UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D.C. 20460
OFFICE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
August 8, 2008
EPA-CASAC-08-017
The Honorable Stephen L. Johnson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, NW
Washington, DC 20460
Subject: Clean Air Scientific Advisory Committee's (CAS AC) Peer Review of
EPA's Integrated Science Assessment (ISA) for Sulfur Oxides - Health
Criteria (Second External Review Draft, May 2008)
Dear Administrator Johnson:
The Clean Air Scientific Advisory Committee (CAS AC), augmented by subject-
matter-experts to form the CASAC Sulfur Oxides Primary NAAQS Review Panel
conducted its review of EPA's Integrated Science Assessment (ISA) for Sulfur Oxides -
Health Criteria (Second External Review Draft, May 2008) on July 30-31, 2008. The
first draft ISA was reviewed by CASAC in December of 2007 and subsequently revised
in response to CASAC advice transmitted in January 2008. This letter presents
CASAC's advice on the second draft ISA issued in May 2008.
The CASAC finds that the second draft ISA is greatly improved and that the
Agency has been responsive to the CASAC's earlier review. There are some remaining
concerns as described in our answers to the Agency's charge questions in the main body
of this letter. Our major concern is the conclusions in the ISA regarding the weight of the
evidence for health effects for short-term exposure to low levels of SO2. Although the
ISA presents evidence from both clinical and epidemiological studies that indicate health
effects occur at 0.2 ppm or lower, the final chapter emphasizes health effects at 0.4 ppm
and above. As discussed in our responses to Charge Questions 3 and 5, CASAC believes
the clinical and epidemiological evidence warrants stronger conclusions in the ISA
regarding the available evidence of health effects at 0.2 ppm or lower concentrations of
SO2. The selection of a lower bound concentration for health effects is very important
because the ISA sets the stage for EPA' s risk assessment decisions. In its draft Risk and
Exposure Assessment (REA) to Support the Review of the SO 2 Primary National Ambient
Air Quality Standards (July 2008), EPA chose a range of 0.4 ppm - 0.6 ppm SO2
concentrations for its benchmark analysis. As CASAC will emphasize in a forthcoming
letter on the REA, we recommend that a lower bound be set at least as low as 0.2 ppm.
-------
With this and other modifications as recommended below, CASAC finds the scientific
quality of the ISA acceptable for rulemaking.
CASAC's response to EPA's charge questions are summarized below. Individual
recommendations from Panel members to strengthen the final ISA are appended in
Enclosure B.
Charge Question 1: The framework for causal determination and judging the
overall weight of evidence is presented in Chapter 1. Is this the appropriate
approach? Is it appropriately applied in the case of SOx? How could the
framework or its application be refined?
Chapter 1 has been improved, particularly by drawing on recent reports that offer
models of approaches for causal inference and classification schemes for the weight of
evidence for inferring causation. The ISA utilizes a five-level hierarchy for causal
determination to be consistent with the Guidelines for Carcinogen Risk Assessment (EPA,
2005). We concur with using the five levels but recommend that the descriptions be
changed to better reflect the level of certainty or confidence in the classification of the
level of evidence. The phrasing of the second level is particularly problematic in its
addition of the wording "likely causal relationship." The approach to evidence
interpretation should avoid using statistical significance as a criterion for evidence
interpretation. CASAC recommends that EPA reconsider the language used to describe
the weight of evidence, particularly for the first three categories which cover a range of
certainty or confidence in causal inference that extends from full certainty to lesser
degrees. The language used should be consistent with other such schemes used by EPA.
For the criteria pollutants, the ISA needs to acknowledge that the pollutants exist
in complex mixtures and to broadly consider the implications for interpreting the
evidence in the ISA. We also specifically recommend improvement in the presentation
of the epidemiological concepts of effect modification and confounding that are
particularly challenging in the face of multi-pollutant mixtures. Other aspects of the
chapter also need greater development including the treatment of uncertainty. The
concept of "margin of safety" also needs elaboration. Figure 1 is not effective and unless
the text is expanded to explain many of the relationships depicted, it needs to be
abandoned or replaced.
Finally, we applaud the specification of questions at the beginning of the first
chapter that outline specifically the intent of the ISA, and the integration of the answers
to these questions, as obtained from the evidence in the ISA, in the concluding chapter.
This structure will provide a clear scientific foundation for the development of the Risk
and Exposure Assessment document.
Charge Question 2: Have these revisions to Chapter 2 improved its assessment of
the currently available scientific knowledge on atmospheric sciences and exposure
and its relevance to the evaluation of human health effects presented in later
chapters?
11
-------
Chapter 2 is significantly improved from its earlier version and previous
comments from the Panel have been taken into account. There are, however, some points
that need to be addressed in greater detail and other points require some clarification.
Since the REA will focus on 5-minute averages of 862 concentrations, a more
comprehensive discussion of the available 5-minute SC>2 concentration data is warranted.
For example, a table presenting information similar to that in Table 2-4 would be useful.
A better discussion of the discussion of the distributions and time-trends of 5-minute
average concentrations is needed as is greater consideration of which distribution best
represents ambient data (normal or log-normal). Further, the analyses should address
correlation between the 5-minute and one-hour average concentration distributions.
Figure 2-11 needs to be redrafted to provide a better depiction of diurnal SO2 ambient
concentrations. The fact that the trend in 862 concentrations is due to changes in 862
emissions rather than to changes in measurement technique should be stated more
definitively.
The rationale behind the selection of the six states selected for analysis should be
better described; some consideration may need to be given to areas where the potential
effect of ship emissions on SC>2 could be significant (i.e., major U.S. ports). Analyses
(e.g. stratification for source proximity and strength and use of maps) of selected urban
areas should be presented to characterize the representativeness of ambient monitors for
characterizing population exposure. Since AERMOD is used in the REA, there is a need
to describe it in the Annex.
Adsorption of 862 onto particles which is discussed in more detail in Chapter 3
needs to be further discussed in Chapter 2. However, this will need to be placed in
perspective because concentrations of particulate matter (PM) and SC>2 used in the animal
toxicological studies discussed in Section 3.1.5 were several orders of magnitude above
typical ambient concentrations. In addition, it should also be noted that atmospheric
chemistry studies do not indicate significant amounts of sulfite on atmospheric PM
Chapter 2 correctly points out the difficulties associated with characterizing
exposure (e.g., limited information on indoor concentrations, and vertical and horizontal
SC>2 concentration gradients). However, the implications for the characterization of
health effects could be better described and discussion on the implied uncertainties
expanded. More discussion of the effect of exercise on health effects in exposure studies
is warranted. In particular, the ISA points out that whether 862 is inhaled via nasal vs.
oral breathing makes a difference for SCh uptake, however, the implications for the
analysis of the health effects studies need to be discussed in greater detail.
Charge Question 3: In the revision, we reduced redundancy, added summary
sections and reorganized Chapter 3. In addition, discussions on potential
confounding by, and interactions with, co-pollutants have been added. The 2nd
draft ISA also includes additional analyses of individual-level data from human
clinical studies (Sections 3.1.3. and 4.1.1) that builds upon the analysis included in
the 1994 Supplement to the Second Addendum. The toxicology sections were
in
-------
reorganized to focus on studies using more relevant concentrations of SO2 and
sections were added to better discuss mode of action and potential particle-SOi
interactions. We are requesting CASAC review specifically on these analyses as
well as on the integration of the overall evidence from the human clinical, animal
toxicological, and epidemiological studies.
Chapter 3 is vastly improved from the previous draft of the ISA. The new
analyses of human clinical studies are helpful. CASAC believes, however, that more
emphasis should be placed on effects seen at concentrations lower than 0.4 ppm. The
Panel is concerned about the potential underestimation of the proportion of asthmatic
persons affected by short-term exposure to SO2 because only mild and moderate
asthmatic adults were recruited to participate in the clinical studies. People with more
severe asthma or poor control of their symptoms may be more susceptible and may
respond more adversely to SC>2 and discussion of this point should be expanded in the
chapter. The potential influence of medication use on asthmatic responses to SO2 should
also be discussed. In general, the evidence from the animal toxicological, controlled
human exposure, and epidemiological studies has been satisfactorily integrated.
However, the gap between the greater level of SC>2 used in the clinical studies and the
relatively lower ambient concentrations associated with respiratory morbidity in the
epidemiological studies needs to be clearly acknowledged. The relevance of the higher
dose animal studies also needs to better developed. The results of the studies involving
extremely high levels of metal oxide or carbon-SC>2 mixtures suggest that SC>2 effects
may be potentiated by co-exposure to particles, but the relevance of these results to
ambient exposures is unclear. Section 3.1.5 should be greatly condensed to give a more
focused discussion of the mechanistic implications of the mixture studies without
presentation of experimental details. While the new summary sections are a welcome
addition to the chapter, their quality is uneven. It would be helpful if these sections were
written in a style that applied the causal inference approach outlined in Chapter 1 in a
consistent manner.
Charge Question 4: The section on concentration-response relationships in Chapter
4 was reorganized and revised to include analysis of individual-level data from the
human clinical studies and some additional discussion of the difficulties of
discerning a threshold in population-level data. In addition, revisions were made to
better characterize groups likely to be susceptible or vulnerable to SOx and the
potential size of the population at risk for SOx-related health effects. Finally,
revisions were made to reduce redundancy with material presented in Chapter 3.
Have the revisions made to Chapter 4 improved the characterization of the potential
public health impact of SOx exposure?
Revisions to Chapter 4 have improved the characterization of potential health
impacts of SOx exposure, but additional clarification, evaluation, and amplification is
needed. The Panel recommends that EPA revisit the definitions of susceptibility and
vulnerability. The definition of susceptibility should consider the form of the dose-
response relationship and the magnitude of response at a given dose. Greater clarity and
specificity is needed in the definition of this concept if it is to be used without ambiguity.
IV
-------
The current definition is limited in concept and inconsistent with toxicological and
epidemiological usage. EPA should harmonize the definitions of susceptibility and
vulnerability across documents for intra-Agency and inter-pollutant consistency.
Additional analyses of the cited clinical chamber data could be performed to gain
insights regarding the population distribution of effects at exposures lower than 0.4 ppm
SC>2; there appears to be an indication of effects below this level when the pooled data are
considered. Specific details are provided in individual comments. See comments from
Drs. Avol, Hattis, Kinney, Pinkerton, Samet and Sheppard In addition, the discussion
should do a better job of acknowledging that epidemiology studies generally lack power
to distinguish between linear and non-linear response forms.
A more complete discussion of potentially at-risk (i.e. susceptible and/or
vulnerable) sub-populations is needed. The discussion should emphasize the importance
of linking populations at potential risk due to age and disease status with considerations
related to exposure and activity distributions in these groups. The joint distribution of all
these factors defines the population at risk of adverse health impacts from 862 exposures.
In addition, discussion of the geographical location of at-risk populations with respect to
pollution sources needs to be included and related environmental justice concerns
addressed, if appropriate. EPA needs to provide additional justification for its decision to
focus on children and older adults with respect to short term respiratory health effects of
SO2.
Charge Question 5: Revisions were made to better integrate findings from
atmospheric sciences, ambient air data analyses, exposure assessment, dosimetry,
and health evidence in Chapter 5. To what extent do these findings support
conclusions regarding causality of SOx related health effects at relevant exposures?
The Panel concurs with the conclusion that short term exposures to SOx are
causally related to changes in respiratory responses, however we do not accept the levels
on which the ISA focuses. The clinical and epidemiological studies warrant a stronger
conclusion about health effects at lower levels. The range of exposures emphasized in
the ISA (0.4 - 0.6 ppm) has clearly carried over to the thinking used in the REA.
Adoption of this range might leave substantial numbers of exercising mild asthmatics at
considerable risk. Table 5-1 of the ISA presents studies showing that 5 - 20% of mild to
moderate asthmatics experience moderate or greater decrements in lung function at 862
concentrations as low as 0.2 - 0.3 ppm. For ethical reasons severe asthmatics were not
part of these clinical studies, but it is not unreasonable to presume that they would have
responded to even a greater degree. In addition, the epidemiological evidence shows
emergency room visits and hospitalizations for respiratory illnesses associated with 24-
hour SC>2 levels below the current standard (0.14 ppm averaged over at 24-hour period).
Collectively, this evidence should lead to a conclusion that 0.2 ppm or even a lower level
of short-term exposure is an appropriate lower bound value for EPA's benchmark
analysis.
-------
The Panel was impressed by the way the preceding chapters' information was
summarized in a series of tables that effectively collated the textual materials, tables, and
figures from the preceding chapters as well as the wealth of data contained in the
appendices. The conclusions regarding causality with regard to short-term exposures are
reasonable. A better discussion is needed on the range of exposures at which effects are
observed in the clinical studies and the much lower range at which effects are observed in
the epidemiological studies. In addition, the importance of identifying susceptible
subgroups of subjects, and estimating the contribution of these groups to population shifts
in response characteristics may be a task for the REA, but that document should be
informed by a discussion of this issue in the ISA.
The larger question is whether the additional research carried out in the last 10
years has led to a substantial change in the scientific basis for making decisions regarding
the basis for setting standards for SC>2. Although the ISA answers this question
affirmatively, the final chapter should be organized to answer all of the policy-relevant
framing questions introduced in Chapter 1. The answers to these questions would
greatly improve the usefulness of this document as a scientific prelude to the Risk and
Exposure Assessment.
With respect to the entire ISA, there remains the longstanding issue associated
with the Clean Air Act's single pollutant approach to analysis and rulemaking. In the real
world, we breathe complex mixtures of multiple pollutants, thus isolating the effects of a
single pollutant requires simplifying assumptions that depart from real world phenomena.
With SO2 in particular, the reality of having to adjust for the possible confounding
contributions of other pollutants makes definitive determinations of assigned causality a
challenge.
In closing, the CASAC is pleased to provide the Agency with advice and
recommendations in the development of this ISA for Sulfur Oxides, which is a
fundamental part of the new NAAQS review process. We look forward to providing
additional advice as the Agency completes the NAAQS review process for sulfur oxides.
