of SO2 knowledge.

5. To what extent does the integration of health evidence focus on the most policy-relevant
studies or health findings?

The chapter in which I expected to find an integration of health evidence (Chapter 4 on Public
Health Impacts) seemed to meander between a review of previous research (which would have
been more appropriate for the Chapter 3 Health Effects Review) and a discussion of
susceptibility (which is an appropriate topic for public health impacts, but could also have been
presented in the Health Effects Review chapter and referred to from the Public Health Impacts
section). I do not believe the most policy-relevant studies were well-focused into a complete
integration of health evidence, although some policy-relevant studies were included.  More could
be done and it could be made clearer.
6. What are the views of the Panel on the conclusions drawn in the draft ISA regarding the
strength, consistency, coherence, and plausibility of health effects of sulfur oxides?

The conclusions chapter (Chapter 5) begins as an extended re-visiting of previously described
findings (which could be shortened or consolidated here to make the line of reasoning easier to
follow).  An objective, clear, short table defining the terminology of findings (e.g., causal, likely
causal, inconclusive, plausible, etc) would be helpful in the document in general and in the
conclusions chapter in particular. Much of the basic data needed to make a firm argument is
present, but needs to be re-organized and edited. Strength, consistency, coherence, and
plausibility have not been fully developed or demonstrated, but much of the data is here to do so.

7. What are the views of the Panel on the appropriateness of public health impact and the
characterization of groups likely to be susceptible or vulnerable to sulfur oxides?

Comments about the discussion of public health impacts in the draft are reported in (5) above.
The chapter on public health impacts contained foundational material more appropriate for
earlier chapters. An important presentation about susceptibility (be it by age, genetic, or other
identifying factors) was presented here, but this arguably could have been in a previous chapter
and should have expanded on the inherent difference in susceptibility and vulnerability.

Background information is also included in the chapter on respiratory disease and asthma in the
US, but a clear and direct linkage to SO2 public health impacts is not completely developed in
the integration of this chapter (aside from the implication that if lots of disease is occurring  in the
US and studies have shown that SO2 causes some of it, there must be large numbers of people
affected by sulfur oxides). Integration of the assembled evidence in a coherent manner still  is

8. What are the Panel's views on the adequacy of this first external review draft ISA to provide
support for future risk, exposure, and policy assessments?

This draft is an appropriate beginning but needs some editing, revision, and re-organization to
provide the needed support for future assessments. Much as the NOx ISA draft that preceded it,
the SOx draft lacks a focused logical approach in which each chapter fits coherently with the
next, together creating a synergistic mosaic of current scientific information, and leading to a
concise and supportable number of discrete conclusions and findings. A great deal of the current
draft is consumed with studies reported in the previous  criteria review document or earlier,
raising a question whether no new information is needed, available, or identified.  There is much
in the way of valuable information in the body of the document, but editing is needed to make
this more fully accessible and useful.
Specific Comments:
Chapter 1
Pgl-1, lines 25-26, beginning "For the current review, multiple species of sulfur oxides..." is
awkward and confusing. The following two sentences are easier to understand and explain the
issue, so recommend deletion of this first sentence, and re-phasing to simply utilize the following
two sentences.

Chapter 2
Pg2-3, lines 25-26 - In fact, based on recent emissions inventories (2005), Los Angeles port-
related operations (primarily ship operations) account for over half of the SOx in the LA Basin.

Pg 2-3, lines 27-28 beginning "Even so, SO2 constitutes..." is interesting but irrelevant and
misleading. To imply that SO2 emissions are perhaps unimportant in volcanic emissions
because they constitute a minor fraction by volume ignores the fact that ground-level
concentrations in the plume can be in the ppm-range of exposure.  Recommend deleting this
sentence and let the rest of the paragraph tell the story.

Pg2-6 to 2-7,  Sources of Positive and Negative Interferences in Measurements - There is a page
or more of discussion about possible interferences in the instrumentation measurement.. .but is
this really an issue? The discussion itself points out that there are filters, scrubbers,  and other
commercial approaches that are routinely used to minimize these potential problems, so the
reader is left with a sense of "much ado about nothing" here.

Chapter 3
Pg 3-17, lines!3-32 - All of this has been covered in previous criteria documents; what is new
here that justifies the authors revisiting this information?

Pgs 3-18 to 3-22 - All of this is interesting but has been previously covered in earlier reviews
and criteria documents. What has been reported since the last review that would lead one to re-
evaluate the current standard?

Pg 3-42 to 3-45, Section Integration of Respiratory Effects - The discussion  is useful but
a bit meandering for an integrating section. Perhaps it would be useful to develop an integrated

"summary of effects" table to get a visual perception of the preponderance of evidence of
effects?  Such a table might list various outcomes (lung function, host defense,
hospitalizations,...) and post a "+", "-",  "+/-", or "?" to summarize the current state of
knowledge with regard to a specific outcome. Visually, one might be able to make some
judgment about the strength and breadth of available evidence for effects at a given level (which
could also be presented, if desired, by listing several absolute or relative concentration columns.
Absolute listings that might be considered would be actual concentrations, while relative
concentration columns might be entries like "below current standard", "at current standard", or
"above current standard".

Pg 3-79, linel - Reference is made to unpublished data, which seems inconsistent with the
boundary conditions of the assembled document (i.e., peer-reviewed publications since the
previous criteria review).

Pg 4-2 - Most of this summation of study data seems out of place and more appropriate to the
Chapter 3 presentation.  This chapter is supposed to integrate the previously summarized
information to focus on the public health impact, not on the reporting of individual study

Pg 4-7, Section 4.2.1 Exposure of Susceptible and Vulnerable Populations - By virtue of this
section title, staff has identified a potentially important perspective in the understanding of
affected populations - namely, the difference between "susceptibility" and "vulnerability", in the
context of ambient pollution exposures and effects.  This categorization should be discussed.

Pg 4-7, lines 8-20 - This section reads more like exposure assessment than public health impact.
Shouldn't this be focused on vulnerable or susceptible populations, and not digress into
discussions of specific exposure sources?

Pg4-7, line 25 - Shouldn't something be added to this discussion about genetic susceptibility
based on proposed mechanistic pathways of effect (such as GSTM null and oxidative stress,

Pg 4-8, lines 6-31 - Again, what is presented here is largely repetitive with Chapter 3, and more
logically belongs there.  This chapter should talk about public health impact, not individual study
repots of observed effects.

Pg 4-18 to 4-19, lines 29 to 31 and 1 to 11 - This general health data may well be true, but it is
not linked back to SO2 in this section.

Pg 5-1, line 5 in introductory paragraph - Shouldn't this be corrected to include the phrase
"... since the last criteria review..."?

Pg 5-4, Section 5.2.1 Findings from the Previous Review of the NAAQS for SO2 - Why is this
section here? Why not just present findings relevant to the current and recent information, which
is presumably the justification for this document?

Pg 5-6 to 5-7 - These definitions of terminology (causal, suggestive, likely causal, etc) are
helpful and could be placed in a table for ready reference, to clarify the increasing strength of
connotation for a given word usage.

Pg 5-6, Section 5.2.2 New Findings... - This section title is inconsistent with what it goes on to
contain and refer to, much of which is over 20 years old, and much of which was included in the
previous cycle of document review. The focus of the current document should be the additional
information available since the last document review.

Pg 5-7, Lines 29 to end - What is presented here under "New Findings of Lung Function" are
old studies previously reviewed in earlier criteria review cycles. Either new studies should be
reviewed,  or a conclusion reached that no new studies were found, or that there is a need for new

Pg5-15, Section 5.3 Conclusions, first sentence - This is what the ISA was designed to do, but in
my opinion, this  draft of the document is somewhat off-target.

Comments from Dr. Cowling

               Very General Comments on these NAAQS Review Processes

Before dealing with the details of my specific assignment during the December 5, 2007 Peer
Review of the Integrated Science Assessment for Sulfur Oxides Health Criteria, I would like to
offer a few general comments about these periodic NAQQS Review processes.

In a May 12, 2006 summary letter to Administrator Johnson, CAS AC Chair, Dr. Rogene
Henderson, provided the following statement of purpose for these periodic NAAQS review

       "CASAC understands the goal of the NAAQS review process is to answer a critical
       scientific question:  "What evidence has been developed since the last review to indicate
       if the current primary and/or secondary NAAQS need to be revised or if an alternative
       level or form of these standards is needed to protect public health and/or public

During the past 18 months, CASAC has participated in reviews of three of the existing six
criteria pollutants - particulate matter, ozone, and lead.  CASAC has also joined with senior EPA
administrators in a "top-to-bottom review"  of the NAAQS review processes.  These two
experiences have led to a seemingly slight but important need for rephrasing and refocusing of
this very important "critical scientific question:"

    "What scientific evidence and/or scientific insights have been developed since the last
   review to indicate if the current public-health based and/or the current public-welfare
   based NAAQS need to be revised or if alternative levels, indicators,  statistical forms, or
   averaging times of these standards are needed to protect public health with an adequate
   margin of safety and to protect public welfare?"

I hope this "critical scientific question" will be borne in mind carefully as CASAC joins with
various relevant parts of the Environmental Protection Agency in completing the upcoming
reviews of the primary and secondary National Ambient Air Quality Standards for Sulfur

Thus, I recommend that every chapter in the soon to be completed Integrated Science
Assessment, Risk/Exposure Assessment, and Policy Assessment/Rule Making documents
for sulfur oxides (and the other five criteria pollutants) will contain a summary section
composed almost entirely of a series of very carefully crafted statements of Conclusions
and Scientific  Findings that:
   1)  Contain the distilled essence of the most important topics covered in each chapter,
   2)  Are as directly relevant as possible to the Critically Important Scientific Question

In this connection, I call attention once again to the attached "Guideline for Formulation of
Statements of Scientific Findings to be Used for Policy Purposes." These guidelines were
developed and  published in 1991 by the Oversight Review Board for the National Acid

Precipitation Assessment Program.  The members of the ORB who prepared these guidelines in
the form of checklist questions included: Drs. Milton Russell, former Assistant Administrator for
EPA, Chauncey Starr, former Director of Research for the Electric Power Research Institute
(EPRI), Tom Malone, former Foreign Secretary for the National Academy of Sciences, John
Tukey, Distinguished Professor of Statistics at Princeton University, and Kenneth Starr, Nobel
Prize Winner in Economics.  The intent of these distinguished mentors in science was to assist
other scientists, engineers, and policy analysts dealing with other environmental research and
assessment programs in formulating statements of scientific findings to be used in policy-
decision processes.  These guidelines are the best guides I know of for formulation of statements
of scientific findings to be used for policy purposes:.

   The following guidelines in the form of checklist questions were developed by the NAPAP Oversight Review
Board to assist scientists in formulating presentations of research results to be used in policy decision processes.
1) IS THE STATEMENT SOUND?  Have the central issues been clearly identified? Does each statement contain
   the distilled essence of present scientific and technical understanding of the phenomenon or process to which it
   applies? Is the statement consistent with all relevant evidence - evidence developed either through NAPAP
   research or through analysis of research conducted outside of NAPAP? Is the statement contradicted by any
   important evidence developed through research inside or outside of NAPAP?  Have apparent contradictions or
   interpretations of available evidence been considered in formulating the statement of principal findings?
   statement correctly quantify both the direction and magnitude of trends and relationships in the phenomenon or
   process to which the statement is relevant? When possible, is a range of uncertainty given for each quantitative
   result?  Have various sources of uncertainty been identified and quantified, for example, does the statement include
   or acknowledge errors in actual measurements, standard errors of estimate, possible biases in the availability of
   data, extrapolation of results beyond the mathematical, geographical, or temporal relevancy of available
   information, etc. In short, are there numbers in the statement? Are the numbers correct? Are the numbers relevant
   to the general meaning of the statement?
   CLEARLY? Have appropriate statistical tests been applied to the data used in drawing the conclusion set forth in
   the statement? If the statement is based on a mathematical or novel conceptual model, has the model or concept
   been validated? Does the statement describe the model or concept on which it is based and the degree of validity of
   that model or concept?
4) IS THE STATEMENT CORRECT WITHOUT QUALIFICATION?  Are there limitations of time, space, or
   other special circumstances in which the statement is true? If the statement is true only in some circumstances, are
   these limitations described adequately and briefly?
5) IS THE STATEMENT CLEAR AND UNAMBIGUOUS? Are the words and phrases used in the statement
   understandable by the decision makers of our society? Is the statement free of specialized jargon?  Will too many
   people misunderstand its meaning?
   MISUNDERSTANDING? Are there any excess words,  phrases, or ideas in the statement which are not necessary
   to communicate the meaning of the statement? Are there  so many caveats in the statement that the statement itself
   is trivial, confusing, or ambiguous?
   VALUE JUDGMENTS? Is the statement free of influence by specific schools of scientific thought? Is the
   statement also free of words, phrases, or concepts that have political, economic, ideological, religious, moral, or
   other personal-, agency-, or organization-specific values, overtones, or implications? Does the choice of how the
   statement is expressed rather than its specific words suggest underlying biases or value judgments? Is the tone
   impartial and free  of special pleading?  If societal value judgments have been  discussed, have these judgments been
   identified as such and described both clearly and objectively?
   courses of action and their consequences inherently involves judgments of their feasibility and the importance of
   effects. For this reason, it is important to ask if a reasonable range of alternative policies or courses of action have
   been evaluated? Have societal implications of alternative courses of action been stated in the following general
     "If this [particular option] were adopted then that [particular outcome] would be expected."
   Acknowledgment of potential sources of bias is important so that readers can judge for themselves the credibility of
   reports and assessments.

 My Specific Assignment in this CASAC Peer Review of the First External Review Draft of
          the Integrated Science Assessment for Sulfur Oxides - Health Criteria

My specific assignment in preparation for the December 5, 2007 CASAC Peer Review of the
"ISA for Sulfur Oxides - Health Criteria" as outlined in CASAC Chairman Rogene Henderson's
memo of November 2007 is Charge Question 8 - What are the Panel's views on the adequacy
of this first external review draft ISA to provide support for future risk, exposure and
policy assessments.

Chairman Henderson also gave this same assignment to two other CASAC panel colleagues -
Drs. George Thurston and Jon Samet. Thus, I am very much looking forward to comparing
notes with both George and Jon during our CASAC Peer Review meeting on December 5.

My own view is that Chapter 5 of this First External Review Draft ISA - Key Findings and
Conclusions — is very adequate indeed in providing support for future risk, exposure, and policy
assessments regarding the health effects of sulfur oxides. I offer high praise for this summary
chapter because it fulfills more adequately than any Criteria Document or Integrated Science
Assessment document I have seen before in providing very carefully crafted summary statements
of scientific findings that conform very well to all but the last of the nine checklist questions in
the  above listed "Guidelines for Formulation of Statements of Scientific Findings to be used for
Policy Purposes."

Each major section of Chapter 5 consists almost entirely of simple declarative statements of
policy-relevant scientific findings that very adequately summarize the current scientific
information contained in the earlier chapters of this ISA document including:
       •  Four summary statements about Emissions Sources, Atmospheric Science, and
          Ambient Monitoring Methods as discussed in Chapter  1,
       •  Five summary statements about Ambient Concentrations of sulfur oxides as discussed
          in Chapter 2,
       •  Five summary statements about Exposure Assessment  as discussed in Chapter 3,
       •  Twenty-seven summary statements about New Findings on the Health Effects of
          exposure to SO2 - including separate summary statements derived from the scientific
          data and information discussed in Chapters 3 and 4 with regard to:
          •    Peak (5-15 minute) Exposure to SO2 and Respiratory Health Effects including
               Respiratory Symptoms and Lung Function,
          •    Short-Term (24-hr average) Exposure to SO2 and Respiratory Health Effects
               including Respiratory Symptoms, Lung Function, Airway
               Hyperresponsiveness, Inflamation,  and Respiratory Emergency Department
               Visits and Hospitalizations,
               Short-Term Exposure to SO2 and Cardiovascular Health Effects,
               Short-Term Exposure to SO2 and Other Systemic Effects,
               Effects of Short-Term Exposure to SO2 and Mortality,
               Effects of Long-Term Exposure to SO2 and Mortaility,
               Concentration-Response Function and Potential Thresholds, and
               Susceptible and Vulnerable Populations.
One could of course also quarrel a bit about:

1) Whether it is optimal and correct to use 862 instead of sulfur oxides in very single one of
   the many side headings listed above.

2) Whether adequate attention is given to the chemically "reduced sulfur gases in the
   atmosphere" as discussed at several places on pages on pages 2-1, 2-4, and 2-24.

3) Whether it would be useful to ask each of the authors, contributors, reviewers, and EPA
   scientific staff to acknowledge their "professional biases" (as suggested in the last
   Checklist question in the "Guidelines") as well as to provide their institutional affiliations
   as already done on pages xiii - xx in this document.

4) Whether Chapter 2 has an optimal title. This chapter highlights key concepts or issues
   relevant to understanding the atmospheric chemistry, sources, exposure and dosimetry of
   sulfur oxides, following a "source to dose" paradigm."  The idea of dealing with atmospheric
   chemistry all the way from emissions sources to dosimetry in the lung is a good one; but
   titling the chapter "Source to Tissue Dose" is a little too "cute" to be taken seriously. In my
   opinion, "Chemistry and Dosimetry of Nitrogen Oxides" would be better as a title for this
   important chapter.

5) Design and Content of Figure and Table Captions. In my opinion, every figure and table
   in any Integrated Science Assessment document —that is clearly to be used for policy
   purposes — should "stand alone" to  the maximum extent possible and not be any more
   dependent on descriptions in the text than absolutely necessary for understanding by

6) Etc, Etc

Comments from Dr. Crawford-Brown
This review follows the Charge Questions for the document review (at the end) and provides
additional comments on Chapter 5: Findings and Conclusions. A general comment is that I found
the document generally appropriate, both with respect to the particular studies examined and the
conclusions drawn from those studies. There was some confusion in my mind, however, about
the period of time being covered. The document mentions documents produced in 1982, 1986,
1988 and 1994. It then focuses on information obtained since the 1982 document. I presume this
is because it is the last year an AQCD was produced, but it seemed a long period of time to be
counted as "new" data. Still, once I accepted this premise, the document flowed smoothly.

As with the NOx ISA, I am not convinced that Chapter 5 serves the purpose for which it is
intended. My understanding of such a chapter is that it will provide the input into subsequent
rounds of the NAAQS process. As such, it should be organized around a series of questions an
assessor is likely to ask, and should provide answers to those questions so the assessor need only
go back into the primary chapters to clarify some points. The answers to these questions should
allow the assessor to determine whether anything about the data obtained since 1982 would
cause a change (relative to the existing NAAQS) in any policy relevant question an assessor
might ask in subsequent NAAQS stages. The current document does not do this. I consider the
different subsections of the chapter below.

On Source to Dose Relationship, I believe the authors should have a succinct statement about
whether the existing monitoring methods, including locations and numbers of samples, provide
an adequate basis for estimating ambient concentrations during the period between 1982 and
today. This should include a statement as to the representativeness of the monitor results for
estimating ambient conditions in specific populations in the U.S., so the assessor can eventually
determine which populations might be selected for analysis of scenarios of exposure. It also
should include a  statement of the best judgment of the ratio of ambient over personal exposures.
A lot of data are presented that are relevant to this question, and they seem  to suggest a ratio of
about 5 or 6 with a GSD of perhaps 2, but this summary is not provided in Chapter 5. With
respect to Dosimetry, the authors should provide a conclusion as to the implications of these
dosimetric factors for extrapolation to the general population. If nothing else, the conclusion
should be that the dosimetric results show rather clearly why the switch from nasal to oral
breathing during exercise is important, which in turn alerts the assessor to the fact that the
sensitive subpopulation will be people who are exercising.

On Heath Effects Findings, there should be a succinct summary of (1) the kinds of effects for
which there is a relationship with current  ambient levels of SOx (the authors in part accomplish
this), (2) any health benchmarks suggested by the data (this is accomplished more in the body of
the document than in Chapter 5, where it is most important), and (3) the concentration-response
relationship obtained from the different epidemiological studies (again, the body of the document
contains some excellent summary figures showing odds ratios, etc, but this level of information
is not carried back into Chapter 5). As written, Chapter 5 leaves the reader with a long summary
of findings without trying to focus attention onto any specific set of conclusions that can be used

by the assessor in calculating health endpoints. I realize this might be deliberate, so the ISA
won't constrain the ways in which an assessor will eventually calculate risk, but there should still
be some more summary statements made about health benchmarks and concentration-response

Throughout this same Heath Effects section of Chapter 5, the authors use the phrases "causal",
"likely causal", etc. I am supportive of this classification system, and the conclusions they have
drawn seem to me justified. However, it is not clear whether these phrases are to mean
something like "causal  at all levels of exposure", "likely causal at all levels of exposure", etc.
My concern here is that the causal link depends on level of exposure, and this classification
scheme loses that distinction. It appears to be more like the practice in Hazard Identification - a
practice I don't support - which asks, for example, whether a compound is or is not a carcinogen,
rather than whether it is a carcinogen  at specific levels of exposure and by particular routes of
exposure. I think a more nuanced statement of causality is  required in Chapter 5, one that focuses
on whether there is a causal link at levels of exposure of policy relevance likely to be considered
by subsequent analysts  in the NAAQS process.

There are several places where the  authors summarize a risk coefficient with a relative risk (as on
Page  5-11) but don't mention that it is relative risk. If one took the first bullet in Section
literally, it would appear that a person exposed at 10 ppb would have a percent or two probability
of dying! The real answer is, of course,  an increased probability of dying that is one or two
percent of the background probability. This is an example of where the authors of the ISA must
be very careful to ensure that assessors using the document later apply the correct model of risk.

In Section, the authors go to pains to mention Krewski et al's conclusion that "the
absence of a plausible toxicological mechanism by which SO2 could lead to increased mortality"
suggests that  "SO2 might be acting as a marker for other mortality-associated pollutants".  I
disagree with this statement. The lack of a mechanism being found may be simply a limitation of
the existing studies. It isn't evidence one way or the other for SO2 being a marker.

Throughout the Chapter, there is no mention of the kind of concentration-response model one
might expect to apply. I don't mean the shape of the model (linear, quadratic, etc) but rather the
mechanistic basis. For example, the data in earlier chapters seem to suggest there is some sort of
distributed threshold model at play, with each individual having a threshold but with this
threshold differing from person-to-person. In that case, the shape of the curve depends on the
PDF of thresholds in the population, with the low exposure portion of a population exposure-
response curve being driven by individuals with a low threshold.

On Page  5-16, the authors repeat what I believe is a mistaken logic from an earlier chapter: that
the public health impacts are expected to be large because the size of the susceptible population
is large. A large susceptible population is  important, but so is the level of actual exposures in
relationship to any thresholds. I don't believe enumerating the size of the susceptible population
tells us much about public health impact. Now,  potential public health impact is another story.

Before turning to the Charge Questions, I provide here a few comments from earlier chapters.

1. In Chapter 2, some of the figures (such as 2.4-5) are almost impossible to read because the
data are so overlapping.

2. On Page 2-31, line 11,1 believe the authors mean Consolidated Human Activity Database.

3. On that same page, line 3, the loss processes are not only during infiltration, but due to plating
on surfaces inside the structure.

4. While reading the section on Relationship of Personal Exposure to Ambient Concentration, I
kept being struck by the obvious implications of these results for Tier I in the Draft Assessment
Plan. These results indicate that Tier I could serve reliably as a very conservative upper bound
estimate on effects.

5. In Section 2.5.4, the authors should construct a succinct statement, carried into Chapter 5, as to
the implications of exposure error on slope factors and health benchmarks. They offer some good
hints in this section (that random error results in bias of a slope factor towards the null, and that
existence of a Personal-Ambient Ratio shifts the response curve uniformly - as mentioned on
Page 2-41), but these implications are never used to draw any summary conclusions about the
slope factors and health benchmarks that will appear in Chapter 5.

6. In Chapter 3, the authors provide many useful summary figures (the first is 3.1-1) but then
don't provide any summary conclusions from these figures. I don't see why a summary range of
odds ratios in such figures can't be established.

7. In Chapter 3, there also is clear evidence that very short-term exposures, on the order of a few
minutes in exercising individuals, is sufficient to produce adverse effects at some levels of
exposure.  This has obvious implications for the averaging period selected in a standard, and so it
should be  emphasized here and in Chapter 5.

8. In Chapter 3, the authors provide some past meta-analysis results, but never attempt a meta-
analysis themselves for other results,  choosing instead to provide plots of the range of results. I
was not sure why this is the case, and assume it is because of some mistrust of meta-analyses?

9. On Page 3-78, the authors summarize the results of the Hong Kong intervention study. They
make much of the initial decline in effects after reductions in SOx, but skirt over the rebound. If
the rebound shoots past the baseline mortality (and I am not saying it does  as I did not find a
copy of this study), this could offset the initial decline. At least some assurance that this did not
take place is needed. Otherwise, the overall beneficial effect will be overstated.

10. At least on first glance, the results in Figure 3.4-1 don't appear much less conclusive than for
many of the figures showing morbidity effects. I'm not sure what to draw from this observation,
other than that it somewhat contradicts the position of having much less conclusive causality
claims for mortality than for morbidity based solely on the epidemiological evidence.

11. In Chapter 4, when describing the epidemiological studies, I kept looking for exposure-
response data plotted. It is unsatisfactory to be simply told that the data are linear without seeing
them plotted with error bars.

12. At the end of Chapter 4, the authors provide an estimate of the number of individuals in the
susceptible population. As I mentioned with respect to Chapter 5, the size of the population may
be large but this must be coupled with exposures of health concern to produce a population risk.
The latter consideration has not been applied here, although perhaps the goal was simply to show
that there is a large population that is potentially at risk if exposures are sufficiently high.

I turn now to the 8 Charge Questions, drawing on many of the comments made above.

1. Yes, I feel that all of this was simple to understand and relevant. However, this is not my area
of expertise, so I can't testify as to whether there might be important missing pieces of

2. Yes, I believe the properties are characterized at a level needed for subsequent assessments. I
do, however, believe the authors need to draw a summary conclusion as to the distribution of
personal-to-ambient ratios that might be applied in a variability analysis of exposure and risk.
This should at least be provided in Chapter 5.

3. The information is good, but the reader is left to draw summary conclusions without adequate
guidance. The authors need to provide succinct summaries of the conclusions and place these
into Chapter 5.

4. The health effects discussion is clear, although again there is a need to draw better summary
conclusions and point the reader towards these in Chapter 5. This is particularly true of any
suggestions for health benchmarks and/or exposure-response relationships.

5. The document does  focus on the most policy-relevant findings, but (as mentioned in Question
4) does not draw summary conclusions that will be needed in policy determinations. This is in
part because the authors have chosen a system of causal claims (causal, likely causal, etc) that
does not specify the exposure level at which these claims  are valid. This aspect needs to be

6. My answer here is the same as in Question 5. The causal claims are justified, but need a bit
more nuance by stating the levels of exposure at which they apply.

7. The appropriate subpopulations have been selected, but the document gives the incorrect
impression that a large number of individuals in this subpopulation means  a large public health
impact. This is not true unless these subpopulations are also exposed at levels above health

8. With the improvements I have suggested for Chapter 5, the document can provide an adequate

Comments from Dr. Kenski

General: Kudos to the team preparing these documents! This ISA was in much better shape
than the NOx ISA we reviewed a month ago.  I found it to be well written and thoughtful. It was
also more truly integrated, in the sense that there were useful summaries that drew reasonable
and defensible conclusions, without overinterpreting or putting too much spin on the data.

Section 2: The description of SOx chemistry is adequate. However, the discussion of ambient
air quality data for SO2 would benefit from some additional detail.  For example, the Figures
2.4-2 and 2.4-3 give a good regional picture, showing the decline in regional SO2 and  SO4 over
the last 10-15 years. But because this is CASTNET data, it comes from monitors that are sited in
rural areas that are, by definition, far from sources and people. So these plots probably don't
show us concentrations that the majority of people are exposed to.  Wouldn't a map that used the
AQS data, including all those urban sites, also be informative? A map of the NAMS/SLAMS
network should be included here or in the Annex.  We are given  Table 2.4-2, but it never actually
says how many monitors are  providing data - is it the sum of the 3-yr averages inside and
outside CMSAs? Are these monitors just NAMS/SLAMS, or are CASTNET sites included?
Much is made of the fact that the CMSAs with 4 or more monitors have plenty of data below the
detection limit, but I don't believe it's ever mentioned that concentrations are higher in urban
areas than in rural areas, at least as evidenced by the medians and 75th %iles in Table 2.4-2.
Surely this is important to our basic understanding of SO2 ambient concentrations.

p. 2-13: Figure 2.4-5 is not very helpful as currently plotted; perhaps using a log scale would be
more effective at showing what the real distributions are. Lines  11-13 on p. 2-13 state that
(using this plot as a basis, presumably) highest concentrations are reached at midday or during
the middle of the night.  That may be true, but this figure does not really allow one to draw that
conclusion. If this is a point that should be made, then a better graphic is required, one that
draws our attention to the central tendency of the data and not those pesky outliers.

p. 2-14: The lack of correlation among SO2 monitors was helpful to point out; it would be also
be helpful to augment this discussion with a summary of how the monitors are  sited (as described
in the AQS system at a minimum)—i.e., how many are  source oriented, how many are
community monitors, even if it's relegated to the Annex. Because some monitors  are source
oriented and others population-based, it is perhaps not surprising that there is a lot of intra-city
variability. A review of data that show this intra-city variability  (e.g., saturation studies) would
be useful and would help readers understand that monitors in the national networks have a
limited ability to characterize the concentration gradients that exist in urban areas (for a very
recent example, see Wheeler et al., Intra-urban variability of air pollution in Windsor, Ontario—
Measurement and modeling for human exposure assessment, Environmental Research 106
(2008) 7-16, available online).

p. 2-24: The discussion of correlations between SO2 and SO4 needs a conclusion, even if it's as
simple as noting that there is  no consistent evidence for correlations across the country. This
section also needs some quantitative description of correlation or lack or correlation between

SO2 and other copollutants, especially in light of the health effects that can be confounded by

Section 3:  I appreciate the consistent use of summaries at the end of each of these sections on
health effects.

p. 3-2, lines 1-15: In light of the data presented in Sec 2 that showed no or very weak
correlations between SO2 and SO4, this paragraph is confusing.  Multipollutant models should
perform best when the species of interest are not correlated. Effects of SO2 might be
confounded by copollutants, but SO4 seems unlikely to be one of them.

p. 3-45, line 8: should this be  'There were also no key human...?

p. 4-1, line 31: needs a comma after effects
       line 32: needs a comma after background level
p. 4-10: lines 1-2 are repeated  from previous page, and lines have apparently been dropped from
the bottom of this page

p. 5-2, last bullet: This bullet  should make it clear that it's summarizing concentrations that are
averaged over these regions. Because the NAAQS applies to individual monitors, not to regions,
another bullet should summarize the actual NAAQS-relevant concentrations, i.e, give the ranges
of annual average and max 24-average concentrations at the regulatory monitors (e.g., the
interquartile range of annual average concentrations was from  1 to 6 ppb in urban areas, and the
maximum annual average was 148 ppb at Some city, Some State.) The conclusions put it
perfectly (p. 5-15, lines 22-26) although, oddly, these same numbers aren't presented anywhere
else in this Section.

p. 5-3, line 1:  This bullet is a bit of a red herring.  The generally slow  conversion of SO2 to SO4
and the fact that emissions are often from hot plumes and elevated stacks are sufficient reasons to
believe that the 2 species would not be highly correlated at ground based monitors.  The more
important  conclusion from the correlations presented in Sec. 2 is that there is huge spatial
heterogeneity in SO2 on an urban scale. This finding is the one that has the  most serious
implications for exposure assessment, because it means it will be more difficult to accurately
characterize the concentrations that populations are exposed to.

As mentioned above, we should have some a bullet here about what other pollutants SO2 is
correlated with, if any.

p. 5-9, line 32: amount -> among

Annexes:  These really need a comprehensive table of contents that includes subheadings so
those unfortunate souls reading it without an electronic version can actually  find information
without thumbing through a hundred pages or so.

Annex 2:  It does make sense to include NOx chemistry in this ISA, but it was a little
disconcerting to open up Annex 2 and find the first 20 pages devoted to NOx instead of SOx.

Please add a prominent sentence or two to the introduction explaining the rationale for this
decision, as it is easy to overlook the footnote on p. 1-5.  The aforementioned table of contents
would make the logic and presentation of this information more apparent as well.

Table AX2.6-1 is not very clearly formatted. It's difficult to determine which items are summed
to make up the subtotals.  Offset the totals from the individual entries or otherwise make it
obvious that not all of the numbers in each column can be added together.  Some items look like
they should have been bolded but are not - e.g., solvent utilization? Metals processing?

Comments from Dr. Kinney

Specific Comments:

P. 2.10, figure 2.4-2: add footnote for conversion of ug/m3 to ppb concentration units

p. 2-31, line 18: change "characterized" to "simulated." It is important for readers to understand
that you are not estimating actual personal exposures of individuals by these methods, but you
ARE simulating the population distributional characteristics.

p. 2-32, para ending line 31: Add the limitation that, because of sensitivity and LOD issues, it is
impossible to characterize hourly or shorter personal  exposures, which is a major limitation of
available personal monitoring technology since micro-environmental exposures at these time
scales are likely to be very important.

p. 2-37, para ending line 3:1 question whether these studies warrant such extensive and detailed
review here given the LOD problems with the personal samples. I would replace this with a
short paragraph stating that LOD issues with currently-available technology preclude our ability
to address ambient/personal relationships for short averaging periods.  The paragraph which
follows is probably adequate.  Just refer  further discussion to the Annex.

p. 2-38, line 14: This is an odd way of expressing general concerns over confounding by co-
pollutants, and it coveys a strong presumption that SO2 cannot be the true causal factor. Please
edit to offer a more balanced viewpoint.

p. 2-42, line 9: I don't follow the logic here.

p. 2-45, whole section: Throughout this section, this discussion of dosimetry provides far too
much detail about studies, most of which were presumably reviewed extensively in previous
CDs.  Instead, here it would be sufficient to summarize the common findings regarding regional
dosimetry with and without exercise.  Details are not needed.

p. 2-46, line 16: this summary paragraph should be the main content of this section,  as noted

p. 3-2 line 11:  this statement about SO2 and SO4 is directly contradicted by the data presented
in chapter 2 demonstrating very low correlations between SO2 and SO4 in most locations

p. 3-2, line 13: It would be fair to say that two-pollutant models involving both SO2 and
PM2.5(or so4) are a valid way to assess the relative health impacts of the two pollutants. If one
or the other pollutant is more robustly associated with health, that supports the interpretation that
the robust pollutant may be the more valid measure of risk.

p. 3-6, first paragraph: Were any of the co-pollutants independently assocaited with symptoms,
and if so, were their effect estimates as robust as was SO2's?  See lines 29-30 of same page for
the kind of information that is needed here.

p. 3-9, lines 17-18: This possibility seems to be raised out of context. The question is whether
positive studies for SO2 have examined robustness wrt PM. If they haven't done so, or SO2 is
generally not robust when they did, then it's reasonable to raise this concern, but absent a link to
the findings of the studies, this is a red herring that conveys an a priori bias towards PM effects
on the part of the author.  In the text that follows, the only study that actually examined this
issue found little evidence to support the statement.

p. 3-16, line 2: O3, not PM, was generally the most robust pollutant in the health studies
reviewed above.

p. 3-22, line 28: this is very light exercise

p. 3-23, second para: The Koenig results at 0.1 ppm should be noted here as well, since the co-
exposure regimen used there is a realistic simulation of ambient conditions.

p. 3-27, first full para: Did Boezen examine co-pollutant effects in either study?  Worth a
mention here.

p. 3-27, second full para: what concentration  of SO2 was used in the Nowak study?

p. 3-30, line 25: this is confusing.  Pneumonia isn't part of COPD.  Edit.

p. 3-31, lines 9-11: This statement is not supported by the evidence you've  presented, e.g.,
Wilson et al results for ages 15-64. This is important as this concept of no adult effects is carried
through the rest of the document.  In figure 3.1-8, results jump around a lot for all ages, no more
so for the adults than any other age group in my view.

p. 3-35, line 13: Seems inconsistent with Petroeschevsky  results plotted in fig 3.1-7

p. 3-35, line 18-20: Needs re-phrasing.  Say something like many studies observed positive
associations, some of which were  statistically significant.

p. 3-35, lines 25-29: among the few adult results, it's true that few are statistically significant, but
Wilson is, and several others are consistently positive.  There's really no basis to make claims
about effect modification by age based on the available data displayed in figure 3.1-10

p. 3-39, lines 1-3: this is a more accurate summary than the one that led off the section.

p. 3-40, lines 15-17: My read of figure 3.1-11 is that SO2 has sometimes been robust and other
times not in co-pollutant models.  This summary statement, and the similar one leading off the
paragraph,  is not consistent with the evidence.

p. 3-42, line 5: replace "generally robust" with "often robust"

p. 3-44: first para: Should mention the Nowak et al. human controlled exposure study in this
paragraph too. It seemed like the most clear indication that BHR was associated with SO2

p. 3-44, line 27: change "were not sensitive" to "were moderately sensitive". Also, delete second
half of sentence.

p. 3-45, lines 1-2 (and previous lines): I find this statement about biological plausibility to be
unwarranted given the hugely different SO2 concentrations used in the epi vs. experimental

p. 3-45, line 8: Did you mean to say "there were also no key.."

p. 3-45, line 15: add a sentence explaining what these markers have to do with cardiovascular

p. 3-46: The reader does not need so much detail on these studies of HRV since they say very
little about SO2 effects per-se. Reduce to one short paragraph summarizing overall findings and
interpretational problems wit SO2.