Sincerely,
/Signed/
Dr. Rogene Henderson
Chair
Clean Air Scientific Advisory Committee
Enclosures
VI
-------
Enclosure A
U.S. Environmental Protection Agency
Clean Air Scientific Advisory Committee (CASAC)
Sulfur Oxides Primary NAAQS Review Panel
CASAC MEMBERS
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
PANEL MEMBERS
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
Did not participate in the July 30-31, 2008 review of the second draft ISA.
-------
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
SCIENCE ADVISORY BOARD STAFF
Dr. Holly Stallworth, Designated Federal Officer, Science Advisory Board Staff Office,
Washington, D.C.
11
-------
Enclosure B
Individual Comments on the Integrated Science Assessment for Sulfur Oxides -Health
Criteria (Second External Review Draft, May 2008) from Clean Air Scientific Advisory
Committee (CASAC) Sulfur Oxides Primary National Ambient Air Quality
Standards (NAAQS) Review Panel
Mr. EdAvol 2
Dr. John Balmes 5
Dr. Doug Crawford-Brown 9
Dr. Terry Gordon 13
Dr. Dale Hattis 15
Dr. Donna Kenski 18
Dr. Patrick Kinney 21
Dr. Steven Kleeberger 24
Dr. Timothy Larson 26
Dr. Kent Pinkerton 32
Dr. Edward M. Postlethwait 35
Dr. Jonathan M. Samet 37
Dr. Richard Schlesinger 40
Dr. Christian Seigneur 42
Dr. Lianne Sheppard 47
Dr. George Thurston 48
Dr. James Ultman 51
Dr. Ronald E. Wyzga 53
-------
Mr. Ed Avol
Comments on SOx ISA 2nd Draft
Question 1 (a modified Chapter 1 on the framework for causal determination)
The continuing EPA-and-Clean-Air-Act self-imposed restriction to consider individual
pollutant effects (in this case, SO2 from particulate SOx) makes the framework for
evaluation of the overall weight of evidence a challenging one, due to the common
sources of pollutant generation (and virtually no difference in the potential targets of
source emission reductions). The reality of having to adjust for the possible confounding
contributions of other pollutants makes definitive determinations of assigned causality a
messier proposition, since the likely presence of other ambient co-pollutants provides a
"smoke-screen" of sorts for the true stimulator (or stimulators) of response and effect.
The framework described in Chapter One, aside from the recurring general concern for an
archaic approach that attempts to evaluate health effects on a discrete pollutant-by-
pollutant basis in the face of ever-growing information about additive and synergistic
physical, chemical, and physiological interactions, seems a reasonable way to proceed.
Question 2 (an expanded Chapter 2 on atmospheric chemistry of SO?)
The additional sections about monitoring information are useful in developing an overall
appreciation for the historical monitoring network and nationally downward trend in SO2
exposure. My personal opinion regarding the inclusion of design criteria for station
placement is that there is more detail than necessary presented here, and most of it is
generic (to any acceptable monitoring site's placement).
The improved information presented does improve the overall assessment of the currently
available knowledge and is relevant to subsequent discussions of human health effects.
Question 3 (a revised Chapter 3 on Integrated Health Effects)
The chapter is still a long compendium of information (over 100 pages), but the summary
figures are very helpful in providing a "meta-view" of numerous studies. So much
information on such a range of health outcomes is presented that, at chapter's end, some
sort of distillation of what the chapter conclusions were would have been helpful. This
might be in the form of a table, listing category (morbidity, mortality, etc), SO2 exposure
(long-term or short-term), health outcome (lung function, symptoms, ED visits,
hospitalizations, etc), and judgment (causal, insufficient, etc), which would assist the
reader in interpreting the weight of evidence from the 100+ pages of data.. .which is
precisely what appears as Table 5.3 in Chapter 5! Accordingly, it might be worth
referring the reader to or providing some linkage to the Chapter 5 summary discussion.
Question 4 (Revisions to Chapter Four, Public Health Impacts)
This revised description of the public health impacts of SO2 exposure is substantively
improved, but could still profit from further clarification, focus, and specificity. The
over-arching questions to be addressed include the following: (1) what insights do current
data provide regarding the shape of the concentration-response function for SO2? (2) Is
-------
there evidence of a threshold of response? (3) Are there populations at risk? (4) If there
are populations at risk, who are they? To the credit of the ISA staff, these issues are
discussed, but they are not always clearly/succinctly summarized or brought to some final
resting point in the document.
For example, the presentation of concentration-response functions and the topic of
thresholds of response are understandably linked, but the presentation in the document is
inter-twined (in sections 4.1.1 to 4.1.3) and thus more difficult to follow than need be.
With regard to the question of concentration-response functions, the presented data from
clinical and epidemiological studies suggest there is support for a linear concentration-
response function claim (more supportive from the clinical chamber studies, less
convincing but somewhat supportive from the epi studies). With regard to a threshold of
response, the data are inconclusive. This discussion, particularly in Section 4.1.3, should
be separated into two sections (one on the concentration-response function, and the other
on threshold response).
The discussion of susceptible and vulnerable populations (Section 4.2) is also in need of
some improved specificity (but is admittedly improved over the previous version). Much
of the appropriate information is contained in the discussions, but structurally, the
respective sections do not always follow the specified title or deliver on the promised
topic. There remains some confusion about definitions of "susceptibility" and
"vulnerability". Reasonable working definitions are provided in the text (at the bottom of
p4-9 and top of p4-10), but then these are lumped together into a discussion of
"somewhat sensitive subgroups" a few lines later (lines 8-14, p4-10).
A discussion about the potential importance of genetic factors is presented in Section
4.2.2, but much of the two-page discussion is generically about air pollution and not
specifically about 862.
Section 4.3 (which needs to be re-titled to "Populations at Adverse Risk" or something,
since the current title is the chapter title) is a valuable perspective on how to think about
the multiple health outcomes being presented.
Section 4.3.2 (Estimation of Potential Numbers of Persons in At-Risk Susceptible
Population Groups in the US) doesn't quite deliver on its title promise, in that numbers
are estimated for only a few of the several susceptible sub-groups identified. Still, the
point is made that many people are potential at increased risk for exposure and possible
response. The concluding paragraph in the chapter (lines 10 through 18 on p4-23)
provides a declarative overview on the issue (in essence, that there are could be a
considerable public health impact, because there are large numbers of susceptible sub-
groups), but seems to fall short by not saying something about the likelihood of exposure
to levels likely to trigger the range of health outcomes identified.
(One final note - is the comment regarding Figure 4.6 "... demonstrating that the SO2-
related excess risk for asthma is, on average, 50% higher among children when compared
-------
to risk estimates that include all ages..." (lines 7-9, p4-23) over-reaching, given the error
bars, available data, and risk range [~RR 1.05-1.1]?)
Question 5 (revised Chapter 5 on Integrated Findings of Causality)
This chapter has a significant task and a large body of multiple outcome data from over
20 years to summarize. The summary Table 5.3 is especially useful, in distilling the
current judgments to brief paragraphs of conclusions for endpoints of potential health
interest.
The document provides an interesting series of comments regarding "previous
conclusions" and "current conclusions" in Table 5.3, but it is not clear to what "previous"
review these comments refer. (Presumably, these "previous" comments refer to the 1996
NAAQS SC>2 review, but one could also interpret this as a reference to an earlier version
of the ISA). This should be explicitly specified in the document.
The larger question here is the following: have the additional years of data, research, and
observations changed our judgment regarding the health effects of 862? The comments
in Table 5.3 regarding previous and current conclusions make it difficult to determine,
and should be simplified to reflect a clearer conclusion of where we are today.
-------
Dr. John Balmes
Comments on SOx ISA -- 2nd Draft
Charge Question 3: In the revision, we reduced redundancy, added summary sections
and reorganized Chapter 3. In addition, discussions on potential confounding by and
interactions with copollutants have been added. The 2nd draft ISA also includes
additional analyses of individual-level data from human clinical studies (Sections 3.1.3.
and 4.1.1) that builds upon the analysis included in the 1994 Supplement to the Second
Addendum. The toxicology sections were reorganized to focus on studies using more
relevant concentrations of SO2 and sections were added to better discuss mode of action
and potential particle-SO2 interactions. We are requesting CASAC review specifically
on these analyses as well as on the integration of the overall evidence from the human
clinical, animal lexicological, and epidemiological studies. 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?
GENERAL COMMENTS
The revised Chapter 3 is much improved by virtue of the reduced redundancy, added
summary sections, reorganization, and additional discussions involving co-pollutants.
The new analyses of human clinical studies in Sections 3.1.3 and 4.1.1 are helpful. In
general, the discussion of the results of the animal lexicological, controlled human
exposure, and epidemiological studies that have been reviewed is technically sound. The
major exception to this comment is the discussion of the effect of SO2 on lung function
responses to inhaled house dust mite allergen in asthmatic adults with pre-existing
sensitization to this allergen on page 3-31. As written, the draft indicates that SO2
"enhanced sensitization to house dust mite". This is incorrect. In fact, it would be
unethical to induce or enhance sensitization to a common aeroallergen in human subjects.
The integration of the animal lexicological, controlled human exposure, and
epidemiological studies that have been reviewed is also technically sound, balanced, and
clearly communicated with several exceptions on page 3-91. The summary statements
about the evidence in Sections 3.4.2.2 and 3.4.2.3 are too strong as worded. I have
suggested revised wording (see below in Specific Comments).
Perhaps my biggest concern about the current draft of Chapter 3 involves the references.
Many papers cited in the text are simply not listed in the References, and some citations
are obviously wrong (see some examples below in Specific Comments).
SPECIFIC COMMENTS
p. 3-13, line 28 The word "allergic" should be deleted from this sentence. SO2 is
too small of a molecule to elicit an allergic response on its own. The presence of
-------
eosinophils in bronchoalveolar lavage fluid does not mean that SO2 induced an allergic
response. In addition, the Gong et al. (2001) study is not listed in the References.
p. 3-30, line 11 The Gong et al. (2001) study is both not listed in the References
and is not a sheep study.
p. 3-31, lines 8-14 The two studies cited in this paragraph did not show that SO2
"enhanced sensitization to the allergen in asthmatic individuals." Rather, these two
studies showed enhanced airway responses to inhaled allergen in asthmatic individual
with preexisting sensitization to house dust mite.
p. 3-31, line 7 The word "preexisting" should be deleted from this sentence.
Increased morbidity with infection does not necessarily mean that there was preexisting
disease.
p. 3-32, lines 24-25 For clarity, the sentence should read "... and was also higher in the
winter months (mean 25.7 ppb [SD 15.8]) than in the summer months (mean 10.6 ppb
[SD 15.1]."
p. 3-39, line 11 The Ito et al. (2003) study is not listed in the References.
p. 3-39, line 13 This sentence should be revised as follows: "A study conducted in
New York City..." In addition, this study is not listed in the References.
p. 3-42, line 25 For clarity, this sentence should be revised as follows: "In
summary, only a few studies provide results for respiratory health outcomes other than
asthma and COPD, and these results are mixed."
p. 3-45, lines 14-15 I would delete "and to a more limited extent the human clinical
studies" from this sentence.
p. 3-46, lines 24-26 This sentence should be revised as follows: "These findings of
increased airway resistance are in concordance with the limited epidemiological study
results of Taggart et al. (1996) that showed SO2-induced increases in AUR among
asthmatic adults.
p. 3-46, line 31 I would add "non-elderly" before adults in this sentence.
p. 3-48, line 4 "bronchial responsiveness" should be changed to lung function
responses. A decrease in FEV1 is not necessarily a bronchial response.
p. 3-49, lines 1-3 The dog study discussed in the 1982 AQCD should be cited.
p. 3-51, line 25 and p. 3-52, line 1 For clarity, this statement should be split into two
sentence as follows: "... of sulfite^J)ose-dependent decreases... and above were also
observed."
-------
p. 3-53, line 8 The Peters et al. (1996) paper cited appears to be the wrong
reference for the Hong Kong study.
p. 3-58, line 4 The Routledge et al. (2006) study cited is not listed in the
References
p. 3-59, line 3 The Gold et al. studies cited are not listed in the References.
p. 3-61, line 4 The Peters et al. (2000) study cited is not listed in the References.
p. 3-61, line 15 The Dockery et al. (2005) study cited is not listed in the
References.
p. 3-64, line 1 The Liao et al. (2005) study cited is not listed in the References.
p. 3-67, line 1 The Ballester et al. 2001 study cited is not listed in the References.
p. 3-75, line 4 ".. .the association between air pollutants and mortality was
examined..."
p. 3-76, line 2 The Le Tertre et al (2002) and Ballester et al. (2002) studies cited
are not listed in the References.
p. 3-77, line 26 and p. 3-78, line 1 For clarity, I would revise these two sentences into
one as follows: "..health effects studies in Asia that summarized the results from
mortality and hospital admission studies published in the peer-reviewed scientific
literature from 1980 through 2003."
p. 3-91, line 2 I would revise this statement as follows: "... studies do not provide
sufficient evidence to infer that long-term exposure to ambient SO2 has a detrimental
effect on lung function."
p. 3-91, lines 18-19 Similarly, I would revise this statement as follows: "... studies do
not provide sufficient evidence to infer that long-term exposure to ambient SO2 causes
prolonged effects on lung morphology.
p. 3-94, line 11 I would add "ambient" before SO2 in this sentence.
p. 3-97, line 28 "Furthermore, the epidemiological studies do not provide..."
p. 3-102, line 19 The Liu et al. (2003) study cited is not listed in the References.
p. 3-102, lines26-27 For clarity, I would revise this sentence as follows: "The most
robust association with intrauterine mortality was observed for an index of three gaseous
pollutantsNO2, SO2, CO)"
-------
p. 3-105, line 2 The Japanese study discussed in the 1986 Secondary Addendum
should be cited.
-------
Dr. Doug Crawford-Brown
Comments on SOx ISA - 2nd Draft
My comments here focus largely on Chapters 4 and 5, although it was necessary to
review the earlier chapters as well because Chapter 5 especially draws on materials from
those earlier chapters. It is clear to me that the EPA is getting better at this process of
creating an ISA. The document is well organized; it is easy to follow the reasoning used
by the authors; and they appear to me to have reached the appropriate summary
conclusions on the ability of SOx to produce effects even at levels below the current
NAAQS, at least for short-term exposures. The document overall provides an adequate
basis "to provide support for future risk, exposure and policy assessments", in the sense
that it establishes the existence of effects in policy-relevant ranges of exposure, identifies
issues associated with exposure, and identifies sensitive subpopulations. It would not be
possible to use the document alone as the basis for performing an actual risk assessment
because there is inadequate assessment of exposures nationally or in specific
subpopulations. But I presume such a formal, quantitative risk assessment would be the
task of the Exposure and Risk Assessment document. Still, I continue to have trouble
understanding the role of the IS As in meeting the needs of risk assessment. This
document continues the pattern of being an inadequate basis for such a quantitative
approach, even if it does systematically review the evidence on which such a quantitative
assessment might ultimately be based.
My specific comments are:
1. I like the formulation of the questions on page 1-2. They seem to be a refinement of
these questions in previous ISA documents, recognizing the need not to just consider new
information but to integrate new information into past assessments.
2. On page 1-5, the authors note that associations are not sufficient for proof of a causal
relationship. Later they mention that epidemiological studies provide evidence of an
association. Still later in the document, they use the epidemiological evidence as a basis
for concluding that there is a causal relationship. This is not consistent, although I think
the inconsistency begins with the first claim: that associations do not prove causality
sufficiently. This needs some nuance, as associations ARE an important component of
causal reasoning, even if not fully compelling. I disagree with views that associations are
not relevant in claims of causality - they can in fact be compelling, rational evidence if
sufficiently well established and reliable.
3. In the figure on page 1-6,1 note that some aspects of the bottom half of the figure,
such as lifestyle, CAN be influenced by community norms, and so are not strictly
individual-level factors.
4. The topic of multifactorial causation on page 1-9 is interesting. But there are real
questions as to whether policy can deal with such complex mixtures, leading perhaps to
-------
different standards in different regions where the mix differs. The common basis could be
overall risk, not the allowed concentration of any one contaminant, but there will still be
significant issues of equitable treatment of the class of emitters in different regions if the
mix differs between those regions (which in turn would require different levels of control
on the same class of emitters in different regions).
5. On page 2-39, the authors state that exposure misclassification may result if total
human exposure is not disaggregated. I am not sure WHICH exposure is then
misclassified. Is it the exposure to ambient levels? If one is using total exposure to SOx,
perhaps through personal monitors, in epidemiological studies, there is no exposure
misclassification due to a lack of disaggregation. There WOULD be misclassification if
one is trying to get at the effect of ambient, policy-relevant exposures alone. I assume
they don't mean misclassification of total exposure.
6. Table 2-9 presents a number of studies, and then the authors appear to settle on 0.13 as
the slope. This value is the highest one in the table by a factor of 2 to 3, and I don't feel
the authors have explained adequately why they chose this particular result from amongst
those in the table. I'm sure there is a rationale, but it doesn't come across in the writing
and so there will be suspicion that the choice was based on conservatism rather than this
being the best study.
7. The discussion of Berkson type errors on page 2-53 is correct, although I don't see
where the authors eventually conclude whether the errors in the specific studies used here
are or are not Berkson type. I presume they assume the errors will lead to
misclassification and bias towards the null, but I can't see this stated clearly. If there IS
bias towards the null, they should state this as a final conclusion, which then has
implications for application of any slope factors and the degree to which a standard is
health protective with a margin of safety.
8. On page 2-54, the authors have not noted that the slope factor can be incorrectly
increased if location near a source is correlated with sensitivity. For example, if poverty
causes sensitivity due to poor nutrition, and land prices decline the closer one gets to a
major source, then exposure and sensitivity are correlated and beta will be increased
above that found when this correlation is not present.
9. In the section on Dosimetry, the authors have summarized the studies adequately.
However, there remains the problem of how SOx gets into the deep lung given the rapid
absorption in the nasal passages, or how its movement into the surface of hygroscopic
particles might affect this penetration. I came away from the chapter not quite clear
whether the authors were saying penetration to the deep lung is insignificant, significant
only for mouth breathing, or unknown quantitatively.
10. On page 2-61, the middle paragraph seems inconsistent. How can uptake in a 3 year
old be slightly higher than for an adult but uptake per unit surface area be the same, given
the smaller surface area in a 3 year old? It seems to me a higher uptake in a 3 year old
10
-------
would produce an even larger difference in uptake per unit surface area. Or could the
authors be in some sense conflating uptake fraction and total uptake?
11. The figures in Chapter 3 are good and useful. It is especially good that they are in a
consistent format. The authors could improve the text by describing how the studies
could be combined quantitatively or at least qualitatively. Otherwise, the reader is left
with just "eye-balling" the graphs for a mean value - or choosing a single study - IF
quantitative analysis is needed. Is meta-analysis not appropriate, and if not, why not? But
perhaps the figures are there only to suggest a robust claim of an association. Still, the
authors should state that if it is the case. The reader also needs to be told whether the
studies in the figures represent a full sampling of such studies, or have been selected from
a larger set to form a subset - this subset would then need to be defined.
The following questions are related to Charge Question 4:
12. In Chapter 4,1 continue to have a problem with the ISAs separating exposure levels
and sensitivity in determining the public health impact. The authors do a good job of
describing the sensitive subpopulation, and then of estimating the size. It is evident that
there are a significant number of people who fall into this category. But having sensitive
people present does not equate to a public health impact unless exposures are sufficient,
and this chapter makes no attempt to quantify the exposures to these people. So I am not
convinced that tables such as 4-2 provide a quantitative measure of public health impact.
13. On page 4-1, there is mention of a population-response threshold. I don't know what
this term means. Populations don't have thresholds, individuals do and these thresholds
are distributed. There are several places in this chapter where the authors make it sound
as if there are population characteristics such as thresholds that can be found in studies,
and I don't believe there are. The belief in population thresholds also is inconsistent with
earlier claims in the chapter that intersubject variability is one of the causes of linearity of
population response at low exposures.
14. I don't understand why the exposure-response section is not placed into Chapter 3.
Surely this is both important information for that chapter, and is in fact the MOST
important information for understanding the effect of any lowering of the NAAQS.
15. It is unfortunate that the clinical E-R curves do not extend down into the region of
policy interest. Figures 4-2 and 4-3 suggest there may be some leveling of the curves in
the lower region of exposure, which has significant implications for the incremental or
marginal effectiveness of policies that push exposures below these levels.
16. It would be useful to show the existing NAAQS level on these E-R curves, so the
reader can see where extrapolation is needed. This is particularly important because the
units of the X axis change between several of the figures.
11
-------
17. The complexity of the curve in Figure 4-5 is intriguing, as we find the same thing in
radiation dose-response curves. It usually occurs when there are competing beneficial and
detrimental effects at low exposures, due to compensatory mechanisms kicking in.
18. On pages 4-9 and 4-10, a susceptible subpopulation is defined as one that might
show effects below concentrations needed in the general population. This is much too
vague of a statement, as it applies to any individual whose individual threshold is below
the population median (50% of the population therefore being susceptible). And again,
there IS no general population response, only the aggregate response of individuals with a
distribution of thresholds. Better definitions of susceptible and sensitive are needed, ones
related in some way to percentiles of the intersubject variability distributions and some
notion of biomodality of these distributions.
Based on Comments 12-18,1 judge that Chapter 4 is better overall than the previous
draft, and is nicely organized for the kinds of information presented. But I continue to
have problems with the way in which public health impact is quantified. I don't believe
this chapter provides quantitative information on actual impacts, but rather the
POTENTIAL for impacts if exposures are sufficient.
19. Figure 4-6 suffers from having a wide range of ages in the young group (0-14 years).
Is there no way to narrow in on the much younger ages where sensitivity seems
especially pronounced?
The following comments apply to Charge Question 5:
20. Chapter 5 continues the pattern of having little formal framework for integrating
information. The EPA seems to cite a different source for frameworks in each ISA (here
it is the NAS Institute of Medicine Report). I don't have an alternative to offer, so just
believe it would be best if the authors just state the principles they are using directly and
not rely on a particular source (requiring the reader to go back to those sources for
clarification).
21. On page 5-2, line 20, a causal relationship is inferred (reasonably), but needs to be
accompanied by a statement of the levels of exposure at which the relationship applies.
The existing statement is too broad. There are causal relationships between exposure and
adverse effect for all things in the world, at some level of exposure.
22. The discussion of uncertainty throughout is inadequate. It could lifted out of here and
placed down into any of NAAQS ISAs, it is so generic. There needs to be a better
summary of the uncertainty and its implications for specific judgments that might be
required in exposure and risk assessments.
Overall, Chapter 5 provides a good compendium of information that should prove useful
for the exposure and risk assessment stages. It adequately summarizes causal claims,
although I don't believe it properly places caveats with respect to the exposure levels at
which these causal claims are strong.
12
-------
Dr. Terry Gordon
Comments on SOx 2nd Draft ISA
General Comments:
This is an excellent draft ISA and for the most part is clear and logical. The
authors have made logical connections between the epidemiology, clinical, and animal
studies and arrived at fair conclusions regarding the adverse health effects associated with
ambient exposure to SC>2. Also, the order of the sections in each Chapter is very good.
Perhaps, NCEA should formalize the order and format of sections so that future ISA's
follow this order (to avoid everyone's suggestion to change the order to their individual
preferences).
Charge Question 3: "In the revision, we reduced redundancy, added summary sections
and reorganized Chapter 3. In addition, discussions on potential confounding by and
interactions with copollutants have been added. The 2nd draft ISA also includes
additional analyses of individual-level data from human clinical studies (Sections 3.1.3.
and 4.1.1) that builds upon the analysis included in the 1994 Supplement to the Second
Addendum. The toxicology sections were reorganized to focus on studies using more
relevant concentrations of SO2 and sections were added to better discuss mode of action
and potential particle-SO2 interactions. We are requesting CASAC review specifically
on these analyses as well as on the integration of the overall evidence from the human
clinical, animal lexicological, and epidemiological studies."
The reduction in redundancy and the addition of summary sections has aided in
making this Chapter clear and concise. The authors have reduced the animal studies so
that a true picture (of their contribution to understanding the biologic plausibility of the
adverse effects of SO2) is now presented. The addition of the toxicology section that
discusses potential particle-gas interactions and enhanced responses in exposed animals is
very appropriate in the level of detail and its relevance to the ISA.
Charge Questions 4 & 5: The revisions are excellent and made the chapters very readable
with minimal overlap with Chapter 3 and each other. The dose-response evaluations are
appropriate, but the final conclusions in Chapter 5 might be worded more strongly so that
the REA working group has more guidance as to which health endpoints and data sets
(epidemiology and/or clinical) are appropriate for risk evaluations.
Detailed Comments:
Page 1-1, line 13 - While this statement is probably true for sulfite, given the added
section(s) on particle-sulfur dioxide interactions, it seems premature to conclude on the
first page of the ISA that sulfuric acid (potentially layered on particles) are present in
insignificant concentrations in the ambient environment.
Page 1-5, line 11 - typo: data complement
Page 2-44, line 11 - Chao ref is missing here (2001 ref?).
13
-------
Page 2-52, lines 5-16 - This section is a bit redundant with similar discussions on page 2-
47 lines 7 & 8, page 2-48 lines 1-5, and page 2-42 lines 19-28. Possibly combine in one
place?
Page 3-6, lines 3 and on - Should a brief discussion of sulfuric acid be added to the
discussion of sulfite chemistry/interactions/metabolism?
Page 3-7, line 9 - Because line 5 mentions 'cholinergic' pathways, saying that
acetylcholine is 'also thought to be involved....' is incorrect.
Page 3-12, figure 3-1 - Add 'provocative' to the definition of PC(SO2).
Page 3-14, lines 19-30 (and onto the next page) - This section should be condensed to a
couple sentences or cut because of the questionable relevance of 100-2000 ppm sulfur
dioxide even if the dosing was briefer in vitro.
Page 3-27, line 8 - Please add refs for the 'several other studies..
Page 3-31, line 5 - Insert 'pseudo-' before 'measure of airway obstruction'
Page 3-37, Figures 3-6 & 3-7 - Would a meta-analysis be appropriate for these studies?
Also, The visual separation of the ED visits and the Admissions studies should be more
distinctly labeled.
Page 3-39, line 13 - 'study conducted in ??' Where?
Page 3-40, line 13 - A positive association for the Lee, 2002 study is not presented in
Figure 3-7. The only Lee study in the figure has a negative association (2006). Typos?
Page 3-50, line 15 - Chen, 1992?
Page 3-55, line 22 - Is this statement true regarding adults? Page 3-25 says inconclusive
evidence. Should the text read 'older adults'?
Page 3-57, line 1 - Should the word 'peak' be inserted before 'ambient'?
Page 3-57, line 18 - What are 'consequent cardiac deaths'?
Page 3-83, line 13 - 'and/or' is unclear
Page 3-85, lines 4-10 - |ig/m3 should be converted to ppm for ease of reading.
Page 3-92, line 10 - This statement about 3 other recent studies should be deleted or
expanded/referenced.
Page 3-99 - This section is nicely written and clearly hits the right level of discussion of
high dose studies and their quasi-relevance.
Page 4-3, figure 4-1 - Inclusion of 0.10 and 0.50 in the bar graph implies that zero
percentage changes occurred vs. the actual absence of data at those exposure
concentrations.
Page 4-6, line 8 - This sentence is not consistent with figure 4-5 which shows no real
change at 10 ppb.
Page 5-2, line 1 - Is this sentence accurate? 'west to east' implies a gradient from the
west coast to the east coast.
Page 5-10, 5-11 - A nicely written evaluation.
Page R-24 - Zeger ref has a typo on the pages.
14
-------
Dr. Dale Hattis
Comments on SOx ISA - 2nd Draft
Response to question 4 and to some extent question 3 on the analysis of individual-level
clinical concentration-response data for the short term bronchoconstriction response to
SO2.
I was encouraged that the ISA authors included empirical information from three papers
by Linn et al. (1987, 1988, and 1990) that are similar to the Horstman et al. individual
data for these responses that I analyzed in my comments to the earlier draft of the ISA.