The same is true of section on cardiac arrhythmias. The epi studies reviewed there tell us
little or nothing about SO2 because of the co-pollutant mix.  I would eliminate most of the detail
and simply state in a few sentences that there is some evidence for SO2 but that it's confounded
by co-pollutants that are likely more relevant, i.e., PM.

p. 3-50, section these first two paragraphs could be reduced to one sentence stating that
there is no clear epi  evidence for robust SO2 effects on blood pressure.

p. 3-51, lines 12-13: this says it all, and is all that needs to be said.  The annexes are where
detailed study reviews need to be.

p. 3-51, section same thing

p. 3-53, para starting line 14: what happened when other pollutants were added in these studies?

p. 3-57, section on cerebrovascular effects: I wonder whether it is necessary for this document to
catalogue every published health outcome for which SO2 results have been reported, regardless
of how biologically  implausible or far-fetched the cause-effect relationship is. Just because it's
been reported doesn't mean it meets the standard of relevance to understanding the health effects
of SO2. The limitations of epidemiology, especially ecologic time series or cross sectional
studies, adds another major layer of uncertainty, rendering these  findings largely irrelevant to the
current purpose.

p. 3-58, lines 7-8: While I tend to agree that any observed SO2 effects seem likely to represent
some form of confounding, the results in figure 3.1-13 do not support this statement.

p. 3-60: Exposure levels and/or durations are so far from ambient in most of these experimental
studies that I question the inclusion in this document which is dealing with setting an ambient air

p. 3-62, lines 16-20: this is all that needs to be said about this literature,  just add the relevant
citations to this short paragraph and delete above text and table.

p. 3-64, line 24: which is a good thing in my view,  the BAD thing about meta analyses wrt
multicity studies is the problem of publication bias.

p. 3-75, line 17, insert "to some extent" between "confounded" and "by"

p. 3-79, sentence from line  15-17: delete.  Not relevant.

p. 3-80, lines 5-6:1 agree with this comment about biological plausibility, but it appears denovo
here, without any thoughtful discussion about biological mechanisms and plausibilty.  A section
should be added which provides such discussion if this statement remains here.  Otherwise, can
do later in an integrated assessment.

p. 3-82, line 8: what about patterns of co-pollutant concentrations and health effects?

p. 3-82, line 18: delete "moderately"

p. 3-84, line 13: Add that a major problem with geographic studies comparing several different
communities is confounding by co-pollutants. There are few if any situations where SO2 is the
only pollutant that varies across metro areas. At best, such studies may indicate something
regarding health effects of "bad air" but rarely will provide pollutant-specific information. In
light of this, I think the detailed study descriptions given here are more than is necessary and the
whole section could be reduced to one long paragraph with references.

p. 3-86, line 8: In addition, level and duration of exposure were far larger than ambient

p. 3-88, line 13-14: once again, need to mention the very much higher than ambient
concentrations  used here.

p. 3-93, line 16 and others: this is not air concentration units; what does it mean?

p. 4-1,  line 28: shallow slope not really an "error  source"

p. 4-3,  first para: aren't the Koenig et al.,  ozone/so2 results relevant here (0.1 ppm SO2 effects)?

p. 4-14, lines 7-8: as noted earlier, i'm not sure this exclusion of intermediate ages is justified by
the evidence.

4-14, lines 26-27: what is meant by "reduce expression of function in the lung?"

p. 4-16, line 23; delete "other"

p. 4-18, line 11 e.g.,: there should be a greater focus here on asthma as the primary disease
condition of interest with respect to SO2.

p. 4-19, last para: It would be helpful to summarize population numbers for these specific groups
here, both as numbers and proportion of total population.

p. 5-1, line 31, append: "As a result of this chemical transformation,  SO2 concentrations
diminish downwind of sources as sulfate concentrations increase."

p. 5-2, line 20, insert "of hourly concentrations" after precise measurements..

p. 5-8, line 20: change "generally" to "often"

p.  5-10, line 1: change "generally" to "often"

p. 5-10, lines 2-6; I don't think it's reasonable to invoke biological plausibility when
concentrations are two orders of magnitude higher in the experimental vs. epi studies.  Need
more nuanced statement here that takes this uncertainty into account.

p. 5-10, line 27: change "of to "on"

p. 5-14, line 18: change "respiratory diseases" to "asthma"

p. 5-14, lines 23-24: change "respiratory illnesses, particularly asthma" to "asthma"

p 5-15, line 3: here's another statement about ages, which is only weakly supported by the
evidence in my opinion.

p. 5-16, line 5: insert "somewhat" before robust.

p. 5-16, line 8: delete phrase after "causes"

p. 5-16, line 10: delete "these"

p. 5-16, line 19, insert ",difficulty in separating SO2 effects from other co-pollutants," after "risk

p. 5-17, line 16-21: It seems that the Koenig et al., o3/so2 controlled exposure study would be
relevant to mention here, or at least the observation that ozone appeared to potentiate so2 effects.

Comments from Dr. Larson

General Comments on Chapter 2 Charge Questions

1.  To what extent are the atmospheric chemistry and air quality characterizations clearly
communicated, appropriately characterized, and relevant to the review of the primary SO2

This chapter has an adequate but succinct summary of the relevant atmospheric chemistry of SC>2
as it relates to the ultimate levels of SO2 in the ambient air. The air quality characterizations are
also clearly communicated. However, due to the preponderance of below detection values in the
regulatory observations, it is difficult to get an accurate picture of the actual concentrations in
most urban areas. For example, to assess the spatial variability across metropolitan areas there
are only 12 MS As with four or more monitors, with some of these located near industrial
sources. Contrast this with the SAVIAH epidemiological study by Pikhart et al,(2001) where
outdoor passive samplers were deployed  at over 100 sites to estimate spatial variation.

While it is true that more  and better measurement methods would provide more precise
characterizations (for a more sensitive method, see Matsumi et.al, Atmospheric Environment 39
(2005) 3177-3185), it is not clear that the focus should be on a broad regional characterization.
It would seem just as reasonable to focus on those areas with relatively high SC>2, either selected
MS As or locations near major sources. One might then provide more a more thorough air
quality characterization in a more limited geographical  region. From an epidemiological
perspective, the relevant co-pollutants may be different in these higher SC>2 regions than in the
country at large or in the European and Asian cities where adverse health associations are

2. Are the properties of ambient sulfur oxides appropriately characterized, including policy-
relevant background, spatial and temporal patterns, and relationships between ambient sulfur
oxides and human exposure?

The characterization of policy relevant background levels is mostly appropriate. However it is a
little misleading to say that these levels can comprise up to 70% of the SO2 in  the northwestern
U.S. Most urban areas in this part of the  country are not strongly influence by  volcanic

I would also take exception to the conclusion on page 2-37 that "when personal exposure
concentrations are above detection limits, a reasonably  strong  association is observed between
personal exposures and ambient concentrations".  This statement is essentially repeated in the
conclusions chapter in section 5.1.3. As best I can tell,  this is  based primarily on the results of
Brauer et al.,  1989. While this might be a perfectly good study, it was done in one city over two
seasons.  One study does not justify this important generalization.

Is the information provided on atmospheric sciences and exposure sufficient for the evaluation of
human health effects of sulfur oxides in the ISA?

From an epidemiological perspective, there is not sufficient information. The current ambient
levels are very low and poorly measured in most locations.  There is also insufficient
information to predict personal exposure levels from these low ambient levels measured at most
community monitors. From a toxicological perspective, perhaps one can draw conclusions in a
limited number of locations based on the short-term effects seen in controlled human exposure
studies and the measured hourly or daily ambient levels at the upper end of the concentration

Specific Comments on Chapter 2

page 2-13      The mean values  are visually difficult to distinguish from those above the 95th
percentile in Figure 2.4-5.

page 2-14      Are the correlation coefficients based on hourly or daily values?

page 2-30      Lawn equipment??  Perhaps this refers to PM or VOC emissions.

page 2-35      Is there a reference for the general statement that SO2 concentrations increase
with height above ground? This may be true near elevated point sources, but not during
nighttime inversions.

Page 2-36     What can we conclude from these negative slopes? That personal exposures
       decrease with increasing ambient 862 concentrations??? This discussion is confusing.
Page 2-28      The last sentence starting with "Thus," is poorly worded.

Comments from Dr. Russell

       Like the NO2 ISA, I like the idea of the second chapter being nicely trimmed down,
getting to the point of going from the source to the dose. Similar to before, there are areas to be
strengthened and refocused.

       First, the discussion of the gas phase oxidation of SO2 to sulfuric acid, and then what
happens to sulfuric acid monomers needs to be strengthened.  Sulfuric acid monomer is, as
stated, very water soluble. However, upon formation, its very low vapor pressure is rather key.
It will condense on preexisting particles or can nucleate to form nanoparticles. It will not only
transfer to water droplets.  Further down page 2-2, (line 25), I would say that at pH's above 5.3,
ozone oxidation becomes increasingly important.  Another piece of chemistry to be brought in to
this discussion is that as sulfuric acid is formed, the pH of the aerosol drops,  and this may lead to
increased formation of VOC. Perhaps this is to be covered under the PM review.

       On page 2-3, line  16, one should note they are referring to US emissions.  Lines 19-23 are
a bit confusing at present.  What is important, here, is that virtually all of the fuel-bound sulfur
gets oxidized to a volatile component (SO2 or  SO3),  and that there is virtually no sulfur in air, so
the sulfur emitted from burning a fuel is quantitatively related to that in the fuel.

       I particularly think that the section on Measurement Methods needs to be refocused. The
major question to be addressed here is if the current methods employed in the field provide
reliable measurements of SO2 for levels of interest, and this should be answered quantitatively.
At present, there is significant discussion of the various measurement approaches (a little is
needed) and lots of discussion on possible interferences, but never does one get the answer to
what is the typical uncertainty in the measurements at a typical monitor in the US. I suspect that
the method employed, while subject to some interferences, provides perfectly fine data, and that
the level  of uncertainty is such that we need not concern ourselves with possible interferences
and biases. Quantify the problems,  let the reader assess if they are of concern.

       Figure 2.4-5 "... in focus" in focus of what?  Actually this figure is not overly instructive
since most of the data is very much  at the bottom end, and it needs to include more information
(what years...).

       In considering emissions, it would be good to also provide  some information as to future
emissions for perspective. CAIR is going to significantly lower emissions in areas where they
are currently high.  This is important for our further consideration as to how  a standard might
impact air quality.

       What are the PRB levels of sulfate? What are the PRB levels of deposition? I  am not
sure this has to be added here, as part of the Sox primary document, but the section title suggests
it will be.

       Page 2-23, line 12: I think you mean months, not seasons,  (or does Philadelphia have
three summer seasons... which in some places might be nice, but having spent time in
Philadelphia in the summer, I am not sure it is  good there.)

       The Findings and Conclusions Chapter is still rather rough.  I would also focus on what
parts of the prior chapters significantly impact how the NAAQS may be revised. SO2 is a
slightly soluble gas, and the fraction that is oxidized in the aqueous phase is quite dependent
upon location. I am not sure what is meant by quantitatively on page 5-1, line 30. (Both lines 23
and 30 should be reworded.) Section 5.1.2 should also look to the future, given CAIR.  It is
interesting that the ISA says that it is inadequate for measurements at or below 3 ppb, but then
notes that the average on the West Coast is ~1 ppb.
Post meeting additions:

       While suggested above, I want to re-emphasize the need to better characterize the
potential for significant exposure measurement error with SO2. This is strongly suggested by the
Sarnat et al., results, but the is significantly more at issue than might be appreciated based upon
that work alone.  Exposure to high levels of SO2 will come predominantly from being in the
plume of a point source that burns coal or oil.  A direct mass balance suggests that a high
observation will be from a plume that is rather thin, thus impacting a rather small portion of an
area.  On the other hand, that plume, on a day that has a wind blowing at a slightly different
direction will miss the monitor, but impact a similar area, and likely  a similar population.
Chapter 2 needs to address this in more detail, and  Chapter 3 should  discuss the epidemiologic
studies in light of how well they have addressed this potential error, and also the co-exposures.
Further, if one looks at the possible concentration gradients in plumes,  one can see that a near-hit
can have a significantly different concentration than a direct hit.

       In an effort to help assess how to extrapolate the information we currently have to better
understand the distribution of pollutant concentrations at shorter averaging times,  it would be
good to provide, as the data allows, CDFs of 24 and 1  hr, and 5-minute, concentrations, and
correlations between 1-hr and max 5-minute average concentrations. I might also include a box-
and-whisker plot of 5 min concentrations, possibly by hour of day.

       To better assess the possibility of confounding, it would be good to provide information
as to how SO2 correlates with concentrations of other pollutants, including metals.

       In regards to the points brought forward to Chapter 5,1 might suggest that the bullets
need to be modified.  First, the route of SO2 oxidation is heterogeneous spatially.  I would not
call gas phase oxidation of secondary importance (I would also suggest that one try to use results
from a more detailed, US-specific model, e.g., CMAQ, if such an analysis exists).

       I take issue with the second bullet in 5.1.2:  The LOD may vary, and I suspect the various
QA tests support that the monitors in many locations are good down to 1  ppb. Don't be so
sweeping in the bullet.  Likewise, in the fourth bullet, I would suggest that there is little (not no)

       5.1.3 Needs to have a MAJOR bullet on the issues with exposure  measurement  error.  It
is potentially large, and needs more work throughout the document.  The bullet  should

specifically note that exposure measurement error is likely larger for SO2 than other pollutants.
Further, this is potentially larger as the reporting time gets shorter.

Comments from Dr. Schlesinger

Overall, this is a very well written document.

Chapter 3 is especially clear, in that each section has at the end a concise summary indicating the
overall conclusions from the studies discussed.

p. 1-1, line 24. Put")" and then "and" after ]

p. 1-4, line 11. The lab studies in the document are not all at or near ambient SOx levels.

p. 2-42, line 23. Add "and systemic" after "... respiratory tract"

p. 2-42, line 24. Add "irritant" after "epithelial"

p. 2-42, line 32. Add "respiratory  functional  parameters, e.g	" after layers;

p. 2-43, line. 20. "sooner" is not really a good scientific term to use here.

p. 2-43, line 24. Add "at ambient concentrations" after ".. .at rest." This makes it more
consistent with p. 2-45, line 5 that indicates that absorption  is related to concentration.

p. 3-55, line 15. Delete ".. .the collective evidence that..." and replace with ".. .any association

p. 3-55, line 16. Delete ".. .has an effect of..." and replace with ".. .and..."

p. 3-62, line 20. Define what is considered to be a "high concentration."

p. 3-64, line 21. What were the criteria for selecting "some" of the studies.

p. 3-68, line 20. The differences between the constituents should be elucidated.

p. 3-74, Section The potential confounding by copollutants has been presented in prior
sections of the chapter related to health outcomes so it is not clear why this section is needed. It
is somewhat redundant.

p. 3-75, Section 3.2.2. Material presented here has been discussed in earlier sections of the

p. 3-81, line 1. Add the word "...subtle changes in..." before cilia. Some changes could be seen
with light microscopy.

p. 3-84, line 17. Be more specific  about the aspects of respiratory health affected. For example,
the sentence could read, ".. .health, such as chronic bronchitis, asthma or respiratory symptoms."

p. 3-94, Section 3.4.1. Some of the results in thesestudies related to carcinogenesis should have
been discussed in Section 3.3.2.

p. 3-103, line 14. Delete "...sulfur agents..." and replace with "...paniculate sulfur oxides..."

p. 3-104, lines 3-4. This sentence does not really indicate what the association is with in terms of
health outcomes.

p. 3-104, line 9. After "...confounding..." add "...and lack of underlying biological

p. 4-6, line 14. Add "... measured..." after "... entire..." and then add "... over the measured
concentration range." After "... effect..."

p. 4-7, Section 4.2.1. This section is not very coherent. The discussion presented should be
melding within other relevant sections.

p. 4-8, Section Details of the studies presented here should have been described in
Chapter 3. Chapter 4 should ideally be an integration and synthesis chapter taking the material
from prior described studies to provide  a conclusion as to public health impact.

p. 4-12, line 17. Add ".. .or a general reduction in immune competence." After ".. .network..."

On page 4-12, lines 8-9, it is noted that two susceptible groups are children and infants. On page
4-14,  lines 7-11, it is noted that there is limited evidence that children are more  susceptible to
SO2.  This is somewhat of a contradiction and should be addressed. Similarly, in Section 4.3, it is
noted on page 4-16, lines 15-16, that exposure to SO2 is associated with various outcomes
particularly among asthmatic children. Again, there is the appearance of contradictions in the
discussion presented. This issue arises again on page 4-17, lines 24-25 and line  32 and page 4-
19, lines 11-18. There needs to be more consistency in the issue of susceptible populations.

p. 5-4, lines 17-18. It is stated here that the thoracic region is more sensitive than the upper
airways. However, there are receptors in the upper airways that could trigger an asthmatic attack.

p. 5-6, line 24. After "effects" add, depending upon the relationship between exposure
concentration and actual ambient levels."

p. 5-7. Section In this and other sections, the term "peak" appears  to refer to duration of
exposure rather than actual  concentration. There may be some confusion among readers since
many will relate peak to concentration and not time.

p. 5-9, lines 13-14. Change sentence as follows: ".. .biologic plausibility, but no concentration-
response information to allow a mechanistic understanding of..."

p. 5-9, line 32. "amount" should read "among"

p. 5-10, line 27. For consistency, the word "weak" should be replaced with "inconclusive."

p. 5-16, line 30. Fog droplets are usually large and therefore would not be carriers of SO2 to the
distal airways.

p. 5-17, line 15. After "... mixture..." add, "or other chemical component within it."

p. 5-17, line 21. After "metals" add "on particulate matter"

Chapter 5 should be retitled, Summary and Conclusions.

Comments from Dr. Seigneur
Chapter 1, Introduction

Charge question 3:  This chapter states that all sulfur oxides will be treated in the ISA. Yet, the
emphasis in the rest of the document is clearly on SC>2. EPA may want to consider whether it
may be more appropriate to mention in the Introduction that although all sulfur oxides must be
addressed, the ISA focuses on 862 because (1) sulfur oxides in the  atmosphere are mostly 862
and sulfate and (2) sulfate is treated under the PM NAAQS.

Chapter 2. Source to tissue dose

Charge question 1: The discussion of the atmospheric chemistry of sulfur oxides covers all the
important points but some revisions would help make the discussion more precise.

       Page 2-1, lines 24-25: Replace "partitions into the aqueous phase of particles" by "forms
       new particles by nucleation or condenses on existing particles" (note that under dry
       atmospheric conditions, H2SO4 may form non-aqueous ammonium sulfate or bisulfate).

       Page 2-2, lines 3 and 4: The same logic applies here, under dry atmospheric conditions,
       H2SO4 may not be transferred to an hydrometeor (since there may be none) but instead to
       a dry solid particle. Also, solubility in water is not the main reason why H2SO4 is
       transferred to the paniculate phase (HNO3, which is also very soluble in water, tends to
       remain mostly in the gas phase because it has a high vapor pressure); the very low vapor
       pressure of H2SO4 is the driving force for its rapid transfer from the gas phase to the
       particulate phase, regardless of the ambient relative humidity or ammonia concentrations.
       I suggest replacing the existing sentence by the following one: "Because H2SO4 has a
       very low saturation vapor pressure, it will be rapidly transferred from the gas phase to a
       condensed phase (particulate matter or droplet).