Unfortunately there does not seem to have been any effort to compare the risk assessment
implications of the Linn et al. data with those of the Horstman data, let alone put them in
to a combined analysis for derivation of an overall estimate of the concentration-response
function for asthmatics. The concentration-response section of the REA also lists several
other sources of individual data on similar responses that have not been summarized or
analyzed as of yet.
Three types of information can readily be extracted from the existing body of clinical
observations: The ED50 for particular responses, the breadth of the population
distriubiton of thresholds, and the form of the distribution (i.e. does it appear to be
lognormal, or some more complex shape?)
For reference, the Horstman et al. paper gave individual SO2 threshold concentrations for
27 asthmatics for a doubling of specific airway resistance. These data are reasonably, but
not perfectly compatible with a lognormal distribution (Figure 1). (In this figure, the
straight line represents a fitted lognormal distribution for the data. The slight departures
of the data above the line on both the low and higher-concentration ends of the
distribution may suggest some tendency toward bimodality in the underlying distribution
of individual thresholds, but these data are not sufficient to make a firm conclusion on
that point). By averaging the logarithms of the individual thresholds for this effect we
find an ED50 (geometric mean dose expected to produce the response in 50% of an
exposed asthmatic concentration) of 1,040 ppb, with 90% confidence limits of 739-1475
ppb. From the standard deviation of the logarithms of the individual values we find a
geometric standard deviation for interindividual variability in the thresholds of 2.38—
meaning that based on these data, 90% of the individual thresholds in a similar
population of asthmatics should be expected to be between 250 and 4340 ppb.
The newly presented information from the Linn et al papers for the same response are
unfortunately not given as individual values, but essentially in histogram form—the
number of asthmatics studied who respond at a few discreet dose levels. I have analyzed
these assuming a lognormal distribution of individual thresholds using an Excel program
adapted from Haas (1994) and used extensively in my prior work on human
interindividual variability in susceptibility (Hattis et al. 2001). The data provided from
the Linn et al. papers suggest an ED50 for the doubling of specific airway resistance in
exercising asthmatics of about 580 ppb—a little lower, but probably not significantly
lower than the range indicated by the Horstman et al. data. (I did not assess the
15
-------
confidence limits for the ED50 in this case) The Linn et al. data also indicate a slightly
greater amount of interindividual variability than those of Horstman—a geometric
standard deviation of 2.85 with 90% confidence limits of 2.10-5.75. Together, if we
assume a lognormal distribution of individual thresholds as before, the central estimate of
interindividual variability and the central estimate of the ED50 from the Linn et al data
would indicate that 95% of exercising asthmatics should have this response at between
104 and 3250 ppb. Unfortunately, there is reason to doubt that the distribution is
perfectly lognormal—a goodness of fit test indicates a barely statistically significant
departure of the data from expectations from a lognormal model at P < .05.
There are a couple of ways to combine these data for an aggregate analysis. One simple
way that I did was to add the Horstman observations where they would fall in the
histogram-like data from the Linn et al. papers. When this is done, the ED50 indicated
by the combined data is just under 700 ppb and the geometric standard deviation for
interindividual variability is 2.52 with 90% confidence limits of 2.01-3.83. The revised
central estimate ED50, combined with the central estimate of the geometric standard
deviation would indicate that 90% of exercising asthmatics should respond between 150
and 320 ppb. Again, however, the combined data depart from expectations of a
lognormal distribution model at a P level of slightly less than 0.02. In further exploration
of the combined data (including experiments of other research groups) I would
recommend an attempt to fit a mixture of two lognormal distributions. I could not
attempt this in this analysis, however because a mixture of two lognormals involves
estimation of 5 parameters (two means, two standard deviations and a parameter to
measure the fraction of the population that belongs to each component distribution) and
the Linn et al. data are aggregated only into 4 ranges. There is likely to be an infinite
number of perfect-fit solutions when one uses only 4 data points to estimate 5 parameters.
If it is of interest to EPA staff, I will provide the Excel spreadsheet showing these
calculations.
References
Haas, C. N. "Dose Response Analysis Using Spreadsheets" Risk Analysis 14:1097-1100
(1994).
Hattis, D., Russ, A., Goble, R., Banati, P., and Chu, M. "Human Interindividual
Variability in Susceptibility to Airborne Particles," Risk Analysis, Vol. 21(4), pp. 585-
599 (2001).
16
-------
Figure 1
Lognormal Plot of the Distribution of Individual Sensititivities
(SO2 Concentrations Needed to Double Specific Airway
Resistance) For 27 Exercising Asthmatics (Horstman et al. 1986)
S
o.
a.
^Ofi
o
-
v = 0.0189+ 0.374x RA2 = 0.934
Z-Score
17
-------
Dr. Donna Kenski
Comments on SOx ISA - 2nd Draft
General comments: The revisions to Chapter 2 have improved it substantially. In
particular, Sections 2.5.3 and 2.5.4 on exposure measurement errors were improved and
clarified. The added tables and discussion were helpful. The added information on 5-
minute averages was a good start, but lacked enough detail to form adequate basis for
decision-making. The dismissal of these data because of the voluntary nature of the
reporting and their variability over time and space is troubling. In light of the number of
health studies showing effects at 5- to 15-minute exposure times, the importance of the
information for the Risk and Exposure Assessment, and given EPA's request that the
states collect this data, it seems unreasonable to dismiss it in this document without even
a rudimentary summary of the data. The REA goes on to glean important information
from the relationship of the 5-minute peaks to hourly means, so the very limited
discussion in this document is surprising. This particular shortcoming was not apparent
in the previous version of the ISA, before the REA had been drafted. The removal of the
NOx information from Annex B was a good decision; it makes it much easier to find the
relevant data for SO2. Some specific comments follow:
2-1 line 13 The initial phrase, "Industrial emissions..." is somewhat
mischaracterized; perhaps the authors meant anthropogenic emissions? Utility emissions
are usually regarded separately from industrial emissions.
2-2 1st paragraph. This section still needs a simple sentence quantifying biogenic or
natural sources. It is unacceptable to spend 2 paragraphs describing natural
sources and still not quantify them except as 'small', which could be 10% or 1%
or 0.01%. Even Table B-3 doesn't do so; the reader is left to infer that such
emissions are less than 0.00 Tg/yr.
2.2. Line 12 Refers to Table B-6, which does not exist. It doesn't look like this
data is available in the revised Annex.
2.3 line 14 Reference to Annex B.5 should be B.6
2.8 Sec. 2.4.1 The addition of the monitor siting criteria was helpful. The open-
path information is probably unnecessary, however. The sampling and analysis
discussion didn't describe any open path instruments for SO2 and since these are a tiny
fraction of the reporting network, if there are any at all, it could be deleted without harm
to the rest of the discussion.
2.11 The new figures 2.1-2.6 are a useful addition. However, the rationale for
choosing these 6 areas for focus needs to be clearly stated. Was it because they have the
most monitors, or the highest concentrations, or the most spatial heterogeneity or
homogeneity?
18
-------
2-11 line 16 ... as well as the subset of monitors in Cuyahoga County.
2-19 line 9 'Overrepresentation'sounds judgmental. 'Predominance'might be a
better word choice
2-23 Fig. 2-11 At the risk of sounding like a broken record, I will point out again
that this figure is a poor depiction of diurnal cycles. If the point is to show diel variation,
as the text says, then the variation in the median (or some higher percentile, if all the
medians are zero) is of much more interest than the outliers, which is all one can see on
this plot. Plot the medians on a scale that can be read, as in Figs. 2-12 - 2-15, rather than
emphasizing the meaningless noise in the outliers. Maybe a log scale would work better.
The plot is not useful as currently configured. Also, it's not clear what the caption means
by 'all cities in focus.' It seems to imply that these are the cities included in the Figs.
2.1-2.6, but the text says it summarizes all the data in AQS, so that should be clarified.
2-26 The description of the 5-minute sample data is still lacking. The text and Tables
2-6 and 2-7 don't agree. For example, Table 2-6 has 80 monitors reporting 5-min
averages and Table 2-7 lists 16 monitors reporting all 12 5-minute values. The
text reports that 108 monitors reported 5-min averages and 15 monitors reported
all 12 values. These need to be reconciled.
2-34 line 6 Annex Section B-6 should be B-3
2-46 line 27 A monitoring site at an elevation of 250 m above street level? Is this
correct?
2-47 line 1 ...should'outdoor'be'personal'instead? Ambient concentrations are
outdoor concentrations.
2-56 line 19 or-> for
2-63 line 5 delete 'of
5-2 lines 5-6 Transport of SO2 plumes from Asia and Europe really wasn't
discussed in the previous sections, so this statement doesn't belong in the conclusions
unless it is supported by the earlier text.
5-2 lines 15-18 Here's another dismissal of the 5-min data as unimportant for
determining concentrations and exposures at short time durations. This just doesn't make
sense. Perhaps the data are insufficient to make sweeping generalizations about national
exposure, but they certainly are useful and valid, and their temporal variability is exactly
what makes them valuable and why we collected them in the first place. A more
reasonable statement here would be that the existing 5-minute data are not adequate to
assess exposure on a national scale - but then, neither are the hourly data.
5-2 line 24 remove the 7 after exposures
19
-------
5-3 lines 8-11 This statement is correct, but would be much stronger if the 5-
minute SO2 concentrations had actually been summarized in this document.
20
-------
Dr. Patrick Kinney
Comments on SOx ISA - 2nd Draft
One general concern I have relates to the way the ISA interprets the public health
relevance of concentrations below those studied in the human clinical studies, i.e., 0.2-0.4
ppm. As noted in the ISA, these studies look at effects in groups of people who do not
represent the full range of susceptibilities present in the general population. In addition,
clinical studies study very small numbers of subjects as compared with the population
exposed to ambient concentrations. These two issues limit the power of clinical studies
to quantify exposure-response relationships at ambient-relevant concentrations. Given
this, it is incorrect to assume, as EPA seems to do in the current ISA, that the lowest-
observed effect concentration represents a threshold below which no effects occur. After
all, we are concerned with protecting the US population, which contains over six orders
of magnitude more people than have been studied in a typical clinical study. A better
way to interpret the exposure-response results from clinical studies is that they provide
points on a exposure-response function that extends down to zero ppb. This concept
should be discussed in the ISA, providing a better foundation for its application in the
health risk assessment in the REA. It also would help to bridge the large gap that
currently exists between the exposure levels at which responses/associations are observed
in clinical/epidemiologic studies.
p. 1-7, line 12: change "occur" to "exist"
p. 1-7, line 24: change "with" to "with respect to"
p. 1-8, line 2: it should be noted as well that control for confounders, whether by
adjustment or stratification, is only successful when the confounder is well measured.
This point is often overlooked.
p. 1-8, first full paragraph: Needs to be reworked. It's not clear whether you're talking
about confounders, the exposure of interest, or both. It seems to change from one
sentence to the next.
p. 1-8, second full paragraph: To the lay reader, it may be unclear what you mean by
covariates as opposed to confounders here. Also, this section on confounders would
benefit from a clear definition of confounding at the outset.
p. 1-13, line 13, insert "controlled human exposure studies," before "epidemiological"
p. 1-14, box on causality categories, in the row labeled "inadequate to infer the
presence...", insert "quantity," before "quality"
p. 2-8, section 2.4. This section should start off with an intro paragraph stating purpose
and approach of the section. The very technical minutia of the initial text is off-putting
as written.
p. 2-8, line 17: need to insert definitions of the three geographic scales in terms of
kilometer ranges or such.
p. 2-9 through 2-11, line 2: this whole section reads like a guidance manual for setting up
a monitoring station. I don't think we need that here. Can't we assume that existing
monitors satisfy these criteria, and move on? Once sentence would be enough to say
that.
p. 2-9, line 3: define "monitoring path", or better yet, delete this whole discussion.
21
-------
p. 2-10, line 6:1 don't think the "open path" analyzer was defined or described earlier.
Please add that. How relevant is this? Is this an EPA equivalent method? If not, why
discuss?
p. 2-11, section 2.4.2: It would be better to start off by presenting descriptive information
and summary statistics on the national data (e.g., table 2-4) rather than focusing on six
particular states. Rationale for these six states is not sufficiently clear. Please expand.
p. 2-21, Figure 2-10: a similar map showing emissions at US ports would be a very
useful addition.
p. 2-21, line 10: what is "diel"?
p. 2-22, Table 2-4: This table should be referred to repeatedly throughout subsequent
sections, especially chapter 4.
p. 2-23, figure 2-11: not clear what is meant by "cities in focus"
p. 2-32 and vicinity: I was surprised not to see any data from the 5-min so2 results. Why
mention it here if no data are to be shown? Must explain why data are not shown.
p. 2-38 and later in Section 2.5: The uncertainties are discussed extensively, but are never
summarized quantitatively in any way.
p. 2-57, line 9 and elsewhere: should be beta hat, not beta.
p. 2-57, line 19: add text on effect of measurement error on SE(beta hat)
p. 2-58, line 5: spelling of physicochemical
p. 3-25, line 15: please provide a quantitative measure of how prevalent these
observations are. e.g., what fraction of 10 min avgs are above 0.4? the term "sometimes"
could be mis-interpreted as something like 10-20%. Reality is probably many orders of
magnitude lower than that.
p. 3-39, line 13, sentence starting "a study conducted" needs grammar fix.
p. 3-43, line 4: state range of ppb for which excess risk is calculated here, and put this
range into the context of SD of ambient 24-hour averages.
p. 3-44, figure 3-8 and following text: I'm troubled that there are no studies reported here
that included PM2.5 as a co-pollutant. This omission is an important caveat that should
constrain any conclusions regarding robustness. Especially so given the dismissal of the
long-term mortality results for SO2 on the basis of concerns over sulfate PM2.5
confounding.
p. 3-45, lines 13-18: This seems too glib. I don't agree that coherence and biological
plausibility are satisfied when comparing chamber results at 200-600 ppb and epi results
are for ambient levels two orders of magnitude lower. Need to include this caveat.
p. 3-55, lines 16 to 25: This section needs to acknowledge the exposure levels observed
on average in epi studies, in metrics equivalent to those used in ch amber studies, e.g., 1-
hr averages - noting means, sds, 95th percentiles etc. This is done for the chamber
results but for some reason is ignored when describing the epi findings. Table 2-4 is the
relevant reference.
p. 4-1, sentence on lines 20-23. This sentence is false. It is not true that most epi studies
evaluate whether there is a threshold. Few if any epi studies evaluate that question
directly. Please edit.
p. 4-3, Figure 4-1: indicate in footnote which of these results are statistically significant.
p. 4-4, section 4.1.2: in linking the chamber findings with the epi findings, it is important
to compare the dosimetry in at least a semi-quantitative way, taking into account the
differences in exposure concentrations and activity levels.
22
-------
p. 4-4, line 10: change "was" to "could not be distinguished from" The key point is that
there is very limited power to formally test whether something is linear or not. Failure to
reject linearity isn't equivalent to "observing that the relationship was linear"
p. 4-5, lines 8-10: this is a good example of the kind of concentration cross-referencing
that I was looking for earlier.
p. 4-6, Figure 4-4: tell us which of these results are statistically significant, or better yet,
provide confidence intervals at each point on the curve. OK to limit to one lag to
simplify display since they all say roughly same thing.
p. 4-9, line 4: insert quantitative LOD (e.g., 3 ppb) in parenthesis after "detection"
p. 4-9, lines 13-14: again the issue of "observing that the relationship was linear"
p. 4-17, Figure 4-6: throughout the document, EPA concludes that ED visits and
hospitalizations are associated with SO2 for children and older adults. This doesn't
come through very clearly from this figure. Can the results for younger adults be added?
Can histograms be created of the RRs within each age range? This would provide
stronger backup for EPA's conclusions about age-related effect modification.
p. 4-19, line 19: Edit grammar. As stated, it suggests that most studies examine effect
modification by SES, which is clearly untrue.
p. 4-22, lines 19-27: need to include the fact that these findings are observed only with
heavy exercise. Then later you need to estimate fraction of asthmatics who spend time at
those exertion levels, and also the fraction of time that people are exposed to these
ambient levels. Both are needed for risk assessment discussions that follow.
p. 4-23, line 9: this is obvious place to insert 1-2 paragraphs discussing the fraction of
relevant exposure levels and exertion levels experienced by sensitive subgroups.
p. 4-23, line 16: risk only occurs with exposure, so need to discuss exposures here too.
23
-------
Dr. Steven Kleeberger
Comments on SOx 2nd Draft ISA
3. In the revision, we reduced redundancy, added summary sections and reorganized
Chapter 3. In addition, discussions on potential confounding by and interactions with
copollutants have been added. The 2nd draft ISA also includes additional analyses of
individual-level data from human clinical studies (Sections 3.1.3. and 4.1.1) that builds
upon the analysis included in the 1994 Supplement to the Second Addendum. The
toxicology sections were reorganized to focus on studies using more relevant
concentrations of SCh and sections were added to better discuss mode of action and
potential particle-SCh interactions. We are requesting CASAC review specifically on
these analyses as well as on the integration of the overall evidence from the human
clinical, animal lexicological, and epidemiological studies.
Overall, the reorganized Chapter 3 is greatly improved over the previous version.
In particular, the following aspects of the Chapter were useful:
• Summary of findings from the previous review - these provide much needed and
useful reference points to enable evaluation of the more current investigations.
• Redundancy was largely eliminated, though some repetition remains that could be
edited or removed.
• Additional individual-level data (i.e. specific numbers of individuals in response
groups) enabled better evaluation of between-individual variation and the impact
on interpretation of results.
• Removal of animal studies that used SO2 concentrations that had questionable
environmental relevance was appropriate, though some remain that likely could
also be removed.
• The end-section summaries were very useful and largely presented a reasonable
evaluation of the literature with respect to weight of evidence for causal
determination.
• Incorporation of the effects of SOx intervention and subsequent change in health
outcomes into evaluation of causality was appropriate and useful.
Additional Comments:
While it may be argued that SO2/particle interactions should be included in the
ISA document, the case was made in the REA that the current review of the SO2
NAAQS would focus on gaseous species of sulfur oxides and not consider health effects
directly associated with particulate sulfur oxide species. These species and their health
effects instead will be reviewed in the PM document. However, one of the sections in the
24
-------
SO2 document (3.1.5) includes effects of SO2 layered on particles and sulfite aerosols. I
did not find the section fit particularly well with the rest of the document and raised
additional questions that could not be answered, or may have been answered in the PM
document. My recommendation would be to eliminate this section.
The REA clearly differentiates between long-term exposure (months to years),
short-term exposures (hours to days), and short-term peak exposures (5-10 minutes)
throughout the document, especially beginning with the Health Effects chapter (part 4).
The ISA differentiates primarily between long-term exposure and short-term exposures,
with the short-term peak exposures embedded in the latter (e.g. section 3.1.3). An
introductory statement or small paragraph placed at the beginning of the major sections
of Chapter 3 (e.g. 3.1) which define the distinction may help the reader to place the
studies in the intended perspective.
Minor
P 3-4, line 5: delete "a" from "presenting a more details"?
P 3-13, lines 26 and 28: is it correct to suggest that SO2 may elicit an allergic
inflammatory response? SO2 is not an allergen.
P 3-17, lines 9-21. I am not sure that this section really provides "biological plausibility
for the effects of SO2 on respiratory symptoms in humans". The relevance of MUC5AC
expression induced by SO2 using in vitro models and in rats exposed repeatedly to 2 ppm
SO2 to respiratory symptoms in humans is questionable, and presenting these preliminary
studies is not necessary.
P 3-103, line 10. "is" should be "are"?
25
-------
Dr. Timothy Larson
Comments on SOx ISA 2nd Draft
Chapter 2:
Section 2.4 is reorganized and shortened with a more appropriate focus on interurban
variability. Tables 2-4 and 2-5 are particularly useful in this regard. The information on
correlations with other pollutants is obviously limited by the lack of co-located monitors.
The discussion of the available 5-minute data is an important addition to the chapter, but
it is very brief, given the importance of this data in the REA. We are left with a
statement that these data are "very difficult to use precisely", without knowing exactly
what that means. Do the authors mean the measurements are difficult to use because they
are inaccurate or that they are accurate but are highly variable across time and space?
Based on my prior review comments, I note that on page 2-52 the statement is still made
that "when personal exposure concentrations are above detection limits, a reasonably
strong association is observed between personal exposures and ambient concentrations".
This statement is based primarily on the results of one study done in one city over two
seasons. In addition, there are no studies that have assessed the relationship between
community exposures and central site concentrations of SO2. Appropriately, the
summary chapter has been changed to place the significance of these results (or lack
thereof) in the context of epidemiological study design and now avoids repeating the
statement on page 2-52.
Other comments:
On page 3-52 the statement is made that "aerosol particles act as carriers and deliver SO2
to the lower respiratory tract". The evidence supporting this is based primarily on studies
done on rats exposed to relatively high levels (1 to 10 milligrams per cubic meter) of
freshly generated zinc, copper or carbon black particles in combination with SO2. These
studies report a synergistic effect of SO2 in combination with the particles and measure
sulfate and/or sulfuric acid coatings on the particles exposed to SO2 as a possible
explanation for the interaction. There is one study listed on page E-17 by Clarke et al
(2000- it is not in the reference list but I assume it is the one in Inhalation Toxicology
v!2, pp!69-186) that is summarized as seeing synergistic effects at more realistic levels
of sulfate particles (-20 micrograms per cubic meter). However, to produce these
particles in the lab required 1 milligram per cubic meter of carbon black particles and 1
ppm SO2 at 85% RH. In summary, the studies listed in support of the statement on page
3-52 were done at very high levels relative to contemporary atmospheric SO2 levels
(effects in the Clark et al study were seen after 5 to 6 days of exposure to 1 to 10 ppm
whereas the ISA summary of typical SO2 levels over this averaging period are typically
below 0.01 ppm; similarly, the PM levels were 1 -10 mg/m3 whereas typical PM levels
of fine carbon are 1-10 ug/m3). Although these studies do support the notion of an
increased response to sulfate-coated particles, they do not support the summary statement
on page 5-11 that "the synergism observed with combined exposure to SO2 and PM in
26
-------
the animal toxicological studies provides supportive evidence for the SO2-related
respiratory effects observed under ambient conditions in the epidemiological studies". In
fairness, this is presented in the document as an interpretation rather than a conclusion.
Given the unrealistically high SO2 and PM levels used in the animal studies (factor of
100 to 1000 relative to realistic ambient levels), its hard to claim that these studies
provide supportive evidence for the relevant epidemiological studies.
27
-------
Dr. Ted Russell
Comments on SOx ISA -- 2nd Draft
The 2nd Draft ISA is an improvement over the first, though could still benefit from some
pointed analyses, particularly in light of the direction of the associated RWA. In
particular, the REA deals predominately with assessing the exposure and risks associated
with 5-minute average SO2 concentrations, and how such concentrations are associated
with 1-hr average levels. Thus, this section of the ISA could be improved. In particular,
added analysis of the structures of 5-minute and 1-hr average concentrations, and how the
two correlate. In particular, the concentrations that lead to the 5-minute maximum and 1-
hr levels are from the same distributions, and thus are strongly linked. Having the
correlation structure would be useful to the staff developing the REA. Indeed, such
information could help streamline their chapter on air quality data. At present, the ISA
has little analysis of the five-minute results. It is interesting to note that the current ISA
says that it is "very difficult to use (these data) precisely", yet the REA does significant
analysis of the data (and I am not sure what they mean by "precisely").
I would think that SO2 concentrations being observed at a monitor will tend to follow a
log-normal distribution, whether one considers five-minute or one-hour average values,
and that the two distributions would have a similar mean, but that the GSD would differ.
(The geometric means of the two would not necessarily be the same, given that the one-
hour average is an arithmetic average). Thus, this section should consider presenting
concentration statistics in terms of the geometric averages and standard deviations (GSD)
(they will need to address how the below detection limit values are treated). From this,
one could assess how the GSD's of the five-minute and one-hour levels correlate, as well
as how the five-minute and one-hour maximums correlate. One might think that the five-
minute maximum can be well approximated as a function of the one-hour maximum and
associated GSD.
Page 1-1, line 12: monomeric, gaseous, sulfur species.
Page 1-9, line 18: The science is not uncertain... the results contain uncertainty.
Page 2-3, line 13: The sulfuric acid will move to a condensed, not necessarily aqueous,
phase. Further, it is not removed.
Page 2-3, line 18: What about sulfate?
Page 2-4, lines 20 on. This section is a abit confusing. They state that most of the
oxidation is via aqueous phase chemistry, then use the gas-phase lifetime to calculate an
overall lifetime. Again, it would be best if they used results from a US-based set of
simulations to address the formation, fate and lifetime.
Section 2-3: Again, provide a straightforward answer to the question as to whether
sampler issues impact the issues under consideration here, and I suspect the answer is no.
Table 2-4: Define units, make the caption more complete.
Page 2-25, line 9: I still think you mean three summer months, not seasons.
Figure 2-11: This figure is still of limited utility. It is impossible to see the frequency of
observations at the low end. Provide a plot that actually provides such information. All
this figure does, really, is indicate an average (which is too low to discern) and the
28
-------
maximums. Use a log scale. Better yet, give a table of arithmetic mean, geometric mean,
maximum, minimum and geometric standard deviation. If you want, you can add a
traditional SD, though I would state this is less helpful.
I was pleased to see Sections 3.1.5.2.1-3.1.5.2.5 dealing with SO2-particle interactions,
though none of these dealt with the dynamics of 862 uptake, and evaded the question of
how likely such mechanisms may be in the atmosphere. Uptake of SO2 on atmospheric
particles is likely very slow, as is seen by the ability for SO2 and particles to coexist in
the atmosphere for long amounts of time. Further, the studies' use rather artificial
surfaces that would be considered much more reactive and potentially absorbing than
atmospheric aerosols. A further limitation to how the results in these sections may be
applied in this assessment is that the products of the interaction between SO2 and the
particles would be monitored as part of the PM matrix, and be regulated as PM. While
there is a significant amount of sulfate in PM, which can be explained by gas and
aqueous phase oxidation of SO2, there is rather little sulfite, suggesting that particles
would provide little avenue as a mechanism to transport significant quantities of SO2
deeper in to the lung. SO2 has a low Henry's law constant, so little will be stored as
sulfite in the aqueous phase in a naturally occurring aerosol. This section should be more
critical in its assessment of the importance of such processes. At present, there is no
indication from observation of atmospheric aerosols that the studies discussed are
relevant, and, conversely, there is ample evidence to suggest what is found is limited to
laboratory generated aerosols that are then exposed to extremely high levels of SO2.
Summary and Conclusions:
Page 5-1, line 25: I would add "... facilities, though emissions from tall stacks may have
a lesser impact on high concentrations at ground level due to diffusion."
5-11, line 2: Here, it says that effect estimates were found to be robust when using 2-
pollutant models, but in Chapter 3, Fig 3-8, suggest that the significance of the SO2
effect decreases when including PM10, sometimes leading to its being insignificantly
different from zero? I probably also take exception to the conclusions in Chapter 3
dealing with their interpretation of the multi-pollutant assessment.
In response to the Charge Questions:
1. The framework for causal determination and judging the overall weight of evidence,
is presented in Chapter 1. Is this the appropriate approach? Is it appropriately
applied in the case ofSOx? How could the framework or its application be refined?
The current version is improved. I will leave the adequacy issue to those with a more
appropriate expertise.
2. The discussion of the atmospheric chemistry ofSOxhas been expanded to provide a
better characterization of the spatial heterogeneity of urban SO 2 concentrations and
correlations of SO 2 with other pollutants. We also included new sections describing
the regulatory network and siting criteria with maps ofSO2 and other monitors. A
29
-------
brief section describing the available 5-minute SO2 data was also included. In
addition, the relationships between outdoor, indoor, and personal exposure to SO2
were clarified with additional details on sources of exposure error. Have these
revisions to Chapter 2 improved its assessment of the currently available scientific
knowledge on atmospheric sciences and exposure and its relevance to the evaluation
of human health effects presented in later chapters?