       Page 2-2, line 11:1 believe that manganese (Mn) is also  a catalyst for the oxidation of
       862 by 62 in aqueous solution. Is Cu an important catalyst compared to Fe and Mn?

       Page 2-2, line 14: Add Finlayson-Pitts and Pitts as one of the standard references for
       atmospheric chemistry.

       Page 2-2, line 24: Why a pH of 5.3 ([H+] = 5 x 10'6 M),  and not 5 or 5.5?

       Page 2-3, line 3: A sentence needs to be added to highlight the fact that the oxidation of
       862 by 63 and 62 is self-limiting, because as sulfate  is formed, the pH decreases and,
       consequently, the kinetics of those reactions decrease.

       Page 2-3, line 9:1 think that this example of (gas-phase) 862 oxidation in power plant
       plumes is misleading because 862 oxidation in power plant plumes varies from nearly
       zero to a value that may exceed the rate of oxidation  in the background air. Near the

       stack, the high NO concentration in the plume depletes the oxidant concentrations
       (including OH) and, therefore, there is almost no SO2 oxidation in the gas phase. Farther
       downwind, the oxidation of SO2 will pick up as the NO concentration is diluted with
       background air. Then, in a NOX limited environment for oxidant formation, the oxidant
       concentrations in the plume will exceed those in the background air and, therefore, the
       SO2 oxidation rate will exceed that of SO2 in the background. A range of 0.5 to 2% fails
       to reflect the complex evolution of the SO2 oxidation process in power plant plumes. I
       suggest stating that SO2 oxidation in background air is  on the order of 1% per hour; then,
       adding a sentence explaining the three stages of SO2  oxidation in power plant plumes.

Charge question 1:  Although it is stated in Section 1 that this  document addresses all sulfur
oxides, Section 2.2 on the sources of sulfur oxides does not address sources of SO3 and sulfate.
Although emissions of sulfate from industrial processes are typically a very small fraction of SOX
emissions, it would be useful to mention those  and to indicate the fraction of SOX emissions that
is sulfate for major source categories (e.g., coal-fired power  plants, diesel  engines).

Charge question 1:  Page 2-9: It would be useful to discuss briefly the issue of Sulfur Emission
Control Areas (SECAs), i.e., those areas where ocean-going ships will have to burn lower-sulfur
fuel to minimize their impacts on air quality inland.  Ship  emissions are a  major issue in ports
and along shipping lanes and as sulfur emissions from point  sources and on-land mobile sources
decrease, ship emissions take more relative importance.

Pages 2-10 and 2-11:1 think that CASTNet is now written CASTNET.

Page 2-11, line 1: Spell out CONUS as continental (or contiguous) United States.

Page 2-23, line 25: Why does SO2 peak during summertime in Los Angeles?

Charge question 2:  Page 2-27, line 24:  Since  it is stated in  Section 1 that this document
addresses all sulfur oxides, some discussion of the policy relevant background (PRB) for sulfate
should be added.

Chapter 5. Findings and conclusions.

Charge question 1: Page 5-1, line 30: delete "in cloud drops and/or in particles" because sulfate
is also formed in the gas phase. The intention was perhaps to state that sulfate ends up in drops or
in particles, then add "; sulfate is then transferred to droplets or particles due to its very low
saturation vapor pressure".

Page 5-2, lines 18-24: It is good news that better detection limits will be made available for
routine monitoring of SO2. Can EPA give an approximate timeline for the availability of those
new monitors in the routine network?

Page 5-15, lines 24 and 46: Did you mean <120 ppb and <600  ppb?

Annex 2.2. Chemistry of nitrogen oxides in  the troposphere

Is this annex needed in the Sulfur oxides ISA? If so, the comments that I provided earlier on the
Nitrogen oxides ISA apply here (in particular, revisions to Figure AX2.1-1).

Annex 2.3. Chemistry of sulfur oxides in the troposphere

Page AX2.24, lines 21-22: Although the solubility of H2SO4 is relevant for its removal by
droplets, it is not very relevant to its gas-to-particle conversion; its low saturation vapor pressure
needs to be invoked here.

Annex 2.4. Mechanisms for the aqueous formation of nitrate and sulfate

Page AX2.28, line 4: Include manganese as one of the catalysts of aqueous 862 oxidation.

Annex 2.7. Methods to calculate concentrations of nitrogen oxides in the atmosphere

Is this section needed in the Sulfur oxides ISA? Also, there should be a similar section on the
topic of modeling sulfur oxides, in particular SC>2.  A discussion of AERMOD is warranted.

Annex 2.9. Policy relevant background concentrations

Table AX2.6-1. Do those emissions include emissions from ships in coastal waters and/or Sulfur
Emission Control Areas (SECAs)?

Comments from Dr. Speizer

Pre-meeting Comments on: Integrated Science Assessment for Sulfur Oxides First External
Review Draft September 2007

Charge Questions (paraphrased):

1-3. Atmospheric Chemistry and air quality characterization appropriate and relevant for review
of primary SO2 NAAQS, properties of ambient SO2, and relevant special, temporal patterns and
exposure estimates.

Page 2-9 Figure 2.4.1. Better resolution of this figure is needed. In fact, it is not clear what state
summarized values mean.  Surely the emission inventories for a state like Ohio are  dominated by
point sources in contrast to a state like Montana or even Pennsylvania.  The text does a little
better job by suggesting that there are localized regions, but the Figure is misleading.
Subsequent figures are better.

Page 2-23, end of first para at lines 6-7 and contrast with remaining two paras.  This is a curious
statement, given the apparent inverse correlation seen in the figure 2.4.6c.  A similar inverse
correlation for Phili and the clear lack of association in Los Angeles and Riverside. . Since the
chemistry is likely to be different in the far west, this probably needs further discussion.

Page 2.28, Figure 2.5.1.  There must be something wrong here. 11 may be in the primary source
of the data.  If the figure is truly representative of "all age groups" (line 3) then 1.8% time spent
in Bar-Restaurant probably is a sampling error.  Also if-60% of people work outside the home
how can the Office-Factory be only 5.4%? This seems to me to be an example of uncritical
acceptance of data to make a point. I have not tried to evaluate the equations on the next two
pages since if the basic data set is wrong there is no point.

Page 2.31, section 2.5.2. Limitations are well described.

Page 2; .35 Iinel8-20.  This seems too dogmatic. Clearly at 250m the proximity to sources will
affect whether measurements made are an over or underestimate of exposure. Away from
sources this  might be a well mixed level, which would truly be representative of exposure.  Since
most urban (and in fact rural) monitoring sites that are not specified as being place for specific
emission control are not measuring point sources, the values measured are not likely to be
overestimates of exposure for people.  Seems sentence on page 2.37, lines 14-15  confirms this.

Page 2.42, Section 2.6 Dosimetry... Almost all reported work is very old. It could have been
summarized in much shorter space.

4-6. Integration of evidence on animal tox, human clinical and epi studies sound, balanced and
communicated. Is the assessment focused on policy-relevant studies or health findings?  Is the
discussion of strength, consistency,  coherence and plausibility of health effects adequate

       The major concern with this chapter (and the next) is that although in places mentioned is
made of the potential for studies to underestimate health effects in certain sensitive
subpopulations, results (particularly null effects reported) generally do not estimate the effects
that are present in the approximately 10-20 and sometimes up to 30 percent of the population
being studied who are the true responders (or sensitive subgroups). Their results are often
ignored when the summary statistics are given, and yet, they represent a substantial sub-segment
of the population who are responding adversely. Throughout these chapters this need to be
brought out better. It might be helpful to reverse the order of the presentation and show the
human controlled data first. Here it would be easier to bring out the point that there are sub-
segments of what are generally believe to be similar healthy people who respond differently.
This would point the argument for the epi studies to bring out the sub segment of these
populations. The studies discussed are in fact the appropriate ones.  The issues of strength,
consistency or lack thereof are brought out.  However, coherence and certainly plausibility are
not well discussed.  The end of each section that seems to indicate where the coherence is
lacking is often stated as an uncertainty of lack of ability of the writer to come to a conclusion.
The raising of the biologic plausibility simple as a statement at the end of a  section (see below) is
often totally unjustified. SO2 has been around a long time. There is both a large toxicological
and human study literature that goes back almost  50 years, without getting into the classified
literature that goes back even  further.  To raise the question of biological plausibility without
documenting why the writer thinks such, is simply disingenuous.

Page 3.7, Figure 3.1.1 I do not find this figure very compelling since almost all the estimates are
the same (as stated in the text)

Page 3.10, Figure 3.1.3  The graphics in this figure are not clear. The size of the population in
each study is reflected in the width of the confidence intervals.  But the size of the central
tendency estimator (although  stated in the description to represent some weighting) is not
consistent.  The Schwartz study is 300 kids , the Neas 98 yet dots are the same.  Also not clear
what lag 0-6 means.  The last  entry is either a sum of all the data and if so not clear the lag used
or and inappropriate and really unbelievable value for the 70 children in the Romieu study. Ditto
problem with Romieu study in Figure 3.1.4.

Page 3.25, line 14-16. This statement is too definitively negative. The limited data is true, but to
conclude that no effect because the tox data was studied at too high a level is premature. This
needs to soften to indicate that there are suggestive but limited data. See the way it is said at the
end of on airways hyperresponsiveness. The difference between the two sections is 5-10
ppm in animals in the former, and 5ppm in the later.

Page3.30, line 19-21.  Here again, the issue  seems to be that the animal tox studies simply
haven't been done beyond the one species of mice.

Page 3.45, summary of short term exposures. There is not sufficient discussion in this section of
the fact that a substantial fraction (15-20%)  of the population may be driving the effects noted in
normals. Even in those  studies in which the overall effects appear to be null there are these
"hyperresponders".  This may explain why in studies of diseased subjects the positive findings
occur more frequently or they appear to be more responsive. Needs more discussion.

Page 3.45 to end of section on Cardiovascular short term effects. Though mention is made of the
increased responsiveness in those with preexisting disease, the summary statements in each
section seems to ignore this phenomena (in spite of presenting what I would deem as relatively
consistent findings that those with preexisting disease are more responsive).  Suggest authors
look more closely as this phenomena.

Page 3.71 discussion of meta analysis. Not clear if authors are planning to redo this analysis.
The suggestion is that the wrong data were used.  Is this the comment made by the original
author or by the writer of this document?  In either case need to clarify what was used, and
indicate its limitations or change it.

Page 3.80 and section of short term mortality. The last line of the summary line 5-6 seems
inappropriate. The purpose of this section was to comment on the epidemiology. That done to
throw in "absence of strong biological plausibility" shows bias rather than good sense. The
previous 80 pages in the chapter provide several arguments that there is a biologic reason for
looking at SO2 as a potential putative pollutant that affects biological systems. Enough said, so
that I do not insult the author further.

7. Public Health Impact
Page 4.3, line 10-11. This is a repeat of an above theme. The Gong study (among others) clearly
identifies a sensitive sub-group even among a larger group that would all be considered
potentially sensitive. The public health impact is  clear that we need to quantify better who these
particularly sensitive groups are. This issue comes up several time in the chapter. Although the
tables define the likely groups with regard to respiratory conditions, it seems not complete.
Perhaps a couple of more summary tables with different parameters are necessary (E.g. Age, sex,
other preexisting disease).  I would think that this might be useful if population estimates are to
be made in subsequent risk assessment documents.

Page 5.2, line 29. I would have thought that a gradient should be going up.  Therefore should
this be an "west-to-east" gradient rather than as written?
Section—Agree with conclusions on short term peak exposures
Section  5.2..2.2—Agree with all except Inflammation.  This is too firm a negative as the issue
really hasn't been studied effectively.
Section—Short term cardiovascular effects—Stated as "inconclusive" is too
conservative. Need to bring into this section more that subjects or patients with preexisting
disease appear to be more responsive. This would make the conclusion more consistent with a
subsequent section ( short term effects on  mortality, since much of the mortality reported
is likely to be cardiovascular.
Section—other systems, agree.
Section—Long term effects on morbidity. Agree
Section—Long term effects on mortality.  Do not agree. The epi argument is fine.  The
lack of a plausible toxicological mechanism,  is because  of a lack of study rather than the actual
testing of any mechanisms.  Concentration-Response function—There is a distinction here that needs to be made.
Although there is a lack of data for a threshold the evidence that there is a concentration-

response function (above some undefined threshold level) is rather good.  In most of the studies
there is either an increase in response with increasing dose or a greater recruitment of "sensitive
subjects" with increasing dose.
5.2.29 Sensitive subgroups. This might be expanded somewhat. Although I would agree with
groups specified, I still have concern that within groups there are individuals who are more or
less sensitive. What this means is that for those subgroups who are deemed "inconclusively
responsive" or with "weak evidence" of responsiveness there are individuals who are truly and
unequivocally responsive but we are just not smart enough to subdivide the  groups adequately to
detect them and define them as a specific subgroup.  I guess this is why the  original framers of
the Clean Air Act thought it was a good idea to have "a margin of safety".

Comments from Dr. Wyzga

Chapter 2

Figure 2.4-1.  Since SO2 emissions will be reduced greatly in the near future, can the estimates
of future emissions be given in this figure?

P. 2-36,1 10:  since the slopes are negative, it would be preferable to use a word other than
predictor; e.g., significantly associated with.

Chapter 3

There are several studies that are not included in this chapter.  Some of them are quite recent and
may not have appeared after the draft of the ISA was written; others report negative results;
hence "SO2" may not have been included among the list of key words in literature searches. A
comprehensive review of the literature should include them, however. I cite the following:
Ho  et al.  (2007) Environmental Research 104:402-409.
Ko  et al.  (2007) Thorax: 62:779-784.
Ko  et al.  (2007) Clinical and Experimental Allergy 37:1312-1319.
Sinclair and Tolsma, (2004) J. Air & Waste Manage. Assoc. 54:1212-1218
Klemm et al, (2004), Inhal. Toxicol.  16: Supplement I, 131-141
Metzger et al. (2007) Epidemiology 18(5):585-592
Peel at al (2007) Am J Epidemiology, 165(6):625-633

p. 3-6,1.  12: replace "likely due  to" with "but there was".
      1. 22: insert at end "for CO and NO2, but not for PM10" if Figure 3.1-1 is correct.

p. 3-42:  it would be helpful to indicate the levels of SO2 in the various studies; it could provide
some clue as to why results are divergent.

p. 3-59: Why is Metzger et al. not included in this figure?

p. 3-59:  The conundrum is that the central eastern cities had higher levels of SO2.

Chapter 4

P. 4-4: Studies are cited which used the  default convergence criteria with GAM. This is clearly
stated, but the implications of this use should be noted.

p. 4-6,1. 8: It should also be noted that exposure measurement error can interfere with
estimation of thresholds.

Comments from Dr. Postlethwait
1.  In reading the document, especially immediately after the NOx ISA, it would appear import
   to address whether the epidemiologic cohorts overlapped in geographic locale and/or time
   with  other studies  which identified differing pollutants as "causal"  to  deleterious health
   outcomes. In other words, were the same people counted more than once?  Because of the
   apparent potential confounding from other pollutants in numerous studies, as noted within
   the ISA,  such an analysis might prove useful in selecting which studies provide the most
   robust support for direct causality of SOx induced health effects.

2.  It appeared that a number of the cited studies reported SO2 concentrations within ranges that
   Chapter 2 identified as potentially problematic in terms of measurement error. Consequently,
   there is either a  modest disconnect between stated  measurement  concerns and the health
   outcomes assessments or the ISA is simply accepting reported concentrations at face value.
   The ISA would benefit from this being clarified.

3.  This reader found some of the concluding statements among  Chapters 3-5 to lack  internal
   consistency with regard to effects levels between clinical and epidemiological studies.  In
   general, it appears that  ambient concentrations are approximately an order of magnitude (or
   more) less than levels that induce  observable  pathophysiologic effects during controlled
   studies.  Thus, while  controlled  studies  show reproducible effects at > 500  ppb, such levels
   are rarely attained  during environmental exposures.  Consequently, it would be useful to
   include, if possible, what short term peaks are attained within the US to provide a stronger
   basis for causal and biological plausibility  statements laid out in Chapter 5.  This becomes
   especially important  when  one considers the dosimetry  data  that  suggests little  SC>2
   penetrates to the  distal lung and  how reported outcomes  correlate with  estimates of
   intrapulmonary distribution.

4.  There appeared to be select suggestions regarding what types of additional studies are needed
   and when biological plausibility  appeared to be evident.  A more consistent format regarding
   additional studies  suggestions  and assessment of  the mode/mechanisms  of biological
   plausibility would strengthen the document.

5.  In general, it appeared that the overarching take  home message was at times suggesting that
   the majority of reported health  outcomes were  either substantially confounded due to co-
   pollutants or the lack of clearly observable effects.  However, in other portions of the ISA
   there  were  fairly  dogmatic  statements  regarding robust  associations  and biological
   plausibility. The document would be improved by minimizing such divergent conclusions.

6.  Because of the recognized  co-pollutant confounding, the very low ambient levels, and a
   relative paucity of mechanistic data, it is suggested that a short section be included in the end
   that identified key areas of research needed to reduce the levels  of uncertainty.

Comments from Dr. Gordon

Charge Questions

Question 4. The discussion and integration of evidence of sulfur dioxide health effects is quite
good in this ISA.  Unlike the recent first draft of the NOx ISA, the health effects chapter is of the
appropriate length and efficiently discusses the key studies and health effects in a manner that
leads to the obvious conclusion that respiratory hospital admission and ER visits, as well as acute
bronchoconstriction in asthmatics, are the adverse health effects of concern for sulfur dioxide.
The integration of the human clinical and epidemiology studies in the ISA, however, was much
better than the animal toxicology data.  Admittedly, there are sparse animal toxicology data at
concentrations relevant to ambient exposure levels, but Chapter 3 should be restricted to include
animal exposure studies of 1 (possibly 2) ppm sulfur dioxide or less.  Such levels of exposure are
appropriate for interpretation of and integration with epidemiology studies which evaluate
associations generally below 50 ppb or with 1 hr peak values higher than 100 or 200 ppb. The
high exposure levels used in the described extra-pulmonary studies (e.g., nervous system effects)
lend little to our understanding of the mechanisms by which sulfur dioxide causes adverse
respiratory or extra-pulmonary effects in susceptible people.  For communication purposes, it's
important to have an integrated analysis that draws key conclusions from the available
epidemiology and toxicology data sets and includes the magnitude of the concentration response
for the different health endpoints - this latter is the key to Chapters 3 and 5 but also the overall
quality of the IS A.