While the new chapter is, indeed, improved, as discussed above there are still some
further needs. In particular, the section on 5-minute averages is inadequate given the
reliance of the REA on 5 minute data. This chapter should have CDFs for 1-hour and
associated 5-minute SO2 data, as well as a characterization of the correlations. The
chapter would still benefit from a table of sources and their amounts. Given the
discussion in Sections 3.1.5.2.1-3.1.5.2.5, this chapter should provide a scientific
foundation for the relevance of the proposed mechanisms and how they are supported by
observational evidence in the atmosphere.
3. In the revision, we reduced redundancy, added summary sections and reorganized
Chapter 3. In addition, discussions on potential confounding by and interactions with
copollutants have been added. The 2nd draft ISA also includes additional analyses of
individual-level data from human clinical studies (Sections 3.1.3. and 4.1.1) that
builds upon the analysis included in the 1994 Supplement to the Second Addendum.
The toxicology sections were reorganized to focus on studies using more relevant
concentrations of SO 2 and sections were added to better discuss mode of action and
potential par ticle-SO 2 interactions. We are requesting CASAC review specifically on
these analyses as well as on the integration of the overall evidence from the human
clinical, animal toxicological, and epidemiological studies.
As discussed above, Sections 3.1.5.2.1-3.1.5.2.5 should be assessed for their likely
importance in the atmosphere and to resulting exposures. Such sections should not be
added without appropriate foundation and discussion of how likely they are to be of
actual importance.
4. The section on concentration-response relationships in Chapter 4 was reorganized
and revised to include analysis of individual-level data from the human clinical studies
and some additional discussion of the difficulties of discerning a threshold in
population-level data. In addition, revisions were made to better characterize groups
likely to be susceptible or vulnerable to SOxand the potential size of the population at
risk for SOx-related health effects. Finally, revisions were made to reduce
redundancy with material presented in Chapter 3. Have the revisions made to
Chapter 4 improved the characterization of the potential public health impact ofSOx
exposure?
5. Revisions were made to better integrate findings from atmospheric sciences, ambient
air data analyses, exposure assessment, dosimetry, and health evidence in Chapter 5.
To what extent do these findings support conclusions regarding causality of SOxrelated
30
-------
health effects at relevant exposures?
The greatly streamlined Chapter 5 is an improvement. I will take issue with the use of
the studies cited suggesting that the particles act as ways to deliver SO2 significantly
more efficiently to the deeper lung, and that the sulfur associated with the particles is
predominantly sulfate formed from chemical reactions in the atmosphere, and very little
is from SO2 interacting directly with dry particles. The molecular kinetics and
atmospheric observations both argue against this.
-------
Dr. Kent Pinkerton
Comments on Sox ISA - 2nd Draft
General Comments: The second external review draft of the integrated science
assessment document for Sulfur Oxides/Health Criteria is much improved over the 1st
draft. The chapters are well organized and logical. The regulatory history for sulfur
oxides is interesting to observe with no changes since 1969, although comments were
solicited for a 1 hour standard of 0.4 ppm as well as a 5 minute standard of 0.60 ppm 862
in 1988. The current standard for sulfur oxides remains at 0.14 ppm averaged over a 24
hour period, not to be exceeded more than once per year and 0.030 ppm annual arithmetic
mean with SO2 as the indicator.
The authors of the document have focused each chapter to address the most relevant
issues and have presented a rather thorough and updated review of the literature since the
most salient features of past reports. New publications presented in the 2nd draft ISA are
highly appropriate, well-presented with a balanced interpretation. The presentation of
positive and negative studies is well balanced with possible interpretations provided
based on a solid scientific basis.
Question 4: The section on concentration-response relationships in Chapter 4 was
reorganized and revised to include analysis of individual-level data from the human
clinical studies and some additional discussion of the difficulties of discerning a threshold
in population-level data. In addition, revisions were made to better characterize groups
likely to be susceptible or vulnerable to Sox and the potential size of the population at
risk for Sox-related health effects. Finally, revisions were made to reduce redundancy
with material presented in Chapter 3. Have the revisions made to Chapter 4 improved the
characterization of the potential public health impact of Sox exposure?
Response:
The concentration-response relationships are clearly presented in Chapter 4. Human
clinical studies provide compelling data to demonstrate SCVinduced decrements in lung
function (FEVi) and specific airway resistance (sRaw) in asthmatics following limited
(10 minute) exposure to SC>2. An increasing proportion of asthmatic responders to
increasing levels of 862 further confirm a strong concentration-response.
Epidemiological studies, although less definitive than human clinical studies, also
provide strong evidence of increasing asthma hospitalizations with increasing 1-hour
maximum SC>2 levels ranging from 0.20 to 0.60 ppm. An important point to emphasize in
both the human clinical and epidemiological studies is the observation that significant
effects were noted at SO2 levels beginning slightly above 0.20 ppm.
Based on the human clinical and epidemiological study findings reported in Chapter 4, it
appears the approach to be implemented in the REA by the EPA of these findings is to
use 0.4 to 0.6 ppm as the beginning level for consideration to derive an appropriate level
32
-------
of safety to public health. Although data (especially the human clinical study) shows an
increased response as SC>2 levels are increased in a classical a concentration-response
profile, a strong rationale needs to be made for not beginning at a maximal SC>2
concentration of 0.20 ppm.
Once again, similarities between human clinical and epidemiological studies should be
emphasized, although I would agree that the human clinical studies provide the most
compelling evidence in considerations for setting the standard for SO2.
The difficulty in establishing a threshold in population-level data is well presented in
Chapter 4. Human clinical studies are likely to be the best source for setting risk levels,
but epidemiological studies continue to provide strong supportive evidence as well. It is
questionable whether one should consider non-linear concentration-response
relationships to SC>2 levels, based on epidemiological studies, especially those studies
illustrated in figures 4-4 and 4-5. These studies demonstrate effects beginning in the 0.2
to 0.3 ppm level. I would suggest (as somewhat already stated in the text of Chapter 4) it
is important to not over-interpret these epidemiological findings as being non-linear. To
do so, would also require an explanation in the Schwartz et al study (1994) (Figure 4-5)
for a negative odds ratio at a low level (0.10 ppm) of SC>2 and a non-linear increase in the
odds ratio as levels of SO2 increase to 0.4 ppm.
Chapter 3 contains a comprehensive overview the scientific literature along with a
complete and thorough review of new studies published since the previous review. The
presentation and stated interpretation and conclusions of this data in general are logical
and balanced. The summaries found at the end of each section in Chapter 3 are helpful
reminders and a reasonable and rational reflection of the section conclusions.
Chapter 4 is written to address critical issues of public health impact as discussed above.
The chapter contents are presented in such a fashion to utilize critical data presented in
Chapter 3 without being redundant. The definition and significance of susceptible and
vulnerable populations is discussed. These are important factors that require a clear
explanation in order to identify the appropriate populations at risk. Should these two
populations identified as "susceptible" and "vulnerable" have overlap, is this also a
critical component in estimating the appropriate size of the population at risk.
Excellent examples of epidemiological studies are presented to further confirm sensitivity
of children with asthma to 862. Similar sensitivity in adults has not been found. Genetic
factors related to health outcomes were also considered with regard to air pollutants as an
important factor to consider in future studies. A study by Winterton and colleagues
(2001) has found a significant association between 862 and the homozygous wild-type
allele for TNF-alpha.
Within chapter 4 is also presented a number of studies to examine age-related
susceptibility with a good balance of studies and outcomes. The tables and figures
throughout the chapter are highly complimentary to the text. To estimate the size of the
33
-------
population at risk is an important determinant, but will need to incorporate the
appropriate parameters.
In summary, this is an excellent chapter. Since the standard has not changed since its
original inception in 1969-71, if the potential decision by EPA in the REA is to suggest
changes in the standard to promulgate a 5-10 minute peak or 1-hour maximal level of
SC>2, this potential to recommend a very different standard from the current standard
needs to be clearly justified in this chapter or in the REA.
Specific comments:
Page 4-2, line 26: please state the meaning of sRaw in the text
34
-------
Dr. Edward M. Postlethwait
Comments on SOx ISA - 2nd Draft
The second draft of the SOX ISA is clearly improved. Suggested points for consideration
include:
1. Much of the document focuses on the potential impacts of short term, relatively
higher concentration exposures. In this regard, it is important to describe and/or
illustrate the characteristics of short term spikes. This may be important for
exposure/risk predictions since it is assumed that 5-10 minute exposures will be
modeled as steady state exposure conditions but in the real world such spikes may,
for example, be very sharp rather than the square wave exposure patterns that are
generally employed in most human clinical investigations.
2. Do the tables which include compilations of relative risks computed from population
studies include points which are not statistically significant? If so, then there may be
an unintentional bias in representing potential health impacts. If none significant
results are to be presented, they should be identified as such to appropriately inform
the reader.
3. While the human clinical studies of asthmatics show apparent thresholds at SC>2 levels
between 0.4 and 0.6 ppm, other data may suggest that some individuals may respond
at lower concentrations. Thus it is recommended that predictive estimates utilize
concentrations below the proposed benchmark 0.4 ppm level.
4. The above highlights an appreciable challenge for conducting risk estimates in that
some population studies suggest effects occur at levels well below those observed
during clinical studies. The ISA would clearly be strengthened by an improved
integration/synthesis regarding the potential genesis of effects threshold differences,
especially since the chapter on dosimetry clearly describes that > 90% of inhaled 862
is scrubbed in the upper airways so that during environmentally relevant exposures
only very small amounts of SC>2 would penetrate to anatomic regions involved in
bronchoconstrictive events. Thus at least an attempt to clarify the apparent
discrepancies is warranted.
5. The concluding language in the various health effects sections relating exposures to
the various biological endpoints of interest would be strengthened by more internal
consistency. For example, it is stated that there is sufficient evidence "to infer a
causal relationship between respiratory morbidity and short-term exposure to SCV',
which is unequivocal based on clinical studies. However, in other sections statements
regarding causality have the caveat "at current ambient levels." Thus, while
asthmatics, and likely most individuals, will respond at some threshold exposure
level, whether this occurs at ambient levels is not consistently addressed. Because of
the dearth of more current data and the confounding associated with multipollutant
35
-------
exposures in population studies, the ISA would benefit by more thorough discussion
of the potential for SC>2 to induce biological effects versus those effects that are/can
be detected under ambient conditions.
36
-------
Dr. Jonathan M. Samet
Comments on SOx 2nd Draft ISA
General Comments:
Overview:
This second draft Integrated Science Assessment (ISA) has been redrafted following the
initial review by CAS AC. Along with the draft ISA for Nitrogen Oxides, it represents
the second use of the new evidence review approach by the Environmental Protection
Agency. As such, it remains critical that a useful model be established for future reviews.
In this regard, I still find the second draft ISA to be deficient. While changes have been
made, the approach to evidence identification, review, and synthesis remains
inadequately specified. In the revised Chapter 1, while the Agency has begun to more
formally set out criteria for evidence evaluation and standard approaches for classifying
the strength of evidence for causation, the overall approach still needs development and
its application in the subsequent chapters of the ISA is not sufficiently thoughtful. My
responses to the charge questions follow:
Question 1:
This question asks whether the framework set out in Chapter 1 is appropriate and
adequately applied, and whether the Agency's approach for causal inference needs to be
strengthened. The Agency, in developing its approach, has relied on a number of recent
reports that offer models, including the approach taken in the Surgeon General's Reports
and the conceptual framework set out in the recent Institute of Medicine document,
Improving the Presumptive Disability Decision-Making Process for Veterans. These
documents along with the others that are cited do provide useful approaches. The
Agency's adoption of a five-level set of descriptors of strength of evidence is appropriate
and should prove useful for decision-making based on the degree of certainty provided by
scientific evidence considered. Through its application, the Agency will be able to refine
the review and synthesis process and, in particular, set precedents for responding to
evidence reaching different levels of certainty for causation. For example, would
evidence reaching the "suggestive" level be a sufficient foundation for altering a
NAAQS?
There are still limitations of Chapter 1, however. The discussion of association versus
causation is not well formulated and the authors are not clear in discussing key concepts
that affect interpretation of epidemiological data. For example, confounding and effect
modification are addressed rather superficially, as is measurement error. I urge the staff
to continue to refine this material, because it will become a model for future ISAs. The
authors comment on the complexity of determinants of health and of response to an
environmental agent, such as sulfur dioxide. However, their approach to doing so is
limited and Figure 1-1 is presented with little explanation.
37
-------
The discussion of uncertainty is brief and needs expansion. A more systematic approach
to describing uncertainty would be useful, both around determination of causation and
with regard to the quantification of effect. This important issue is handled in only one
page.
I was concerned by the application of the approach for causal inference in subsequent
chapters. The discussion is limited around evidence synthesis, particularly in Chapter 3.
There needs to be a better template for assuring standardization of discussion of the
strength of evidence.
Charge Question 4:
The revised Chapter 4 adequately sets out the quantitative information from the human
clinical studies. The plots are clear in showing the observed concentration-response
relationships. The epidemiological data are also described, although the findings cannot
be so readily summarized.
The chapter offers definitions of "susceptible" and "vulnerable" that need some
strengthening. Is a susceptible population as defined or one that has a steeper exposure-
response relationship than a non-susceptible population? The definition given here
emphasizes response at a lower concentration than in the general population. The
distinction between susceptibility and vulnerability lies in a greater biological response
for susceptible individuals and a greater risk for exposure (higher exposure?) in a
vulnerable population. Later in the chapter this distinction becomes blurred.
Specific Comments:
Page#
1-5
1-5
1-5
1-5
1-5
1-6
1-6
1-6
1-7
1-7
Line#
1
6
8
9
15
5
8
8
3
6
Comment
Need introduction here that a hierarchy of evidence may be identified
and that for some pollutants human clinical studies may provide the
most compelling evidence, e.g., CO, but not for all, e.g., PM.
Delete "can"
A panel study is a cohort study
".. .occasionally in epidemiology:" and offer the opportunity to
investigate a change in exposure.
"species" and the model used
"The results are" delete and substitute "observed risk represents"
The alternatives are chance and bias
"subjects in a population" usually volunteers
Randomization needs better description: of exposure (pollutant vs
sham or of people (pollutant or sham).
Delete randomness, substitute "the role of chance"
38
-------
Page#
1-7
1-7
1-7
1-7
1-7
1-8
1-8
1-8
1-12
1-13
1-14
3-1
3-1
3-2
O O
3-2
o o
J-J
3-3
3-3
o o
J-J
3-8
5-2
5-2
5-2
5-3
5-3
5-3
5-4
5-6
5-9
5-11
5-12
Line #
16
17
19
24
25
1
4
10
10
1
19
7
21
9
23
2
11
14
22
25
20
26
27
1
8
12
24
6
12
17
4
Comment
"Nutritional deficits" Such as?
"from epidemiological studies" Not inferred from epi studies alone.
Delete statistical
Delete "with", substitute of
Effect modification can also be examined with models
"possible mechanisms" and the distributions of these factors in
population under study.
the
"In multivariate analyses. . ." true? On what basis? References?
But not necessary
Delete "the absence of, Substitute not meeting
"recent publications" All relevant?
"The complex molecular and. . ." What does this mean?
Delete "nature", substitute role
May provide evidence
And to precisely characterize exposure/response
Delete "test", not a test-an application
"thus, attempts results of models that attempt to distinguish
Delete errors, substitute problems
Why?
"reliability" or validity?
How were studies selected?
"led to the conclusion that rt" Substitute the evidence.
T4ie7 substitute An immediate response
Depending on concentration
"human clinical studies (finding reporting of reporting respiratory
symptoms)"
"5-20%" Of asthmatics
"internal coherence" What is this?
How do they provide plausibility?
"much uncertainty remains" What does this mean? What kind of
uncertainty?
"Individuals in, substitute having a chronic"
"data support the finding. . ." Not the proposed language
39
-------
Dr. Richard Schlesinger
Comments on SOx ISA - 2nd Draft
Overall, the 2nd draft improves on the 1st in integrating the various disciplines. However,
there is still room for improvement in some of the sections as noted below. In many
cases, the various disciplines are not really integrated but rather separate discussions of
each are just placed in the same section. A true integration would interweave the epi with
mechanistic support from the controlled studies.
The inclusion of a section on mixtures, i.e., 3.1.5, is an improvement, but it is not clear
why that section was placed where it was. I think that a section on mixtures and the
potential for interaction related to all health outcomes from SO2 for all exposure
scenarios should be developed and placed at the end of the chapter prior to the
Conclusions.
Table 5-3 is excellent and provides a very nice summary overview of the conclusions of
the ISA and comparison to the previous Criteria Document.
Specific Issues
1. Section 3.1.3 is titled, Respiratory Effects Associated with Peak Exposure. In
reality, this is a discussion of studies that involve exposures for 1 hr or less to
single levels of SO2 without any baseline in animal and human clinical
evaluations. Thus, the use of the term peak may be misleading since this may
imply some higher level relative to some lower baseline. This section should also
be renumbered as follows:
3.1.3.1 Clinical Studies
3.1.3.2 Animal Studies
2. Section 3.1.4 discusses Epi studies and this should be reflected in the section
heading. Furthermore, lines 7-21 on page 3-17 should be moved to section 3.1.2.
3. Section 3.1.4.3 is an attempt to integrate animal, human clinical and epi studies.
4. Page 3-48, lines 15-17. The inconsistency may be due to the fact that the specific
nature of the interaction may depend upon the specific co-pollutant.
5. Section 3.1.6 seems to be out of place here. Perhaps the material could be
integrated into other sections.
40
-------
6. Section 3.2. title should perhaps read "systemic" rather than "other."
7. The conclusions in section 3.2.2.are not totally consistent with the statements
made on page 3-59, lines 24-27. In one case, the evidence is noted as inconsistent
while in the other it is noted to be inadequate.
8. On page 4-1, add "and vulnerable" following susceptible on line 10.
9. The comment on page 4-12, lines 1-2, that epi studies suggest that people with
preexisting respiratory diseases are more susceptible to effects from SO2
contradicts the comment on page 4-11, lines 12-13.
10. The sentence on lines 17-19 on page 4-20 is not clear.
41
-------
Dr. Christian Seigneur
Comments on SOx ISA - 2nd Draft
Charge Question 2. Have these revisions to Chapter 2 improved its assessment of the
currently available scientific knowledge on atmospheric sciences and exposure and its
relevance to the evaluation of human health effects presented in later chapters?
The second version of the ISA for the health criteria of sulfur oxides shows significant
improvements compared to the first draft version. Comments made last year on the
discussion of the physico-chemical processes that govern the evolution of sulfur oxides in
the atmosphere have been addressed in this revised version. Therefore, the following
comments on the second version tend to address points of detail or clarification rather
than fundamental issues.
Chapter 2. Source to tissue dose
Section 2.2, p. 2-3: "Because the saturation vapor pressure of H2SO4 is extremely low, it
will be removed rapidly by transfer to the aqueous phase of aerosol particles and cloud
drops". This is true only if aqueous particles or cloud drops (or fog droplets) are present.
For example, under dry conditions, gaseous sulfuric acid may be transferred to particles
to form dry particulate ammonium sulfate.
p. 2-20: The SO2 and sulfate concentration maps shown in Figures 2-8 and 2-9 originate
from the CASTNET data using some spatial interpolation methodology. Note that EPA's
Clean Air Market Division (CAMD) has developed some data fusion approach that
combines the CASTNET data with a CMAQ simulation to produced concentration maps
that are improved compared to the simple CASTNET interpolation maps (CAMD
contact: Melissa Rury).
Annex B
CTMs are sometimes referred to as chemistry-transport models (e.g., p. B-15, line 1),
sometimes as chemical-transport models (e.g., Title of Section B-5). It may be preferable
to use a single term throughout the document.
p. B-7, lines 12-13: The statement that transportation related sources only have a minor
contribution to SO2 ambient levels suggests a strong east-coast bias. In California, where
point sources have low SO2 emissions compared to the eastern U.S., ships can be a major
contributing source to SO2 concentrations. As a matter of fact, it would be appropriate to
highlight the recent State of California regulation on sulfur content authorized for diesel
fuel of ships that are within the proximity of the Californian coast.
p. B-17, line 6: The Weather Research & Forecast (WRF) model is now the preferred
meteorological model; add it in the parentheses before RAMS.
42
-------
p. B-22, line 14: The discussion points out that a plume-in-grid treatment affects the
simulation results for ozone concentrations. This effect is due to the improved treatment
of the NOX emissions when using a subgrid-scale representation of major point sources.
In reality, the plume-in-grid treatment redistributes the ozone concentrations with little
effect on the overall ozone budget (e.g., Karamchandani et al., Development and
application of a state-of-the-science plume-in-grid model, J. Geophys. Res., 107, 4403-
4415, 2002). Since the document focuses on SOX, it should be noted that a plume-in-grid
treatment of SOX (and NOX) emissions has a significant effect on secondary sulfate (and
nitrate) formation that may lead to up to 15% less sulfate (or nitrate) formation from
those major point sources treated with a subgrid scale representation within the 3-D grid-
based model (Karamchandani et al., Plume-in-grid modeling for particulate matter,
Atmos. Environ., 40, 7280-7297, 2006).
p. B-25: It is true that compensation of errors may lead to good model performance for
the wrong reasons and that the performance evaluation of individual model components
is desirable. Therefore, it would be appropriate at this point to mention the effort made to
evaluate the aqueous chemistry component of SOX models either in the laboratory
(Waltrop et al., J. Atmos. Chem., 12, 1-17, 1991) or in the atmosphere (Daum et al.,
Atmos. Environ., 18, 2671-2684, 1984). Also, the evaluation of reactive plume models
against aircraft data should be addressed.
p. B-25: This section only discusses regional and global CTM for SOX and does not
present any information on plume dispersion models that are used to estimate local (e.g.,
within 50 km) impacts of major SOX emission sources. Since a plume dispersion model
(AERMOD) will be used to conduct the Risk and Exposure Assessment, it is desirable to
have a section in the Annex that presents this type of models, their limitations and
available performance evaluation results.
43
-------
Dr. Frank E. Speizer
Comments on SOx ISA - 2nd Draft
Chapter 1 :
Page 1.6, Figure 1-1. I do not find this particular figure that useful. Perhaps Gee &
Payne-Sturges, 2004 have sufficient text that accompanies the figure to make it useful,
but simply putting it out without adequate descriptors do not help. I could imagine many
of the arrows going in both directions if not in opposite directions. My suggestion would
be either to expand the text of abandon the figure.
Page 1-6, section 1.3.3 the first sentence I this section needs to be modified. There is
another alternative way of stating what is being said. After all there are no clinical
studies of tobacco smoke and lung cancer. The studies are all epidemiological
"association" studies. But the consistency and reproducibility and the magnitude of the
associations are so overwhelming that there are no alternative tenable explanations (there
were early on but these have all but gone away).
Page 1-14, last box in table from previous page. I think it would be worth putting a limit
on "any level of exposure" Perhaps something like reasonable level or at least in human
non-toxicological levels.
Page 1-15 end of section 1.3.8. Two issues that are not discussed here, (will need to be
addressed somewhere) but I thought should be part of this overall assessment of causality
are "uncertainty" and "margin of safety". There are a various interpretations of each and
laying out some definitions, even if they might be different definitions used by different
people, would be useful. What constitutes a level of uncertainty that is acceptable to all,
90% of regulators, 50% of regulators etc? Similarly, are there percents of uncertainly
that are judged by advocacy groups that are different? With regard to margin of safety
ditto. Is it 20% margin, 50%, 100%, 500%. A discussion up front in the document might
be a useful exercise and might put CASAC, Staff, and administrative policy makers all
on same page.
Focus Question 1.
I think the cataloging of the descriptions of how causation might be determined is
reasonable. The issue really isn't how we generally understand how epidemiology,
clinical studies and animal toxicology data are used. More important issues are the ones
raised in my last point above. I know there have been focus group discussions on the
topic and there may even be some published research on the subject. The more I think
about it the more important is seems to me that it be part of this chapter. Adding this
material would substantially change the discussion and perhaps aid in what will be a
difficult discussion is subsequent chapters when causation gets interpreted in making
recommendations for standards.
Additional Comments related to other Chapters.
44
-------
Figure 3-1, Page 3-12: This is an important figure that should be more thoroughly
discussed. The figure indicates, from one of the larger early studies that there are marked
discrepancies in the responsiveness of individual subjects. About 25% of the subjects
were essentially unresponsive in terms of airway sensitivity below 2 ppm and another
25% were responsive below 0.5 ppm. Thus, in assessing group data it is obviously the
case that within any group the responders and the non-responders need to be considered
separately if at all possible. In this study of asthmatics, in which the presumption would
be that the response pattern would be random across the group, this clearly was not the
case. I am sure that I and others argued in the past that this raises concern that in
attempting to justify a short term standard the fact that which 25% of the asthmatic
population would be affected cannot be determined prior to exposure argues for
consideration in the setting of a standard with an adequate margin of safety. I would
contend that 25% of the population of asthmatics is a larger fraction of the most sensitive
population than one should be comfortable to ignore in setting a short term standard.
This argument did not win the day in 1994. The data since that time has further
established the issue that there are what might be called "super sensitive" groups of
people and the whole issue of both what proportion of the population should be protected
and what the margin of safety should be will need to be revisited. In this case margin of
safety is not a surrogate for uncertainty of effect but a measure of the size of the
population to be protected.
In reaching a conclusion on page 3-46 the general impression is that the evidenced is
suggestive. However, not taken into account in reaching these conclusions is the
evidence that not all subjects within any given population are equally sensitive to the
impact of SO2. In my mind this phenomena tips the balance that regulation of short term
exposure is necessary.
This point is made even more effectively in Section 4.1.1 on page 4.2-3 along with figure
4-1. Given that sRaw is measured as doubling and FEV1 change is measured as 15%
decrease the range of changes from 0.2ppm to 0.6 is quite substantial and is consistent
with a significant adverse effect that will need to be modeled in the risk assessment.
A better job needs to be done in indicating that subgroups of subjects appear to be more
sensitive and may be driving the overall assessments in any specific study. These
subgroups appear to exceed 5% to 13% for sRaw and FEV respectively, at the lowest
levels of exposure tested (0.2 ppm). The lack of a threshold at these (relatively) high
levels of exposure as compared to "background of less than 15ppb makes straightforward
extrapolation uncertain. I am sure there are people who are sensitive at these lower levels
but finding an appropriate level and population at risk will be difficult.
With regard to long term exposure and mortality the larger (and better) conducted US
studies are consistent but when assessed in the context of all studies that might be
interpreted as negative leave the impression that the results are inconsistent and therefore
no prudent declaration of causal association can be made. This seems slightly off base as
there is adequate reasoning given as to why some studies might be positive and some
negative and one simply cannot sum the total of studies.
45
-------
Summary and Conclusions Chapter
The conclusion about the "very few monitors" Page 5-2, line 16 suggests that one strong
recommendation in this chapter is that an expanded network of monitors are needed.
This may or may not be true depending upon the distribution seen at existing monitors in
that with a 1 hour mean of 13ppb and a 99th percentile value of 120-700ppb little
information is supplied other than the conclusion given that the max is not contributing to
the mean. Some full distributions might help lead to the conclusion as to how much
additional monitoring is necessary outside potential known hotspots.
Having the key finding summarized and reproducing the summarizing tables from the
appendices is useful.
46
-------
Dr. Lianne Sheppard
Comments on SOx 2nd Draft ISA
Overall assessment: This document is much improved from the previous version.
Generally it flows better and more closely attains its goal of presenting an integrated
science assessment of SO2 and its effect on health.
Charge question 1: The Chapter 1 framework is reasonable.
Charge question 2: Chapter 2 is better.
• There is better characterization of the spatial distribution of SO2 monitors, but the
discussion is still lacking in the sense that there is no stratification of analyses by
micro vs. neighborhood scale monitors. This is an important feature to take into
account in the interpretation of epidemiological studies. What fraction of
monitors in each area falls in these categories? How many areas with SO2
monitors lack any neighborhood scale monitors?
• Section 2.5 is a good addition and a good effort was made in 2.5.4 to discuss
important features of exposure assessment for epidemiological studies. The
availability of neighborhood scale monitors should be brought into this
discussion, as well as the reactivity of SO2.