Question 6. While the ISA has drawn important conclusions regarding the robustness of the
respiratory epidemiology data and a lack robustness in the cardiac mortality data, the issue of
confounding co-pollutants deserves more attention and discussion. This is critical to the use of
the ISA in the Health Plan's risk assessment because the data suggest some studies on sulfur
dioxide's role in adverse respiratory effects stands up to inclusion of single co-pollutants while
others do not.  Given that the high concentrations of sulfur dioxide used in many of the reported
animal toxicology studies make plausibility interpretations quite difficult, it is even more critical
to clearly discuss the epidemiology findings given the potential confounding by co-pollutants.
Along that line, the discussion and integration of the adverse effects of multiple pollutants would
be aided by a more complete discussion of numerous  studies, particularly animal toxicology and
human clinical studies, which have investigated the interaction of particles with sulfur dioxide.
These studies are presented in the Annex Tables, but the ISA only mentions one human clinical
study (Koenig,  1983) whereas there are several animal studies that demonstrate that particles,
especially in the present of moisture, can enhance the effects of sulfur dioxide. These studies
may not help in setting benchmark values for health effects, but they will provide biologic
plausibility for the benchmark health effects which occur at very low sulfur dioxide

Question 7. The respiratory effects (hospital admissions, ER visits, acute bronchoconstriction)
have been correctly identified, discussed, and justified in the ISA.

Major Comments:

       As key chapters, Chapters 3, 4, and 5 are fairly well-balanced in integrating the exposure
and health data needed for risk assessment. The chapters are of appropriate length and detail and
do integrate the science of SOX's health effects without providing too much detail (a repeated
exception would be the inclusion of animal studies utilizing high (> 5 ppm) of sulfur dioxide).
Chapter 4 has fairly similar goals to other parts of the ISA (summarizing the adverse health
effects and discussing which concentrations and time frames are of concern), and EPA should
consider combining the chapters or more clearly delineating why topics are included in each
chapter. For example, a large part of Chapter 4 relates to a discussion of susceptible
subpopulations, information which could be included in Chapter 3 amongst the other health
effects discussions on atopies, asthmatics, and children/elderly.
       The bronchoconstrictive response of asthmatics to sulfur dioxide is fairly rapid.  EPA
should expand their data collection to include 15 min (or shorter) interval data for use by
epidemiologists to examine associations between shorter interval peak exposure levels and
adverse pulmonary endpoints in future studies.

Medium Comments:
       Throughout the ISA, the animal toxicology studies should be included only when they are
of relevant concentrations.

Minor Comments:

Chapter 2
page 2-3, line 31 - Stating that sulfur is bound to amino acids implies somewhat that amino acids
are grabbing or binding free sulfur rather than that sulfur is a key component of some essential
amino acids.
page 2-12 - An additional figure or table which presents the decreasing ambient concentration of
sulfur dioxide over the last 2 decades would be helpful. Although the text states the decline in
sulfur dioxide concentration over time, a clear graph showing trends in 1-hr, 24-hr, and annual
values would be more valuable than 2-yr trend example data for sulfur dioxide and sulfates in a
few different cities.
page 2-13 - Figure 2.4-5 needs a better explanation of what is being presented.  This same
comment applies to all figures in the ISA - the figure legend should allow the reader to fully
understand the figure without searching through the chapter's text.
page 2-13, line 1 - The use of 'aggregate' is unclear.  The use of 1-hr data to estimate daily or
annual values is important but what does 'aggregate up' mean?

Chapter 3
page 3-4, line 1 - Sulfur dioxide concentrations units are usually given as ppm, yet there is
inconsistent use of |ig/m3 included sporadically throughout this chapter - ppm values should be
page 3-10 - In the Figure legend, the 'size of the box of the central estimate' doesn't really add
much information when there is already a mean (95% CI) - it seems redundant with 95% CI.
Also, when the 95% CI is very large, the box becomes so small that it is hard to see the dash
representing the OR.
page 3-15, lines 22-26 - In the ISA, it is a good decision to not present every study that finds the
same result, but such sections would be more lucid if information was presented on why  some

studies deserve detail and others don't (i.e., better studies or equal in quality but the same
page 3-15, line 29 - Were the increases of 15 and 23% due to sulfur dioxide alone or sulfur
dioxide plus NaCl?
page 3-16, line 5 - typo - delete first 'of
page 3-19, lines 2-6 - Something is unclear in this section. If the pooled data from mild and
moderate/severe groups totaled 40, why does line 6 state 15/40 'moderate/severe subjects'?
page 3-19, lines 8 - 23 - This is an important point/study in understanding whether asthma
severity is linked to responsiveness to sulfur dioxide. It is somewhat unclear as the first sentence
states that the moderate/severe asthmatics had the greatest physiological and symptomatic
responses, then the paragraph points out various sides of the interpretation.
page 3-20, line 9 - typo? Change 22 to 24.
page 3-24, lines 1-13 - Why is this animal toxicology data included by itself when there is data
going back to the 1980's investigating vagal pathways in human subjects exposed to sulfur
dioxide.  Two  examples:

Myers DJ, Bigby BG, Calvayrac P, Sheppard D, Boushey HA. Interaction of cromolyn and a
muscarinic antagonist in inhibiting bronchial reactivity to sulfur dioxide and to eucapnic
hyperpnea alone. Am Rev Respir Dis. 1986 Jun;133(6):1154-8.

Yildirim Z, Kilic T, Koksal N, Kotuk M. 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. 2005

page 3-25, lines 4-13 - These animal studies using high concentrations of sulfur dioxide have
little relevance to the ISA.
page 3-26, line 16 - Unclear. Cut 'using Mch .. .relatively'?
page 3-27, last para - These animal studies using high concentrations of sulfur dioxide have little
relevance to the ISA.
page 3-28, line 15 -16 - Should this read 'sensitized to Ascaris' or challenged with Ascaris in
already sensitized sheep?
page 3-30, lines 24-25 - Pneumonia and bronchitis (acute) are not usually included collectively
under COPD, are they? Later, in Figure 3.2-3, data is given separately for COPD and
page 3-32 - Figure 3.1-7 - Again, the size of the box interferes with interpretation of the OR and
95%CI. A couple boxes are so big that they obscure the reader from seeing if the 95%CI are
above the 1.0 Relative Risk level.
page 3-34, lines 15-16 - Is it appropriate to give a risk factor per 10 ppb, when the range given
for sulfur dioxide annual means is only 0.9 to 4.8 ppb?
page 3-35, line 13 - ibid
page 3-36, Figure 3.1-9 - Typo? 25 degrees C. The legend says 2 different risk ranges (10 and
40 ppb) - should the differences be delineated in the figure?
page 3-37, Define NR in the figure
page 3-38, line 3 - Age should be stated for these results.

page 3-40, lines 13-17 - This conclusion statement should possibly be qualified a bit more.  The
sulfur dioxide RR does not appear to be robust in the Schwartz (1995) and Thompson (2001)
studies which have fairly tight RR/95% CI.
page 3-41- Define * in the figure.
page 3-45, lines 29-30 - HRV is predictive for survival after an MI, but is there strong evidence
for this statement as written?
page 3-48, line 24 - These animal studies using high concentrations of sulfur dioxide have little
relevance to the ISA (nor is it referenced).
page 3-54 - The figure has some shading that is not explained. Also, the ED visits and hospital
admissions data in this figure are not 'clearly distinguished' as stated on page 3-53, line 8.
page 3-60- This page describes high dose animal studies and should be summarized in a sentence
or two or cut out completely. Also, the final para has no exposure concentration or ref.
page 3-61 - Given the high  doses used in the animal studies, Table 3.1-1  should be eliminated.
The first para on this page may have high concentrations of sulfur dioxide (not given) and could
be cut as could the  Singh study in the last line of this page (32 ppm sulfur dioxide).
page 3-62 - ibid
page 3-81, lines 1-2 - This is not a chronic study and is a high concentration (10 ppm).
page 3-84, line 17 - 'respiratory health' would include Lung Function (next section), so this
statement is a bit too broad.
page 3-84, lines 1-2- This is not a chronic study and is a high concentration (5 ppm).
page 3-87, line 17 - typo? 65/72?
page 3-93, line 16 - Unclear - are the mg/day personal doses?
page 3-103,  lines 5-6 - Are  risk estimates of 1.02 and 1.04 really  different (i.e., 'smaller' as

Chapter 4

page 4-1, lines 30-32 - The sentence needs editing.
page 4-5, line 6 - It is puzzling why the Ponce de Leon study described here, thus giving it
weight/importance but it is one of many in a figure in Chapter 3 and only appears as a reference
in Chapter 3.
page 4-9, lines 13-16 - This sentence is speculative. Why wouldn't irritative effects be seen in
atopic children?
page 4-9, lines 30-31 - Typo? Repeated on next page.
page 4-10, lines 2-8 - Is the averaging time available for this paragraph?
page 4-10, line 32 - Typo? line ends abruptly.
page 4-12, line 26 - The inclusion of the Ponce de Leon study in the statement regarding the
association between sulfur dioxide and hospital admissions in children contradicts the statement
on page 3-34, line 18 stating other European studies did not find a significant association.
page 4-14, line 26 - Typo? 'expression' or?
page 4-15, line 31 - Define  TNF-1 allele - homozygous for the variant or wildtype?
page 4-18, line 8 - These percentages differ from those given on lines 1-2 on the next page.

Chapter 5

page 5-2, lines 18-24 - While the accurate monitoring of any pollutant is essential to
understanding its health effects, is there really a need to push the detection limit even lower?
Health effects appear to occur significantly above 1 ppb.
page 5-3, lies 14-31 - This is a very good section, but it would be improved if some additional
indoor/outdoor ratios ranges were given. The stated range of 0.03 to 1.01 implies this broadness
is frequently the case (the dosimetry section and table in Chapter 3 says otherwise).
page 5-8, lines 1-10 - This section and the supporting studies presented in Chapter 3 provide
'clear evidence for sulfur dioxide effects with peak exposure.' Therefore, it is not clear why a
peak exposure (5-15 min) Tier III evaluation was ruled out in the Health Assessment Plan
page 5-9, line 22 - Typo? 'symptoms' under Airway hyperresponsiveness subsection?
page 5-9, line 32 - Typo - among = 'amount'
page 5-14, lines 9-14 - This is one of a few possible explanations why population thresholds
may be obscured.  More importantly, why would this be labeled 'obscured'? A sensitive
subpopulation would bring up the lower end of the curve and they are truly responders.  I'm not
sure the shape of the line matters here as much as identifying the concentration that causes an
adverse effect.
page 5-16, lines 14-20 - Do these measurement uncertainties really complicate 'our ability to
attribute' if the measurement levels are significantly below observable effect levels?
page 5-16, line 25 - This statement is incorrect - human clinical studies do examine sensitive
subpopulations such as asthmatics.
page 5A-2 and 5A-3 - These are all high dose studies and should be cut.
page 5A-2 - define NR; Under the Mortimer  study, it states '0-75 ppb (shown in graph)' - what

Comments from Dr. Thurston

General Comments
       This document is an excellent first effort at reviewing and integrating the large body of
lexicological, clinical and epidemiological knowledge regarding the human health effects of
sulfur oxides (SOx).  I especially like the evidence evaluation approach outlined on pages 5- and
5-7.  It makes a compelling case for the need for a new short-term SC*2 standard (of 1 hour or
less averaging time) in order to sufficiently protect public health.  However, the document has
some shortcomings in several regards, as follows.
       First, although the document states on pages 1-1  and 1-2 that "the possible influence of
other atmospheric pollutants on the interpretation of the  role of SO2 in health effects studies is
considered, including interactions of SO2 with other pollutants that co-occur in the environment
(e.g., nitrogen oxides, carbon monoxide [CO], ozone [O3], particulate matter [PM])", this is not
yet sufficiently  accomplished in this document.   In particular, the document does not yet
sufficiently address the interactions of SOx and PM, though much of the information necessary
to do so is already included in the document. As discussed at our July meeting, this document
should comprehensively consider the SOx-PM interaction, considering the potential for the
potentiation of each pollutant by the co-presence of the other. Yet this is not drawn out in the
document,  and  the PM  is  only really  considered as a possible confounder (i.e., a problem in
discerning  SOx  associations) in this document, rather than as a potential  facilitator, of SOx
effects. For example,  in the  dosimetry  section, no mention is made (on 2-43) of Dr. Mary
Amdur's work showing that the respiratory effects of SOx on animals are greatly enhanced by
the co-presence  of particles,  presumably because particles can absorb  the gas  and act as an
"vector" for the  SOx, allowing it to bypass absorption in the upper airways.  (One biologically
plausible mechanism might be that the adsorption of SOx could be making the metals in particles
more acidic and more bioavailable, and thus the SOx makes the particles more toxic.)  Similarly,
the clinical studies and  epidemiology need to consider the potential for a SOx-PM interaction
more consistently.
       Throughout the  document there  is some evidence consistent with the PM interaction
influence on SOx toxicity, but it is not really considered collectively: one example with regard to
clinical studies  appears at the top  of page  3-16, where it says "One  human clinical study
provided evidence that during exercise, peak exposures (10 min) to  SO2 at concentrations of as
low as 0.5 ppm in the presence of hygroscopic particles that can carry SO2 deeper into the lung
can elicit significant changes  in pulmonary  function  in asthmatic adolescents. ".  Also, with
regard  to the London Fog Episode on page 3-63, it says:  "the 1982 AQCD could not resolve the
relative roles of these two pollutants and suggested that the clearest mortality associations were
seen when both pollutants were at high levels ". Thus,  this interaction is touched upon here and
there,  but needs to  be  organized and brought together, and thereby  considered in a more
"holistic" way.   Indeed, these issues need to be handled comprehensively in both the SOx and
NOx documents  Overall,  while there are smatterings of references (here and there) to PM-SOx
interactions as  an possible "confounder" in various passages, I  see PM as the insufficiently
addressed "elephant  in  the room"  of each of these  two new gaseous  pollutant  assessment
documents.  Interestingly, this factor is relied upon in  making conclusions in Chapter 5 (at the
bottom of pages 5-16 and 5-17), but the  support for  this important  point is  not well enough
developed in the prior chapters.  I recommend that the NOX and SOx documents both address
this gas-particle interaction issue more directly and comprehensively.

       Second, although the introduction to the document states about Chapter 3  (on page 1-5)
that " The focus of this chapter is on the strength of underlying epidemiological or toxicological
evidence and the coherence and  plausibility  of the  body  of evidence  for  effects  on the
respiratory,  cardiovascular,  or other system.", these criteria  for the  evaluation of the health
effects considered are not sufficiently set out at the start of Chapter 3,  nor are  these criteria
consistently applied  throughout the document.  Presumably these  choices  are based upon Sir
A.B. Hill's 1965 treatise on causality, but it should be explicitly referenced, and the rationale for
the selection of these specific criteria from Hill's longer list, and how they  will be applied, needs
discussion here.  I feel that  a more consistent application  of the A.B. Hill criteria across the
various sections,  especially  as a function of pollutant averaging time and  concentration when
possible, would enhance the value and usefulness of the SOx document.
       Third,  and related to  the above discussions, the use  of epidemiological results  with
multiple pollutants  considered  simultaneously  is not  useful, yet it  is  done  throughout the
epidemiology  sections.    The problem is  that including correlated pollutants  simultaneously
makes the individual pollutants' coefficient  estimates biased and largely uninterpretable.  As
stated  on page  35  of the  EPA's companion  SOx Scope and Methods  document:  "When
collinearity  exists, inclusion  of multiple  pollutants in models often produces  unstable and
statistically insignificant effect estimates for both SC>2 and the co-pollutants." Thus, one should
consider at most  two pollutants at  a time, and even then only as a sensitivity analyses, not as
useful  estimates of the individual pollutants effect estimates  or their  significances.  In multi-
pollutant models, only the linear combination of the pollutant effects is an unbiased estimate (not
the individual coefficients), so a more useful  approach might be to look at the total effect of all
pollutants in such models, and see if that overall estimate is increased by the addition of another
pollutant in  order to  evaluate if some additional information is provided or not.  At a minimum,
the epidemiology tables  should remove all models  considering more  than  two pollutants at a
time.  In addition, the interpretation of multi-pollutant models should consider the potential for
pollutant interactions  (e.g.,   possible potentiation  of effects)  as  possible  explanations for
variations across models, studies,  and  locales,  rather than merely dismissing pollutant terms
affected by the inclusion of other pollutants as indicative of statistical confounding. The effects
of co-pollutants on the SOx  term may reflect real biological  interactions of effects,  especially
between PM and SOx, and this should be considered as a  possibility, in addition to statistical
       Fourth, units  are not yet consistent throughout the document.  Sometimes it refers to ppm,
sometimes to ppb, and sometimes even ug/m3 (e.g., see top of page 3-12).  Personally,  I prefer
ppb, but some consistent concentration metric should be used throughout to ease cross-discipline
and cross-study comparisons.

Specific Comments

Chapter 1
Page 1-1, lines 23,24: This sentence is muddled, seeming to include particulate sulfates among
gaseous SOx.  Clarify.
Pg. 1-3, line 8: Add a comma after "possible".
Pg. 1-3, Iine31: Change "systems" to "capabilities". The U.S. and Canadian health care systems
are not the same.

Page 1-4, last paragraph.  Fix grammar.  Each consideration is written as a question, but there is
no  question  mark.    I suggest not  having them as questions,  changing  each numbered
consideration to start "whether" instead of "were" or "are", and fixing each from there.
Chapter 2.
Pg. 2-3, line 18.  Note that nearly 90% of the power plant emissions of SOx are from coal fired
power plants.
Pg. 2-8.  line 29.  Note that, as shown in Figure 2.4-1, most of this improvement in SOx
emissions was made prior to 1995, and that progress has slowed in the last decade.
Pg. 2-13.  Also present the distribution in the cumulative frequency of occurrences of 5-minute
average SC>2 concentrations exceeding given levels (e.g., %>50 ppb,  %>100 ppb, %>250 ppb,
etc.) in the U.S. data collected in the past 10 years (1997-2006) at some 300-400 ambient sites,
as discussed in the draft SOx Scope and methods for Exposure and Risk Assessment document
(Nov. 2007).
Pg. 2-24, last par.: Published evidence documenting intercontinental transport should be added.
For example, from China to the Eastern U.S., see:  Lall,  R and  Thurston G. Identifying and
quantifying transported vs. local sources of New York City PM2.5 fine particulate matter air
pollution. Atmospheric Environment 40 (2006) S333-S346.
Pg. 2-43. last par.: Add discussion of Dr. M. Amdur's work showing the copresence of particles
increases penetration and effects  of SOx.  This will provide support needed for statements made
on page 5-16, line 30.
Chapter 3.
Page 3-1:  Add discussion of A.B. Hill's Criteria for causality and lay out the criteria to be used
to evaluate the  various studies and to evaluate causality.  This should be consistent  with the
system utilized in Chapter 5 (pg.  5-6), and be applied and considered throughout Chapter 3. If a
section or a study doesn't help evaluate one of these criteria, eliminate it.  Add studies only if
they are needed to address a criteria not yet considered in each section of this chapter.
Page 3-2, line 10: Change "have little ability" to "cannot definitively"
Page 3-2, lines 14-15. Change the words "and serves as an important tool in addressing the issue
of confounding by copollutants."  to read: ", and may provide some insights  into the potential for
confounding or interactions among pollutants."
Pg. 3-2, line 26.  Change "multiple model" to read: "multiple pollutant models
Pg. 3-3, line 1. Start new par with: "While clinical...", and drop "do in fact"
Pg. 3-3,  line 17.  Add summary sentence noting that, given their respective  strengths and
limitations, all the  different types of lexicological, clinical, and epidemiological studies are
needed in order to evaluate the causality of SOx-health effects associations with confidence.
Pg. 3-16, lines 4-7.  This study is quite important, given the discussions on pages 5-16 and 5-17.
This needs further discussion as to how it fits with toxicological studies  that have  shown greater
effects with the  co-presence of  particles (e.g., Amdur, et al.), and how this  shows coherence
across  disciplines.   It may also  account for why epidemiology studies show effects  at lower
levels than controlled SC>2 studies, as PM is always also present in the ambient environment.
Pg 3-19, line 11, change "response with" to "response to"
Pg. 3-23, 2nd par. Note the Koenig study showing effects at lower levels with PM co-exposure,
and  highlight effect of exercise.   Both  of these factors  important to understanding the
epidemiology, and why it might  see associations at lower levels than  most  controlled pure SO2
Pg 3-26, lines 10,11:  Use ppb, not ug/m3.

Pg. 3-35, lines 3,4,5.  Eliminate sentence.  Such many pollutant models are uninterpretable, and
should not be cited.
Pg. 3-40, line 4.  Expand this paragraph to consider biological interactions between SO2 and PM
that may be causing the multi-pollutant model results, not just statistical confounding.
Page  3-40.  lines  29-32.  Note  that the warmer months  are generally  associated with higher
activity levels by children, and more acidic forms of SOx, so these epi results are consistent with
the toxicology and clinical results.
Page 3-44,  line 2.  What averaging time for the 40 ppb? How about for the other values in this
Pg. 3-47, line 17. Beta blockers not a factor in this study?
Pg. 3-56, line 9.  Use ppb, not ug/m3.
Pg. 3-58, line 5.  Consider interactions in this paragraph, too, especially for PM. There is a need
for evaluation of coefficients as a function of PM levels, not just whether the coefficients change
in simultaneous  multi-pollutant  models.  More advanced and comprehensive approaches are
needed to sort this  out than considered here.
Pg. 3-58, line 19. Change "likely confounded" to "may be confounded"
Pg. 3-59, Table3.1-13. Replace all models considering more than two pollutants at a time with
two pollutant models when  available.  Are  PM, SO2 two pollutant models available for most
studies? If so, consider making a table of SO2 alone vs. SO2 with PM.
Pg. 3-67, line 9.  change "confounded by" to "confounded with".
Pg. 3-74, line 9.  Change title and text to include both confounding and interactions. Note where
epi evidence supports or refutes and interaction between sulfur dioxide and PM (e.g., where
single pollutant model SO2 effect larger in places with higher PM).
Pg 3-75, lines 17-18.  Change to read: "... .and Europe generally suggest that SO2 mortality risk
estimates may be confounded by co-pollutants, making a definitive distribution of effects among
the pollutants difficult."
Pg. 3-80, line 4.  Add comma after "is suggestive"
Pg. 3-80, line 5.  Add discussion of possible interactions between SO2 and PM,a nd evidence for
or against within the epi literature.  New  tables and/or analyses of the results from each study
may be needed to add this perspective on the literature.
Pg. 3-87.  Add discussion of the ACS study (pope et al,  2002) vis-a-vis SC>2 and lung cancer
Pg. 3-96, line 23.   Elaborate on the fact that  this lack of specificity by  SC>2 undermines its
credibility as causal for long-term mortality in this study. PM is specific,  SC>2 is not.
Pg. 3-101,  line 23.  Good point.  Also, note that the Harvard 6-Cities study has a lower %
College Educated (closer to  the overall U.S.), and gets higher PM RR estimate, consistent with
this conclusion.
Chapter 4.
Pg. 4-3, line 11.  Note Koenig study showing effects at lower levels in co-presence of PM, as in
the real world case. Also, note that exercise lowers threshold for effects.
Pg. 4-5. Note that the co-presence of PM and its varying interactions with  SC>2 may  also affect
ability to detect a threshold in epi studies.
Pg. 4-7, line 22. Note also  that children are active, and exercise lowers the threshold for SC>2
effects in clinical studies, so  that is another way in which children are more at risk.
Pg. 4-10, lines 21-26. Rewrite with respect to the criteria to be outlined in  the beginning of
Chapter 3, based upon A.B. Hill's criteria for causality.

Pg. 4-16, line 10.  Change "only very limited" to "insufficient".
Chapter 5.
Page 5-1, line 23. Change "coal and oil" to "coal"
Page 5-1, line 27 from "substantially since 1990" to "substantially since 1990, but progress has
slowed in the last decade".
Pg. 5-14, line 16.  Note the increased susceptibility resulting from exercise and co-presence of
particles in this section.
Pg. 5-15, line 12-13. Change "only very limited" to "insufficient".
Pg. 5-15, line 24.  How high is >120? Give actual value.
Pg. 5-16, lines 28-31.  Good discussion, but this needs better documenting in the body  of the
document (e.g., Chapters 2 and 3).
Pg. 5-17,  lines  16-21.   Good discussion, but needs better documenting in  the  body of the
document.  Especially for the effects of SOX on particle bioavailability of transition metals.
Page 5-18.   line 1.  Change "will result in decreased" to "will be associated with a decrease in

Comments from Dr. Hattis

Integrated Science Assessment

2. Spaciotemporal patterns and exposures, policy-relevant background.

       I think the document could have gone a little farther in analyzing the data in Table 2.4.2
on SO2 concentration distributions observed by existing monitors in CSMA's for different
averaging times. Figure 1 shows lognormal plots of the data in this table.  From the
correspondence of the data points to the fitted straight lines, it can be seen that particularly for
the shorter averaging times, the data are well described by lognormal distributions. In the fitted
regression line the intercept is an estimate of the logarithm (base 10) of the geometric mean and
the slope is an estimate of the logarithm of the geometric standard deviation.  For example, the
estimated geometric mean for the maximum 1 hour daily averages of the readings from CSMA
monitors is 100.806 = 6.4 ppb and the estimated geometric standard deviation is 100.524 —about
3.34.  These results allow us to make at least some quantitative estimates of the likely frequency
of ambient outdoor exposures at levels associated with various incidences  of short term
responses to SO2 in populations that have been  studied in clinical settings. I will use these
results for my response to the charge question (#7) on likely public  health  impacts below.

4. Integration of evidence on health effects from toxicological human  clinical and epi
studies technically sound, balanced, and clear?

5. Integration of health evidence focus on the most policy-relevant studies and health

6. Conclusions drawn on the strength, consistency, coherence and plausibility of health

       Generally I agree with the somewhat skeptical treatment of the epidemiological  results as
likely to be confounded with effects of particles. I think this is particularly likely because
although we know how to measure SO2 gas pretty well, we don't know exactly what the most
causally related components of particles really are.  Because SO2 and fine particle levels are
correlated, this makes it quite likely that existing regression related  findings are attributing some
of the mortality and other responses causally related to some particle fractions (size distributions,
composition) erroneously to SO2.

       Particularly in the light of this, I would have preferred a more quantitative treatment of
the issue of human variability in the undoubted causally related responses  observed from clinical
exposures to SO2. I am particularly intrigued by the possibility of a more  quantitative analysis
of the individual subject response data of Horstman et al. (1986) reproduced in Figure 3.1-6, and
any other similar data sets.

       For the analysis of human variability in Figure 2 below I have extracted the individual
Horstman et al. data as best I could from the figure provided in the ISA and the accessible
abstract (I could not easily obtain the original paper). Figure 2 is based on the conventional
assumption for probit analysis that in the population of asthmatics studied  there is a lognormal

distribution of individual thresholds for the response (a doubling of airway resistance during
exercise). In this case the intercept is an estimate of the log of the SO2 level needed to elicit the
response in the median asthmatic (100.0189 = 1.044 ppm = 1044 ppb) and the slope is an
estimate of the log of the geometric standard deviation [the Log(GSD) in our terminology] of
individual response thresholds (100.374 = 2.37).

       In previous efforts my colleagues and I have compiled a substantial database of
information on human interindividual variability for a variety of responses (see the website at
http://www2.clarku.edu/faculty/dhattis). The log(GSD) of about 0.37 in this case is not at all
unusually large-it is actually toward the lower end of observations of variability in responses to
acute inhalation exposures compiled in our data base  (Table 1) (however, it can be seen that in
many of these cases with larger variability the agents  act via specific receptors or via allergic
processes that may well in general be subject to more variability than responses to nonspecific

       Given the variability analysis in Figure 2, it is straightforward to make at least a tentative
projection of the likely incidence of responses for asthmatics similar to those studied by
Horstman et al. (1986) at any air level, assuming that the population distribution of response
thresholds is in fact perfectly lognormal:

                                      expected incidence of response (% of days
                                    expected to cause  100% increase in specific air
                                   way resistance for exercising asthmatics, ignoring
                                     the exposure duration difference between 10
                                    minute studied exposure  and 1 hour duration for
                       ppb         the greatest 1 hour average in a 24 hour period)
                        10                          3.4E-08
                        20                          2.2E-06
                        30                          1.9E-05
                        40                          7.7E-05
                        50                          2.1E-04
                        100                         3.2E-03
                        150                         0.012
                       200                         0.028
                       400                         0.13
                       600                         0.26
                       800                         0.38
                       1000                         0.48
7. Public health impact and characterization of susceptible/vulnerable groups.

       Given the analyses presented earlier in my responses to charge questions 2, and 4-6, and

       * Assuming that both the exposure distribution and the distribution of individual
       thresholds for response in asthmatics are perfectly lognormal,

       * Ignoring for now the exposure duration and intake difference between the one-hour
       exposures measured by the monitors and the 10 minute exposures used to measure effects
       in the exercising asthmatics studied by Horstman et al., and

       * Neglecting any systematic differences there are likely to be between individual personal
       exposures and air concentrations measured in the elevated outdoor compliance monitors

we can derive an estimate of the overall fraction of days that asthmatics similar to those in the
studied group. We do this by cutting the assumed lognormal distribution of air concentrations
from 0 to 1000 ppb into intervals of 1 ppb, calculating the number of asthmatic people who
might be in each interval, and summing up the number likely to respond during the maximum
hour's exposure on each day (Table 2).  Overall the fraction of asthmatic-days expected to elicit a
response of the severity recorded by Horstman et  al. (1986) is about 2.9 per 10,000.
Interestingly, half of the total response incidence is attributable to very rare high exposures (over
about 230 ppb).  This results from the larger estimate of variability in exposures, compared to the
estimate of variability in human response thresholds.
8. Adequacy of first draft to provide support for future risk, exposure and policy

There are useful data here that can support future risk, exposure and policy assessments. As
illustrated by the tables and figures I have constructed for the responses to previous charge
questions, I think the current document can be taken further in quantitative analysis of exposure
levels, and the extent of variability in individual humans' susceptibilities/distribution of response
thresholds. I also think that the authors should make some more ambitious attempt to draw
quantitative conclusions via meta-analytic combination of the very best multipolutant studies of
effects. This will inevitably still be confounded by the fact that we don't have a good idea what
the appropriate causally relevant metric is for the particle exposures (in terms of size fractions
and composition).  Absent this, the fact that it is much simpler to measure SO2 levels, and the
fact that SO2 and fine particle levels are correlated means that multiple regression analyses are
very likely to attribute some of the effect caused by particles to SO2.  Getting a quantitative
handle on how large this confounding really is will be very challenging with existing

                           Figure 1
        Lognormal Plots of Data from Table 2.4.2—Distributions

        of S02 Concentrations (ppb) for Different Averaging Times




                         y = 0.806 + 0.524x RA2 = 0.990

                         y = 0.306 + 0.534x RA2 = 0.998

                         y = 0.442+ 0.418x RA2 = 0.971

                         y = 0.457+ 0.325x RA2 = 0.
1 hr max

1 hr ave

24 hr ave

1 yr ave

                         Figure 2
Lognormal Plot of the Distribution of Individual Sensititivities

(SO2 Concentrations Needed to Double Specific Airway
Resistance) For 27 Exercising Asthmatics (Horstman et al. 1986)

               v = 0.0189+ 0.374x RA2 = 0.934

                                                               Table 1
       Previous Observations of Human Interindividual Variability in Local Lung Function Responses to Inhaled Agents
response studied
Air Cone. Needed to cause 10%, 15%, and 20% decrease
Air Cone. Needed to cause 10%, 15%, and 20% decrease
FEV1 change in relation to CXT of ozone exposure
FEV1 Increase by Antiasthmatic
PD20— concentration needed for 20% increase in
individual baseline value of FEV1
PD20— concentration needed for 20% increase in
individual baseline value of FEV1
Specific Airway Resistance PC50— concentration needed
for 50% increase in individual baseline value
Specific Airway Resistance PC50— concentration needed
for 50% increase in individual baseline value
Specific Airway Resistence-concentration needed for
20% increase in individual baseline value
Specific Airway Resistence— concentration needed for
100% increase in individual baseline value
Specific Airway Resistence-concentration needed for
15% increase in individual baseline value, mean of 2
trials with and without ozone
Specific Airway Resistence— concentration needed for
15% increase in individual baseline value, mean of 2
trials with and without ozone
Specific Airway Resistence— concentration needed for
20% increase in individual baseline value
Specific Airway Resistence-concentration needed for
20% increase in individual baseline value
Specific Airway Resistence— concentration needed for
20% increase in individual baseline value
Specific Airway Resistence-concentration needed for
20% increase in individual baseline value
Specific Airway Resistence— concentration needed for
20% increase in individual baseline value
population studied
Females-general population
Males— general population
Experimental subjects
Atopic subjects
Atopic subjects
Bakers— occupationally
5733 smokers with mild to
moderate airflow obstruction
Healthy athletic adults, 18-50
Allergic asthmatic patients
Allergic asthmatic patients
9 year old New Zealand
Allergic asthmatic patients
General adult population,
Norwegian community, Age
Nonsmoking adults with mild
Nonsmoking adults with mild
Ragweed allergen
Wheat flour dust
Wheat flour
Grass allergen
Ragweed allergen
data source
Paoletti et al, 1995
Paoletti et al, 1995
McDonnell et al. 1995 analyzed in
Hattis 1998
Lip worth 1992, analyzed in Hattis
Meerschaert, 1999
Meerschaert, 1999
Merget, 1997
Merget, 1997
Tashkinetal., 1996
Balmesetal., 1997
Hanania, 1998
Hanania, 1998
Sears etal, 1996
Hanania, 1998
Bakke, 1991
Evans, 1996
Evans, 1996
Source: Human interindividual variability database, updated as of 5/05, available "http://www2.clarku.edu/faculty/dhattis" discussed in

Hattis, D. "Distributional Analyses for Children's Inhalation Risk Assessments." Journal of Toxicology and Environmental Health, 71:1-9, 2008 in press.
Hattis, D. and Lynch, M. K. "Empirically Observed Distributions of Pharmacokinetic and Pharmacodynamic Variability in Humans-Implications for the Derivation of Single Point
Component Uncertainty Factors Providing Equivalent Protection as Existing RiDs." In Toxicokinetics in Risk Assessment, J. C. Lipscomb and E. V. Ohanian, eds., Informa
Healthcare USA, Inc., 2007, pp. 69-93.
Hattis, D., Baird, S., and Goble, R. "A Straw Man Proposal for a Quantitative Definition of the RfD," Drug and Chemical Toxicology, Vol. 25, pp. 403-436, (2002).

                                        Table 2
  Illustrative Calculation of the Expected Fraction of Days on Which Exercising
Asthmatics Might Experience a Doubling of Specific Airway Resistance, Subject to
                           Extensive Assumptions (see text)
   Upper end of conctration      overall contribution to fraction of days with    cumulative total fraction of days
       interval (ppb)                    response in interval                     with response
            10                              1.6E-09                             1.6E-09
            20                              9.0E-08                             9.1E-08
            30                              5.4E-07                             6.3E-07
            40                              1.4E-06                             2.1E-06
            50                              2.6E-06                             4.7E-06
            60                              3.9E-06                             8.6E-06
            70                              5.1E-06                             1.4E-05
            80                              6.2E-06                             2.0E-05
            90                              7.0E-06                             2.7E-05
            100                              7.7E-06                             3.5E-05
            110                              8.2E-06                             4.3E-05
            120                              8.5E-06                             5.1E-05
            130                              8.7E-06                             6.0E-05
            140                              8.8E-06                             6.9E-05
            150                              8.8E-06                             7.8E-05
            160                              8.7E-06                             8.6E-05
            170                              8.6E-06                             9.5E-05
            180                              8.4E-06                             l.OE-04
            190                              8.2E-06                             1.1E-04
           200                              7.9E-06                             1.2E-04
           300                              6.4E-05                             1.8E-04
           400                              4.0E-05                             2.2E-04
           500                              2.4E-05                             2.5E-04
           600                              1.5E-05                             2.6E-04
           700                              9.7E-06                             2.7E-04
           800                              6.4E-06                             2.8E-04
           900                              4.3E-06                             2.8E-04
           1000                             3.0E-06                             2.9E-04

Comments from Dr. Samet

 General Comments
The draft ISA for SOX follows the model used for the NOX ISA and consequently many of
the general concerns that I expressed about the earlier ISA are applicable to the SOX ISA.
In my general comments on the NOX ISA, I noted the failure to adequately describe the
methodology for the review, the lack of transparent criteria for evidence evaluation, and
the incomplete development of a causal framework for interpreting the findings of
epidemiological studies. Interpretation of regression estimates for "SOx effects" is
particularly problematic.  This and other problems noted for the NOX ISA are again
evident and equally limiting; I attach my earlier comments on the NOX ISA.
On reading this ISA, I was particularly concerned by the failure to explicitly describe the
approach for evidence evaluation from the outset. The authors' conclusions reached on
the various health outcomes are summarized in the attached table; these were largely
expressed in summary paragraphs at the end  of subsections. The depth  of analysis was
limited and terminology is not uniform. Toxicological information appears to have been
variably considered in making judgments on the strength of evidence.
These evident limitations  of the ISA appear to reflect inadequate development of the
overall approach to preparing the ISAs.  In response to a request for a protocol for
carrying out the ISAs, I was informed that there was not a formal protocol beyond the
brief procedural charges transmitted by Marcus Peacock in memoranda dated December
7, 2006 and April 17, 2007.  In my opinion, a systematic review process should not be
implemented, absent a formal and documented protocol that describes the approach for
evidence gathering, evaluation, summarization, and interpretation, including uniform
criteria and language for describing the strength of the available evidence.  This
deficiency of the process needs immediate discussion.
The Chapter 5 summary of the evidence and  judgment as to the strength of evidence for
causation is not well grounded in the review  offered in Chapter 3. I concur with the
judgment as  to the  causal  nature of the short-term effect on lung function, derived largely
from experimental  findings in human clinical studies. The ISA uses the terms "consistent
and robust" in referring to the findings on respiratory health, phrases that are not well
supported by the judgments  made in Chapter 3 (see summary table).