Charge question 3: Chapter 3 is much more readable and well organized.
Charge question 4: The revision to the concentration-response modeling discussion is
reasonable. Discussion of vulnerable and susceptible groups is reasonable, but could
possibly still benefit from a catalogue (e.g. in a table) of all potential vulnerabilities and
susceptibilities that warrant consideration. The public health impacts section appears to
be incomplete. While the gradation of responses and the prevalence of respiratory
disorders tables are useful, these are not linked to provide any inference about degree of
public health importance of SO2 exposure in the population.
Charge question 5: There is good integration in Chapter 5.
Specific comments: To be added.
47
-------
Dr. George Thurston
Comments on SOx 2nd Draft ISA
1. The framework for causal determination and judging the overall weight of evidence, is
presented in Chapter 1. Is this the appropriate approach? Is it appropriately applied in
the case of SOx? How could the framework or its application be refined?
RESPONSE: Yes, I feel the staff has done an excellent job of properly incorporating
and explaining both the weigh of evidence criteria and the causality inference framework.
2. The discussion of the atmospheric chemistry of SOx has been expanded to provide a
better characterization of the spatial heterogeneity of urban SO 2 concentrations and
correlations of SO 2 with other pollutants. We also included new sections describing
the regulatory network and siting criteria with maps of SO2 and other monitors. A
brief section describing the available 5-minute SO2 data was also included. In
addition, the relationships between outdoor, indoor, and personal exposure to SO2
were clarified with additional details on sources of exposure error. Have these
revisions to Chapter 2 improved its assessment of the currently available scientific
knowledge on atmospheric sciences and exposure and its relevance to the evaluation
of human health effects presented in later chapters?
RESPONSE: Yes, this is much improved, but I would like to see a better representation
of the trends in short-term SC>2 concentrations (i.e., 1-hr maximum). While there has
clearly been improvement in annual average SC>2 concentrations, as would be expected
given the advent of the acid rain control program, these efforts have focused on lowering
overall tons of SC>2 emissions, and not on lowering maximum impacts. So the question
remains, have we seen a reduction in peak SC>2 exposures near major sources? As such, I
would ask that plots of the distribution of 1-hour maximum values (for which many more
sites are available than for 5-minute peaks) over time to better elucidate how much
improvement we are making in lowering peak SC>2 exposures.
Another issue I have with Chapter 2 is the inadequate coverage of the PM-SO2
interaction, and the potential for the co-presence of PM to increase lung dosage and
health effects of SC>2 over and above that indicated by controlled animal and human
populations using pure SC>2 gas, without associated PM (which would be the case in the
real world). While the document includes research by Amdur on page 2-60, it does not
include her work very clearly showing the way that co-presence of PM along with SC>2
increases the impact of acute SC>2 exposure. While these studies are discussed in
Chapter 3, they are probable even more important to mention here as well, as they are
important to our understanding of SC>2 doses in the real world. Now, while SO2 has a
relatively low Henry's law constant, and it might therefore be expected that little will be
stored as sulfite in the aqueous phase in a naturally occurring aerosol, but the Amdur
studies conversely imply that this SO2-PM interaction is a strong one, so the document
48
-------
needs to more thoroughly investigate this mechanism, and better sort out its possible role
in enhancing the effects of ambient SC>2. This might very well be an important
consideration in trying to extrapolate from controlled studies to effects in populations, as,
if ignored, it may well mean that the human health effects of elevated SC>2 exposures will
be underestimated, and the SC>2 concentrations required for a response would be
underestimated, as well. This consideration also points to the greater relevance of using
epidemiology for risk assessment, rather than controlled exposures to subjects, since
epidemiology will have already incorporated such PM-SO2 interactions that happen in
the real world, but not in controlled exposure environments lacking PM.
3. In the revision, we reduced redundancy, added summary sections and reorganized
Chapter 3. In addition, discussions on potential confounding by and interactions with
copollutants have been added. The 2nd draft ISA also includes additional analyses of
individual-level data from human clinical studies (Sections 3.1.3. and 4.1.1) that
builds upon the analysis included in the 1994 Supplement to the Second Addendum.
The toxicology sections were reorganized to focus on studies using more relevant
concentrations of SO 2 and sections were added to better discuss mode of action and
potential par ticle-SO 2 interactions. We are requesting CASAC review specifically on
these analyses as well as on the integration of the overall evidence from the human
clinical, animal toxicological, and epidemiological studies.
RESPONSE: While this section makes advances, Section 3.1.5.2.5 is wholly inadequate.
It ignores the major implication of these interactions: that exposure studies that do not
include the potentiating effects of PM on SC>2 exposures will very likely have both
underestimated the size of SC>2 effects and the overestimated the SC>2 levels at which
such effects may be experienced in the co-presence of particles (i.e. in the real world).
4. The section on concentration-response relationships in Chapter 4 was reorganized
and revised to include analysis of individual-level data from the human clinical studies
and some additional discussion of the difficulties of discerning a threshold in
population-level data. In addition, revisions were made to better characterize groups
likely to be susceptible or vulnerable to SOxand the potential size of the population at
risk for SOx-related health effects. Finally, revisions were made to reduce
redundancy with material presented in Chapter 3. Have the revisions made to
Chapter 4 improved the characterization of the potential public health impact ofSOx
exposure?
RESPONSE: Again, in this chapter, there is little acknowledgement of the fact that many
of the clinical and animal studies were done using pure SC>2, without co-PM exposures,
as would occur in the real world. This is especially critical in section 4.3.2, where the
clinical studies are used to indicate benchmarks of effects. Such a choice of levels at
which various effects would occur must take into account the documented effects that
PM would have on the size and threshold of effects by SC>2. For example, based on the
work of Amdur and others, a conservative factor to adjust for the copresence of PM
might well be applied to estimate the levels at which such clinical symptoms might occur
in the real world.
49
-------
5. Revisions were made to better integrate findings from atmospheric sciences, ambient
air data analyses, exposure assessment, dosimetry, and health evidence in Chapter 5.
To what extent do these findings support conclusions regarding causality of SOxrelated
health effects at relevant exposures?
RESPONSE: The big issue that needs more discussion is why the epidemiology shows
health effects associations at much lower concentrations than do the clinical studies. The
paragraph at the top of page 5-11 does discuss one likely explanation of this apparent
paradox: the effects of the fact that "aerosol particles act as carriers and deliver sulfur-
containing compounds more effectively to the lower respiratory tract". However, more
discussion is needed here of the fact that this may well account for why epidemiological
associations are found at SC>2 concentrations well below where effects are documented
by clinical studies using pure SC>2 exposures.
50
-------
Dr. James Ultman
Comments on SOx ISA - 2nd Draft
General Comments
This second draft of the ISA provides a significant improvement over the first draft, both
with respect to content and chapter organization.
Apparently, the framing questions that originally appeared on page 1-2 in the first draft of
the ISA have been deleted from the second draft. The answers to several (but not all) of
these questions can be culled from the final "Conclusions" section on page 5-12 of the
revised document. Given the importance of these questions in informing the REA as well
as the final rulemaking, EPA staff should consider reinserting the questions into Chapter
1 and then explicitly answering them in Chapter 5.
Comments on Chapter 2 and Charge Question #1
General
A major improvement in this chapter is the new material on monitor locations,
particularly those 108 sites where 5-minute averages have been recorded.
I am, however, puzzled by the statement on lines 1-2 of page 2-32 that: "Although these
5-minute data meet AQS minimum quality assurance requirements.. .these data (are) very
difficult to use precisely." This appears to be the rationale for not including any actual
data from these sites in the document.
In fact, this very set of data, is used in the first draft REA to model the relationship
between peak and short-term ambient SO2 concentrations. I suggest that disconnect
between the ISA and REA be remedied by including some data, in either a tabular or
graphical form, regarding the peak and 1-hour short-term concentration data obtained
from the monitors listed in tables 2-6 and 2-7. Is it possible, for example, to examine and
compare the relationships between the two concentration averages over time at various
sites?
The text indicates that ambient monitors are allowed to be sited at any vertical location
between 2 and 15 meters above the ground (pg. 2-9 lines 3-4). Yet, there is vitually no
discussion of vertical SO2 gradients nor the errors such gradients introduce when using
ambient concentration as a surrogate for ground-level concentration. For example,
ground-level concentrations near point sources will be much lower than concentrations
recorded from monitors high off the ground. It would be beneficial for the ISA to
provide a better qualitative and quantitative appreciation of this.
51
-------
Specific Comments
Page; lines
2-3; 10 &12 Define terminology (OH for hydroxyl free radical and HC>2 for
hydroperoxyl radical) immediately after the equation in which they first appear.
2-9; 3-4 How large a difference between monitored concentrations is a variation of 2-15
m difference in monitor height likely to introduce? How does this difference depend on
distance to the point source?
2-14; 1 Make sure that SLAMS and NAMS acronyms have been previously defined.
2-23; 9 I assume that you mean "the Pearson correlation coefficients (r) for
concentration data from multiple monitors taken as pairs in these CMS A's..."
2-26; 9-11 104 sites +15 sites is greater than the total of 108 sites mentioned in the
previous sentence.
2-41; 5 Given the importance of exercise in promoting health effects in asthmatics, it is
important that this be included as a personal activity in this list.
2-41;17-21 Incomplete, run-on, sentence.
2-58; 14 For clarification, it would be helpful to add a phase indicating that the Henry's
Law constant is inversely proportional to solubility.
One-way flow versus cyclic flow—
2-59; 4-5 What the text refers to as "total respiratory tract absorption" is presumably the
retained dose of SO2 over a complete respiratory cycle.
2-59; 18-20 In these experiments, the upper airways were isolated from the lower
respiratory tract, and absorption was measured during one-way flow (as opposed to the
bidirection flow that occurs during a complete respiratory cycle). This was also the case
for other investigations that are cited later on pages 2-59 and 2-60.
2-63; 6-8 It is precisely because of desorption during exhalation that cyclic flow
experiments report smaller absorption efficiencies than one-way flow experiments. This
explanation should be worked into the discussion.
52
-------
Dr. Ronald E. Wyzga
Comments on SOx ISA - 2nd Draft
Charge question 3: In the revision, we reduced redundancy, added summary sections
and reorganized Chapter 3. In addition, discussions on potential confounding by and
interactions with copollutants have been added. The 2n draft ISA also included
additional analyses of individual-level data from human clinical studies (Sections 3.1.3
and 4.1.1) that builds upon the analysis included in the 1994 Supplement to the Second
Addendum. The toxicology sections were reorganized to focus on studies using more
relevant concentrations of SO2 and sections were added to better discuss mode of action
and potential particle-SO2 interactions. We are requesting CASAC review specifically
on these analyses as well as on the integration of the overall evidence from the human
clinical, animal toxicological and epidemiological studies.
Response to charge question: By and large I applaud the efforts undertaken in the
revision. It increases the presentation and discussion of the human clinical studies, which
I believe are extremely informative about the health effects of SO2. The focus is clearly
upon respiratory responses, especially among asthmatics. I previously had sent the
Agency staff several papers of human clinical studies, which have hopefully aided in this
effort. They provide comprehensive input to aid our understanding of responses of
asthmatics exposed to SO2 in the chamber. To some extent the results presented in the
ISA could have been extended to indicate the importance of other factors, such as the
influence of routine asthma medication and weather upon response.
I believe the document was largely successful in integrating results across scientific
disciplines although I note that several highlighted toxicological studies do utilize
exposures orders of magnitude higher than those which occur in the ambient
environment, (e.g., 0.8-6mg/m3 and l-2ppm SO2). Special note should be made of the
fact that these concentrations are extremely high and not representative of those that
occur in the real world. The copollutant issue in epidemiological studies is a difficult one
to address largely because few studies address this issue systematically, making it
difficult to resolve this issue definitively. I would urge the document to ask for more
systematic investigation of this issue within individual studies.
Overall comments: This document is substantially improved over the previous draft; it
considers the informative human clinical studies in more detail and correctly focuses
upon respiratory health, especially asthma issues. I believe that relatively small changes
in the document are needed. I give my specific comments below, which generally ask for
further clarification or greater articulation of the presented material.
Specific comments:
p. 2-23,11. 9-20: perhaps some discussion about the dependence of exposure to SO2 on
whether an individual (monitor) is in the plume from a point source is warranted.
53
-------
p. 3-11; 11. 7-18: Another reason for the similarity in response between mild and
moderate/severe asthmatics is that the latter were not able to exercise as intensely as the
mild asthmatics. It should also be noted that studies (e.g., Gong et al., 1996; Gong et al/.
2001; Linn et al. 1988) have shown that medicated asthmatics were not sensitive to SO2
exposure.
p. 3-15,11. 11: insert "exercising, non-medicated" before asthmatics.
P. 3-16,1. 3: Linn et al., 1983 reported a significantly increased response among
asthmatic subjects utilizing a mouthpiece for exposure as opposed to those with
unencumbered breathing. This reference can be cited here.
p. 3-35,1. 10: "statistically"
p. 3-43,11. 10-19: Systematic examinations of the potential influence of co-pollutants on
study results are rare. Studies often consider a limited set of co-pollutant with no
discussion about why other co-pollutants are not considered. It would be useful to
indicate the need for more systematic review of this issue within a study.
p. 3-45,11. 1-7: It should be noted that the panel studies found responses at ambient
levels only when asthmatics were exercising.
p. 3-46,1. 31: reference should be given.
p. 3-47,11. 14-20: It should be noted that exposure was by mouthpiece. See comments for
p. 3-16.
p. 3-49,11. 16 and following: It should be noted that the exposure levels in the studies
cited in this section were extraordinarily high. This comment applies to section 3.15.2.5.
p. 3.55,11. 21-22: It should be noted that controlled human studies did not find
responses in adult atopies. (Linn et al., 1987).
p. 4-2,1. 1: Measurement error can also serve to linearize dose-response estimates.
p. 4.4,11. 10-11. See comment for page 4-2.
p. 4-11,11. 4-13: It might also be noteworthy to indicate that asthmatics on routine
medication do not appear to respond to SO2 while exercising.
54
------- |