Comments on Chapter 5 and responses to  charge questions 5 and 6
General Comments on Chapter 5: Chapter 5 presents a bulleted summary of the
findings of previous chapters, noting through this display, the advances in evidence since
the prior reviews of SOX.  The approach to providing the updates is succinct, although the
methodology for determining these advances lies on the more opaque approaches of the
prior chapters. Of  concern is the methodology for integrating the health findings, which
as noted, is set out  only briefly in Chapter 5.  The focus is primarily on the "positive"
findings in a large and difficult body of evidence. While criteria such as coherence and
consistency are mentioned, I note a failure to be truly integrated. Rather, the focus
immediately settles on the findings of the human  exposure studies with  regard to short-
term effects of experimental exposure on lung function, and selected positive findings in
regard to respiratory health.  The latter do not receive adequate interpretation.

Charge Question 5: Potentially, the clinical studies are policy relevant in certain
settings, specifically those where short-term exposures comparable to the concentrations
used in the exposure studies might occur. The groups at risk would include exercising
persons as well as those with asthma, particularly if exercising. This is likely a relatively
infrequent exposure scenario (to be addressed in the risk assessment). The ISA does not
adequately consider the potential concentration-response relationships for this short-term
outcome.  Given understanding of the dosimetry of SOX in the respiratory tract, what
doses would be anticipated at short-term peaks likely to be experienced at present?
The respiratory health outcomes are policy relevant and much of the literature relates to
exposures that would be encountered in the general population setting.  In this regard, the
findings are more relevant to the present NAAQS.

Charge Question 6: The conclusions drawn in the draft ISA are incompletely grounded
in considerations around strength, consistency, coherence, and plausibility.  The draft
ISA is neither sufficiently comprehensive nor thoughtful in its application of these
criteria. The criterion of strength refers to the magnitude of the effect. With regard to
current levels of ambient SOX, "strong" effects would not be anticipated and "weak"
effects are far more plausible.  The document has no explicit criteria for consistency and
plausibility is variably addressed in the various integrative sections. As I note above,
criteria for evidence evaluation are  offered, only briefly, in Chapter 5 and there appears to
be little consistency in Chapter 3 in reaching judgments.

Response to Charge Question 8: Based on my review of this first external draft ISA, I
find it to be deficient, perhaps largely reflecting an inadequately developed  synthesis
methodology.  I am doubtful that the document will become adequate without
development of a more explicit protocol and rigid adherence to it.

Table Summary Conclusions Concerning Various Health Outcomes in the SOx ISA
Respiratory Outcomes/Short-Term
Respiratory Symptoms
Respiratory Symptoms
Lung Function
Lung Function
AHR and Allergy
Host Defenses
Emergency Medical Care for
respiratory outcomes

Cardiovascular Outcomes/Short-Term
CVD/Short-term Exposure


Respiratory Morbidity/Long-Term

Reproductive outcomes

Long-Term Mortality


Adults/Repolarization Changes
Adults/Blood Markers
Adults/Acute MI
Adults/CVD Emergency care
Adults/Cardiac Emergency Care
Adults/Stroke Emergency Care

Cardiovascular/Re spiratory

Respiratory Health
Lung Function



Mixed/no independent effect
Effects in clinical studies of peaks
Suggestive of increase in AHR
Weakly suggestive epi findings
Suggestive evidence for association

Some suggestive findings
No Conclusion
No effect
No effect
No evidence for increased risk
Collective evidence weak for

Positive coefficients but possible
confounding; overall, evidence is
suggestive but limited
Association but possible confounding

Evidence is suggestive,
OO 5
No indication of an effect
Unlikely to have an effect

Difficult to draw conclusions

Several studies suggestive, limited
interpretation for causality

Specific Comments
Page: 1-3
Number: 1 Author: JSAMET Subject: Sticky Note Date: 11/26/20073:01:51 PM
Confusion here in terminology and concept. Synergism is one form of effect modification and interaction
and effect modification are often used interchangeably.

Page: 2-38
Number: 1 Author: JSAMET Subject: Sticky Note Date: 11/26/2007 3:58:52 PM
But certainly, outdoor-indoor relationships are probably highly variable within and across communities.

Page: 3-6
Number: 1 Author: JSAMET Subject: Sticky Note Date: 11/26/2007 4:16:13 PM
Or, the effect of SO2 occurs through secondary PM formation

Page: 3-40
Number: 1 Author: JSAMET Subject: Sticky Note Date: 11/26/2007 4:32:57 PM
This is a rather "loose" way to summarize a complicated body of evidence. This is not a sufficiently
transparent finding.
Page: 3-42
Number: 1 Author: JSAMET Subject: Sticky Note Date: 11/26/2007 4:35:12 PM
Again, not clear what is the basis for this summary judgement of a very mixed body of literature.

Page: 3-44
Number: 1 Author: JSAMET Subject: Sticky Note Date: 11/28/2007 12:43:07 PM
Not clear how epi studies contribute to biological plausibility.

Page: 3-74
Number: 1 Author: JSAMET Subject: Sticky Note Date: 11/26/2007 4:47:25 PM
This discussion refers to confounding based on changes in estimates comparing single-pollutant with multi-
pollutant estimates. Such changes may have other explanations related to mediation of effects by secondary
PM and differing degrees of measurement error for correlated pollutants.

Page: 3-101
Number: 1 Author: JSAMET Subject: Sticky Note Date: 11/26/2007 4:58:14 PM
What is meant by "concentrated? Main point is that the high exposures are largely in the East?

Page: 5-1
Number: 1 Author: JSAMET Subject: Sticky Note Date: 11/26/2007 6:12:33 PM
This sentence is conceptually vague and needs to be expanded into an introductory paragraph that is far

Page: 5-6
Number: 1 Author: JSAMET Subject: Sticky Note Date: 11/28/2007 3:20:42 PM
These would appear to be the "rules of evidence" for interpretation of the findings. They need description
and development in Chapter  1. Was this set of criteria given to all auithors? Was the same approach used in
the NOx ISA?

Comments from Dr. Ultman

Chapter 1. Well-written chapter.  The inclusion of the framing questions, as was done in
the recent ISA for ozone is a good idea.  On lines 8-9 of  page 1-2 where current SOx
standard is given, I would add values in ng/m3 as well as ppm.

Chapter 2. This chapter rather seamlessly synthesizes a diversity of material from
source-to-dose.  With respect to the dosimetry section, there was no mention of the
extrapolation of animal exposure conditions to equivalent exposure concentrations in
humans using mathematical modeling or default scaling methods. It is not clear whether
this was an oversight, or if there is nothing in the literature specific to  SOx that
adequately informs such an animal-to-human extrapolation.  Also absent from the
dosimetry section was the notion that (because reaction products of inhaled SO2 can
reach the bronchial and pulmonary circulations) exposure of the respiratory system to
SO2 might lead to downstream systemic effects.
Some specific suggestions and editorial comments for the authors to consider follows:
                Define what the symbols OH and M represent, and show OH as a free
2-2      2-2    You may want to show HO2 as a free radical.
2-9   Fig 2.4-1  Change styles of bar fills so that legend entries are more obvious, or use
                color plates in final ISA.
2-10  Fig 2.4-2  Change to grayscale instead of pseudocolor in final plots.
2-11  Fig 2.4-3  Change to grayscale instead of pseudocolor in final plots.
2-13    1-2      It's not clear how this sentence follows from the previous sentence.
2-13  Fig 2.5-5  Contrary to the explanation in the text and the figure title, this is not a
                box plot. Also, it would be helpful to add another plot of this type
                showing the distribution among all cities of the  daily-average
                concentrations, and how this varies throughout the year.
2-14   15       Pearson's correlation coefficient between what two variables?
2-26 Fig 2.4-11  Change to grayscale instead of pseudocolor in final plots.
2-29   10       This equation assumes that ambient and non-ambient exposures are
                additive. This is not necessarily the case.  It is more accurate to say
                that concentrations arising from ambient and non-ambient sources are
                additive as represented in Eq. 2-11.
2-29  26        Equation 2-11 is really a "recasting" of equation 2-10.  Is there a final
                equation for ET that is missing?

2-43  20        Nodelman and Ultman did not make any measurements of SO2.
2-44  6-8       It was not "total respiratory absorption" but rather "absorption
                efficiency that was probably independent of inspired concentration.
                From this point to the end of the chapters the authors must be more
                careful to specify whether which of these two quantities they mean.

Chapter 3.   This is a well-organized chapter covering numerous important issues that
arise in the identification of health effects and sensitive populations.

Page   Line
3-8  Fig 3.1-2  Why does the the right-most portion of the solid curve break up into
               isolated points?
               extrapolated from data at lower concentrations?
3-9    25      At this point and throughout the chapter, the phrase "lower respiratory
               symptoms" is frequently used.  "Lower respiratory tract symptoms"
               would be a more accurate phrase.
3-13   30      Add "of'after "7/8."

Chapter 4.     An important point in this chapter is that public health impact should be
evaluated in terms of the magnitude of a health  effect as well as the number of people it
affects. This is particularly important when trying to identify susceptible subpopulations.

Chapter 3 contained several tables illustrating the odds ratio or relative risk associated
with SOx exposure and how this varied in different subpopulations. On the other hand,
chapter 4 contained tables and text that documented the size of these subpopulations.
Yet, no where in chapter 4 are the two synthesized so that the reader can appreciate the
possible health impacts.

Chapter 5.   This chapter would be more effective if it was organized around the
framing questions laid out in chapter 1.   It would then be more apparent whether or not
the ISA exercise was successful.  As it stands now,  it is difficult and, in some cases, not
possible to infer the answers to the framing questions from the conclusions made in
chapter 5.

Comments from Dr. Balmes


Charge 4  To what extent is the discussion and integration of evidence from the animal
          toxicology and controlled human exposure studies and epidemiologic studies
          technically sound, appropriately balanced, and clearly communicated?


Like its counterpart in the NOx ISA, Chapter 3 on Health Effects is long (104 pages) and
overly detailed in certain parts. There should be less detail about specific studies in the
chapter text; these details are best left to the annex.  That said, Chapter 3 in the SOx ISA
does a better job of summarizing and evaluating the evidence in each section than the
previous effort. Still, by trimming detail and endeavoring to present the information in a
more thematically clear manner (i.e., each section should have a clear "story line"), a
revised chapter will better support whatever recommendations for an air quality standard
emerge from the review process.  The chapter as currently written reads too much like a
mini-criteria document rather than an integrated  synthesis.

In general, the presentation of the results of the animal lexicological, controlled human
exposure,  and epidemiological studies that have been reviewed is technically sound.
However,  the criteria for  selection of specific studies in all three categories should be
clearly stated.  In addition, the criteria for judging the strength of findings from specific
studies as  well as those used to assess aggregate  findings of studies on a relevant research
question should also  be clearly stated.

I am concerned about one section ( Airway Hyperresponsiveness and Allergy).
On pp.3-27 and 3-28, there are several paragraphs that discuss results of animal
toxicological studies  in support of the statement that "A limited number of animal studies
also suggest acute SO2-induced increases in airway obstruction in allergen-sensitized
guinea pigs and sheep."  This statement is the lead sentence for a paragraph that discusses
three papers (Douglas et al., 1994; Lewis and Kichner, 1984; and Scanlon et al.,  1987).
First, none of these studies involved allergen-sensitized guinea pigs or sheep. Second,
only one of these studies  (Douglas et al.,  1994) involved allergen-sensitized animals.
Third, none of these studies provide evidence in  support of the statement.  The next
paragraph in this section discusses a paper by Riedel et al. (1988) in a manner that
misrepresents the results of the study. The Riedel paper actually presents results that
show that  SO2 exposure enhances sensitization to allergen in guinea pigs rather than "the
effect of SO2 exposure in ovalbumin-sensitized guinea pigs." The study animals were
exposed to ovalbumin only on the last 3 days of a 5-day exposure protocol; one week
after the end of exposure  to various concentrations of SO2 or filtered air the animals were
challenged to ovalbumin  and the  SO2-exposed animals, at all concentrations, were more
sensitive to ovalbumin than the control animals.  The next paragraph opens with the
statement  "Similar findings were observed in which guinea pigs were exposed to a single

SO2 concentration."  The Park et al. (2001) paper actually is similar to the Riedel paper
because it presents results that suggest that SO2 enhances sensitization, not that "airway
obstruction induced by ovalbumin challenge was higher in ovalbumin-sensitized guinea
pigs exposed to 0.1 ppm SO2 for 5 days compared to sensitized guinea pigs that werfe
not exposed to SO2."  The results of the Kitabatake et al. (1992, 1995) and Abraham
(1981) papers are correctly discussed. The concluding paragraph of the section contains
the statement that "Toxicological studies that observed increased airway obstruction and
hypersensitivity in allergen-sensitized animals provide biological plausibility."  The
Kitabatake and Abraham papers are the only toxicological studies that support the
increased airway obstruction part of this statement. The Riedel and Park papers actually
do not provide evidence in support of "increased hypersensitivity in allergen-sensitized
animals"; rather, these two papers show that SO2 exposure can enhance sensitization to
an inhaled allergen. This subsection needs to be rewritten.

In section there should be some discussion of papers by Jedrychowski et al. that
present results of the longitudinal study of lung growth in children and adolescents in
Cracow, Poland. Although there was exposure to both particulate matter and SO2 in
Cracow during the study period, these studies provide evidence of a possible effect of
SO2 on growth of lung function in children and merit discussion.

In my view, the epidemiological data are relatively consistent and coherent with regard to
the association of short-term exposure to SO2 and emergency department
visits/hospitalizations for asthma and all respiratory diseases, particularly among
children. The toxicological evidence in Chapter 3 could be more clearly presented to
convincingly support potential mechanisms for asthma exacerbation.

The toxicological data that are best presented in the chapter are the controlled human
exposure data which indicate that asthmatic individuals are especially sensitive to SO2
exposure in terms of respiratory symptoms and bronchoconstriction. While these data do
provide some plausibility for the epidemiological studies that find an association between
ambient SO2 and emergency department visits or total hospitalizations for asthma, they
do not illuminate how SO2 exposure might induce respiratory hospitalizations in non-
asthmatic individuals. The animal toxicological data provide little help because
exposures much higher than ambient are required to induce measurable effects.
Enhancement by SO2 of airway responses to specific allergen challenge is perhaps the
potential mechanism of asthma exacerbation best supported by the animal toxicological
data. While these data are discussed in Chapter 3, they are included in a way that
somewhat misrepresents their meaning (section and p. 5-9 "Airway
Hyperresponsiveness" bullet) in an effort to support the limited epidemiological evidence
that SO2 can induces airway hyperresponsiveness in atopic individuals.

In terms of clear communication, Chapter 3 as currently drafted falls somewhat short.
The text in Chapter 3 needs to be tighter, less redundant, and more thematically organized
(i.e., each section should have a story line). In particular, the summary/integration
subsections should provide an overview of the quantity and quality of the evidence for

the health outcome(s) of interest as well as evaluation of how well the toxicological data
support the epidemiologic findings.


p. 3-3, line 26 The title of section 3.1.1 should make it clear that it is a summary of
literature reviewed in previous documents.

p. 3-14, line 19      In epidemiological  studies assessing the relationship...

p. 3-16, lines 2-4     .. .reported in most  studies mean that the evidence is insufficient to
conclude that short-term exposure to ambient SO2 has an independent effect on lung

p. 3-26 and throughout the document        Airway hyperresponsiveness (AHR) should
be used consistently throughout the document. Refraining from using bronchial
hyperresponsiveness (BHR) or airway hyperreactivity (Chapter 4) will avoid unnecessary

p. 3-27, lines 20-21  See comments on section above. This sentence is
inappropriate here. The papers discussed  in this paragraph do not support this sentence
and require a different interpretation.

p. 3-28, lines 19-20  As noted in the general comments above, this statement is only
supported by two of the studies reviewed in the previous paragraphs in this subsection
(Kitabatake and Abraham).

p. 3-30, line 9 ... and no alterations in pulmonary immune system function were

p. 3-32, Figure 3.1-7 legend        There is no asterisk in the figure.

p. 3-34, line 3 ... and 1-h max in Paris, France.

p. 3-35, line 18      .. .though not always statistically significant...

p. 3-36, Figure 3.1-9 legend        There is no asterisk in the figure.

p. 3-37, Figure 3.1-10 legend        There is no asterisk in the figure.

p. 3-39, lines 31-32  In summary, there are limited studies..., making it difficult to draw

p. 3-42, lines 8-9     "and  the animal toxicological studies that observed SO2  altered
lung defenses" should be deleted.  The interpretation of the relevant toxicological data on
p. 3-30, lines 19-21, is the correct interpretation.

p. 3-45, lines 1-2     Ibid.

p. 3-47, line 20      ...and_coagulation markers...

p. 3-49, line 8 .. .NO2, CO, and PM2.5 were found...

p. 3-52, line 28      ...but not SO2 orNO2...

p. 3-57, line 4 .. .the relationship between ambient SO2 and hospitalizations for cardiac

p. 3-59, Figure 3.1-13 legend        There is no asterisk in the figure.

p. 3-62, line 30      .. .no adverse effects on development or reproduction other than
the limited evidence of neurodevelopmental effects noted above.

p. 3-63, line 13      Exposure to >5-ppm SO2 was found...

p. 3-69, line 11      ... for NO2-mortality associations, but were weaker.

p. 3-77, line 14      ... CHF 1 y_ear before death were compared.

p. 3-78, line 29      ... other constituents.. .are responsible for the adverse effects.

p. 3-84, section       See my comment above about including discussion of
results from the Cracow longitudinal study of lung function.

p. 3-88, line 32      .. .The reliability and validity.. .have been reviewed...

 p. 3-93, line 16      Is this really SO here?


p. 4-8, line 23 Should be airway hyperresponsiveness here.

p. 4-10, lines 1-2     The last two lines of p. 4-9 are repeated here.

p. 4-12, lines 21-26  The Ponce de Leon et al. (1996) and Anderson et al. (1998) papers
should not be cited in support of this statement given the data presented in Figure 4.2-1.

p. 4-14, lines 25-27  .. .(e.g., homozygosity for the null allele at the GSTM1 and GSTT1
loci, the presence of the Val-105 variant allele at the GSTP1 locus) that significantly
affect expression of protein or function in the lung.

p. 4-18, line 8 .. .with approximately 10% of adults or 13% of children having ever been
diagnosed with asthma...


p. 5-9, lines 18-20    As noted for Chapter 3, the first part of this statement is supported
by only two papers cited in the document and the second part is misinterpreted. This
bullet on "Airway Hyperresponsiveness" should be rewritten.

Comments from Dr. Pinkerton

Charge 4  To what extent is the discussion and integration of evidence from the animal
          toxicology and controlled human exposure studies and epidemiologic studies
          technically sound, appropriately balanced, and clearly communicated?

Charge 5  To what extent does the integration of health evidence focus on the most
          policy-relevant studies or health findings?

Charge 6  What are the views of the Panel on the conclusions drawn in the draft ISA
          regarding the strength, consistency, coherence and plausibility of health
          effects of sulfur oxides?

Charge 7  What are the views of the Panel on the appropriateness of public health impact
          and the characterization of groups likely to be susceptible or vulnerable to
          sulfur oxides?


The presentation and relevance of the toxicology, controlled human exposure and
epidemiological studies within this report is excellent. The studies are logically
interpreted.  The length and depth of the material is quite extensive, therefore, some
consideration should be given to 1) shorten and/or combine materials from chapters 3 and
4 and 2) include further materials in the Annexes of the ISA.

The authors  of the chapter have provided a nice balance for each of the studies referenced
and described in  the document. The  scope of the studies presented is reasonably
complete and adequate. The human health studies are a major force to drive the newest
and most relevant information to facilitate updating the criteria document for SOX.

A key issue in establishing the next NAAQS standard for  SOX is the manner in which
SOX is measured and reported. The  importance of exposure  to SO2 over the short
timeframe of exposure (5 minute peak concentrations) is of high relevance as reflected in
much of the  writing of the ISA.  These short-term analyses are likely to be a better
predictor of health effects. Therefore, such measures are critical to be included as part of
the analysis in the next  consideration of the SOX ISA. The authors provide solid support
for the inclusion  of such measures which is highly commendable. The importance of
inclusion of SOX measures over the short  timeframe has been argued previously, but
dropped from consideration in the previous review of the document.  Therefore, the
authors should be applauded for their renewed interest and strong justification for the use
of a 5-minute peak concentration in evaluating 1  and 24 hour averaging times. This
assessment could provide extremely important insights as the tiered approach

The animal toxicology studies in large measure are performed at SOX concentrations that
are orders of magnitude higher than those  for human clinical  or epidemiological studies.
Although these differences in exposure ranges complicate the use of animal toxicology

studies to establish appropriate levels for the protection of human health, never-the-less
these studies do provide biologically plausible mechanisms of effect for exposure to

A critical component of the ISA draft document is to define SOX effects as clearly as
possible, while also making it very clear of the ever-present difficulties to deal with
confounding co-pollutants. Although it is fairly clear SO2 is the most prevalent form of
gaseous SOX, while other forms of sulfur oxides tend to from particulates, it should be
done clear SO2 is the critical measure to represent SOX. However, if this is not the case,
then further clarification should be made in the document.

A critical review of the most current literature since the last NAAQS document for SOX
emphasizes the need for additional research to further explore the importance of genetic
susceptibility and whether genetics further influences susceptibility for those in sensitive

There appears to be a general lack of consideration for the sensitivity of newborns and
children to the effects of SOX.  Asthma has clearly been identified in both young and old
that is a factor to increase risk of the effects of exposure to SOX.  However, it will also
be critical to further investigate the potential and biologic plausibility of children to
exposure to  SOX that may be different from that observed in adults.

Mortality is  clearly an important consideration in assessing the health effects of SOX.
However, it is unclear how aggressive will be policy-decision making and/or the factors
that must be taken into consideration to factor mortality into a revised criteria document
for SOX. Will such a regulation be based on a specific threshold and/or a linear model?

It is also important to note that accurate exposure levels today in the United States today
are almost never as high as that level found in the current  SOX NAAQS.

Comments from Dr. Sheppard

Overall:  The ISA follows a new structure and represents a substantial change from the
previous scientific review. EPA is to be commended for a great start in carrying out the
intent of the new process.

1.  To what extent are the atmospheric chemistry and air quality characterizations clearly
communicated, appropriately characterized, and relevant to the review of the primary
SO2NAAQS? And 2. Are the properties of ambient sulfur oxides appropriately
characterized, including policy-relevant background, spatial and temporal patterns, and
relationships between ambient sulfur oxides and human exposure?

The characterization needs to more directly address the spatio-temporal variation in SO2
and the contribution of monitor siting characteristics to the summary data because this
has direct bearing on the interpretation of the epidemiological studies. Monitors sited
near local sources that don't reflect usual population exposure aren't appropriate to be
used in time series studies. How prevalent are such monitors?  How many cities have
only local source monitors?

Analyses should be added to relate 5-minute averages to 1-hour averages to bring
forward into the health assessment.

The discussion of correlations is too  ambiguous as presented to be useful (p 2-33 and
Table 2.5-1).

I don't find the information in the ISA adequate for understanding personal exposure to
SO2.  Even if measurement data are  limited because of LOD issues, there are studies of
I/O ratios and (lack of) indoor sources that add important information to the
understanding of characteristics of personal SO2 exposure. I would like the ISA to
assemble and integrate the research so that summary statements about estimates of
attenuation of ambient source concentration for personal/indoor exposure (e.g. a or Finf)
are included in the document.

3.  Is the information provided on atmospheric sciences and exposure sufficient for the
evaluation of human health effects of sulfur oxides in the ISA?

No. See my comments w.r.t. question 2.  Make sure to bring forward key points about
exposure with respect to evaluation of health effects into chapter 5.

Chapter 2 comments

Presentation in this chapter needs to  focus on understanding needed to interpret the health
studies and to inform the health assessment.
   •   Revamp the analysis of AQS data to address key aspects of exposure that are
       relevant to health studies.  This includes adding a thorough analysis of 5-minute
       averages and their relationship to 1-hour averages, and analyses separating

       monitors by proximity to source and other key features of spatial distribution.  In
       order to estimate the susceptible population exposed to SO2, spatial analyses to
       predict SO2 features by land use characteristics at the level of census tracts is
       needed. Add in-depth analyses of AQS data to the annex and then bring forward
       the most important features into chapter 2.
   •   Include land use and/or point source analyses to help with later analyses that will
       need to quantify the degree of population exposure to high levels of SO2 from
       point sources
   •   Exposure measurement error in time series studies is likely to be a very important
       factor for SO2 because of the local source nature of this pollutant and the monitor
       siting policies.  (This could actually be an issue for all short-term epidemiological
       studies, also including case-crossover and panel studies.)  Analyses of AQS data
       to try to clarify this issue are needed.  For instance some cities only have SO2
       monitors sited near a local source (e.g. Seattle). This has  huge implications for
       estimates from time series studies. Add an analysis that classifies cities according
       to siting criteria of monitors (e.g. only local source, only population-oriented,
       both types) and provide results that allow this feature to be incorporated into the
       evaluation of the epidemiological studies.
   •   Analyses of personal or indoor concentrations vs. outdoor concentrations need to
       be summarized  in the context of informing the models for personal exposure from
       ambient and non-ambient source exposures.  The summary in the document
       (section 2.5.3) does not address this adequately.  Table 2.5-1 needs to be
       revamped to clarify the correlations presented. I suggest including scatterplots of
       daily data and separating individuals/homes in the plots. Analyses need to be
       conducted so parameters assumed in APEX can be obtained directly  from the
       ISA. (More extensive analyses can be put in the Annex and appropriate
       summaries brought forward.)

Tables and figures in the chapter need to be thoroughly revamped.  Some may be
completely revised based on my recommended thorough analysis of AQS  and other data
that should be added to the Annex.
   •   The Figure 2.5-2 summary of indoor and outdoor SO2 only shows annual means.
       This analysis is so highly summarized it is difficult to apply these results to the
       epi studies. It would be more relevant to look at scatterplots of daily data.
   •   Figure 2.4-5 is mistitled and mixes data from all monitors regardless of siting
       criteria. I think it is important to distinguish extremes due to local sources from
       extremes at locations without local sources because the proportions of the
       population exposed or not to local sources is relevant to the interpretation.
   •   Table 2.4-1 only provides means and does not summarize distributions. At a
       minimum add a measure of spread.  Think about how to make this analysis more
   •   Table 2.4-2 also needs stratification by siting characteristics or other features to
       be identified from thorough descriptive analysis of the AQS data.
   •   Table 2.4-3 is too highly summarized to be useful.  The summary doesn't inform
       the interpretation of epidemiological studies.  It isn't clear how each  of the ratios
       is calculated.  Distinctions such as single vs. multiple sites, single vs multiple

       individuals, individual-level vs averaged data, home ventilation characteristics,
       and presence or absence of indoor sources cannot be discerned from the table.
       Revamp to allow these results to be used to inform the estimates of Finf and/or a
       in the personal exposure model.

4.  To what extent are the discussion and integration of evidence on the health effects of
sulfur oxides from the animal toxicological, human clinical, and epidemiological studies,
technically sound, appropriately balanced, and clearly communicated?

There needs to be a protocol for reporting on studies and selecting what information to
report. As a strong example of information that should not be included, I think it is
inappropriate to report the results of finely stratified subgroup analyses.  As an example,
in summarizing the results of Bozen et al (p 3-27 1 3) the initial sample size of 327 was
reduced to a subgroup of 25 for reporting.  This result is almost certainly an artifact of the
post-hoc stratification of the analysis and reporting. A uniform protocol for selecting
papers and results within papers would certainly not allow reporting of such a result.

Chapter 3 comments

Restating a point made with respect to chapter 2, interpretation of all epidemiological
studies of SO2 effects (both long-term and short-term) needs to be assessed in the context
of exposure assessment, particularly with respect to how evenly distributed SO2 is in
space in a certain area, and the effect of local sources, particularly on the monitoring data
used in the study.
    •   I question whether short-term epidemiological studies based on central site
       monitoring data are reasonable given the local source nature of SO2.  Monitor
       siting needs to be considered as part of the evaluation  of the short-term
       epidemiological studies (e.g. time series, case-crossover,  panel) because of the
       low monitor-to-monitor correlations frequently observed and the prevalence of
       SO2 monitors sited to capture the effects of local sources. Using a monitor that
       doesn't represent usual population exposure in these studies is likely to have a fair
       amount of classical measurement error in the estimates.

It is not adequate to refer to "epidemiological studies" as a  synonym for specific
epidemiological study designs such as time series studies (e.g. p  5-13; not sure this is an
issue in this chapter).

I am concerned that the ISA review and annex summaries for health gloss over very
important subtle issues that affect the analysis and interpretation of the studies. Each
study has unique strengths and weaknesses. Some features are shared by the design,
exposure data, or software, while some are unique to the individual study. While
summarizing such a complex set of features will be very difficult to address completely,
progress  can be made by expanding the detail in the summaries of each study and
following a strict protocol  for reviewing, summarizing, and evaluating each study.
Expanding upon the list of detailed information summarized on page 1-6, the tabular
summaries should include (Ib) characteristics of the monitoring data used in the study

(2b) study design (2c) summary of analysis approach and key features of the analysis (3b)
description of the population selection criteria (5) any major limitations or issues with
study, (6) discussion of study strengths, weaknesses, and key features. (I note some of
this information is mentioned in individual studies.)
    •  As an example, here are some features of Schildcrout et al that I would emphasize
       but which are not included in the ISA or Annex: This paper is stronger than other
       panel studies of its type because of the additional control for confounding
       provided by using within-city estimates of the  odds ratios and its unique focus on
       the combined effects of changes in two pollutants simultaneously. The summary
       should include the average number of observations per subject and unique details
       about the analysis.  Key features of the pollutant data may impact study results,
       including the extensive imputation of the PM data, the limited availability of the
       ozone data, and the siting of SO2 monitors in some cities (e.g. the only SO2
       monitor in Seattle is right next to a cement plant).

Add comments on interpretation of epidemiological studies.  For instance,
    •  Monitor siting in individual studies is an important feature.
    •  Discuss multipollutant models and the interpretation  of studies where a single
       effect in the presence of others, i.e., the effect of a change in SO2 with all other
       pollutants held constant (see e.g. figure 3.1-11  p 3-41) vs.  studies where joint
       effects are reported from changing SO2 and another copollutant simultaneously
       (see e.g. figure 3.1-1 p 3-7). The role of exposure measurement error could have
       important ramifications in multipollutant analyses because SO2 may have more
       measurement error than PM.
    •  The evidence that the population is heterogeneous with respect to exposure that
       induces response has important implications for population-level epidemiological
       studies (e.g. time series studies).  Because of averaging, heterogeneity in the
       threshold is likely to translate into a linear dose-response function in time series

Is it possible to obtain the relevant quantities from epidemiological studies reporting
multipollutant model results in order to calculate the joint effects of the change in two
pollutants simultaneously? It may be worthwhile to contact  authors of key studies and
determine if the relevant quantities can be obtained. If so, ISA authors may wish to
consider this suggestion a possible comprehensive strategy for reporting epidemiological
study results at a given lag:
    1.  Single pollutant model for SO2  [Independent effect]
    2.  Joint effects model for SO2 and selected other pollutants (perhaps only PM)
       (Note: need to think carefully about the assumed increments for comparability
       with the other analyses)  [Mediated effects by another pollutant in addition to
    3.  Multipollutant model estimates of SO2 adjusted for PM or selected other
       pollutants (these others are held constant, again perhaps only PM) [confounding
       effects]. Similar estimate for the PM (or other pollutant) effect adjusted for SO2.

    4.  Commentary on interpretation of each of the above effects, along with discussion
       of the lack of evaluation of effect modification (given the published analyses by
       and large haven't done this).

7.  What are the views of the Panel on the appropriateness of public health impact and
the characterization of groups likely to be susceptible or vulnerable to sulfur oxides?

A  discussion of responders needs to be added to the public health impact chapter. While
such individuals haven't been shown to be a clearly identifiable subgroup they are a
significant fraction of the population and deserve attention.

Chapter 4 comments

Variable sensitivity in the population (i.e. presence of responders) is a key issue for
regulation and needs to be addressed directly in this chapter.

Address overlap with chapter 3 in revision of this chapter.  Look at public health impact
from a regulatory point of view and discuss the evidence from this perspective.
5.  To what extent does the integration of health evidence focus on the most policy-
relevant studies or health findings?

There needs to be a protocol to systematically classify studies with respect to their policy-
relevance and use that to weight the evidence.  Aspects of this protocol must address
adequacy of exposure data, study design, population size and relevance, and publication

6.  What are the views of the Panel on the conclusions drawn in the draft ISA regarding
the strength, consistency, coherence and plausibility of health effects of sulfur oxides?

The organization  of chapter 5 should be reevaluated with respect to the framed policy
questions (page 1-2) and information needed in the health assessment.  In addition, the
criteria for conclusions need to be specified in the chapter, including the classification of
health  evidence, the selection of studies to focus on, evidence for coherence, etc.  The
classification of health evidence should be revamped using published criteria from
NAS/IOM as a guide.

Chapter 5 comments

Organization: It would be useful to consider organizing this chapter with respect to
information that needs to be brought forward into the health assessment.  Important
questions include are the properties of the atmosphere well measured, what is the
relationship between concentration and exposure, what do the monitoring data tell us
about the  epidemiological studies.  In addition  or alternatively, the chapter should address
and can be organized around the policy questions framed in chapter 1.

Add cross-referencing to previous chapters whenever possible to facilitate the reader's
ability to get more in-depth information.

The concluding statement from this chapter (p 5-17 to 5-18) needs to be better
substantiated.  It needs to be explicitly stated if this statement should be interpreted in the
context of the  already low usual population exposures to SO2 or regardless of the existing

Section 5.1.3:  Need to revise the summary to be more useful to the exposure assessment.

P 5-3 1 19-21:  I would like to see a summary statement about the slope or estimate of

P 5-3 to 5-4: Add comments about monitor siting and their effect on different study
designs, particularly the role this may play in interpretation of time series studies.

P 5-9 to 5-10 summary of respiratory ED and hospitalizations: Here is an example where
the alignment  of the epidemiological and clinical/toxicological may be helped by
comments on the representativeness of exposure to SO2 in time series studies.

P 5-13 1 14-17: Is it justified to base the ISA conclusion so completely on the conclusion
reached by the authors of this study?

P 5-13 1 31: Here is a place where  the general term "epidemiological" appears to be used
for the more specific term of "time series".  All users of this document will be better
served by the more exact term.

P 5-14 1 4: Include the references here when specific studies are referred to so explicitly.

P 5-14 1 9-10:  Here is an example  of a place where it would help the reader to cross-
reference earlier chapters and/or the annex.

P 5-16 1 14-20: Representativeness of monitors is key for SO2 study interpretation.
Futhermore, the ability to separate  effects of SO2 from other pollutants is a fundamental
one for interpretation of SO2 effects estimated in epidemiological studies.

P 5-17 1 9-10:  I would hope that the further analysis of AQS data requested will better
inform this statement.  If after that  analysis the statement is still justified, is there a
plausible chemical/atmospheric reason for this?

8.   What are the Panel's views on the adequacy of this first external review draft ISA to
provide support for future risk, exposure and policy assessments?

Overall, this is a good first start, but as noted in detail above and by other panel members,
there are areas that need additional attention or thorough revision.

Additional specific comments

1-5 to 1-6: I note that controlled human studies are not addressed.  Is this an oversight?

2-13 Figure 2.4-5 is a scatterplot not a boxplot.

2-141 20-21: Correlations are very sensitive to the range of the data.

2-40 1 19-20: A huge population would be needed to get a good estimate of XtA, so this is
not feasible in practice for any pollutant.

2-40 1 27: The population size is relevant to this argument and should be stated. Also it
would be informative to comment on how results would compare for SO2 if they were

2-41 1 1-7: A comment on the applicability to SO2 is needed here since it is so highly

2-41 111: Replace "epidemiological" with "time series".

2-411 12-14: It is more correct to say the time series studies estimate a different
parameter because they use concentration instead of exposure.

2-42 1 7-16: This paragraph needs work.  Attenuation of an effect estimate could indeed
change conclusions.  The complex soup of pollutants that exist in the atmosphere could
play an important role in health but the chapter hasn't addressed this directly.  Is this the
place to discuss incorporating evidence from clinical and toxicological studies?

3-26 1 10: Insert the number of tests per person, e.g. "range of	to	per person".
Here is an example where deeper understanding of the key features of the study may help
with interpretation.

3-84 121: The cross-sectional design is a weak design so I would not put much weight
on this result.

3-84 1 28: Does "no consistent" mean "not significant"? Elaboration about consistency
should not be a discussion of positive vs. null findings with no point estimates or
confidence intervals reported.

4-11 30 to 4-2 1 1: Other possible reasons are  heterogeneity of attenuation of ambient
concentration and exposure measurement error due to monitor siting near local sources.

4-3 1 5-8: With only 3 dose levels, it is difficult to fit much other than a linear function.

4-3 1 19-26: Has the Lin et al study been evaluated to assure readers there has been
adequate control for seasonal confounding? There is limited mention of this in the annex,
but the case-control design suggests the extra attention is merited.

4-3 1 27: Please not the design, not just the general term "epidemiological".

4-4 1 5-6:  I think there are many features of a study that are more important to mention in
the ISA than the GAM convergence criteria.

4-5 1 15-25: Can this discussion be made clearer?  Wasn't some of the work done by
these authors (Brauer et al) published in the peer-reviewed literature? See for instance
Brauer, Brumm, Vedal, Petkau 2002 in Risk Analysis, Vol 22, p 1183-93.

5-3 1 19-21: It would be more useful if this bullet were to focus on a summary of the
slope of the association and not merely a description of the strength of the association.

5-3 section 5.1.3: Add a bullet on the effect of monitor siting on exposure, particularly
with respect to different study designs.

Appendix Table 5A-3:  Is it worth adding additional percentiles so there is some hope
that more studies will have some air quality data represent