:(( Clean Air Scientific Advisory Committee (CASAC)
• A Federal Advisory Committee to the U.S. Environmental Protection Agency
March 18, 2022
EP A-C AS AC-22-001
The Honorable Michael S. Regan
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, N.W.
Washington, D.C. 20460
Subj ect: CASAC Review of the EPA's Supplement to the 2019 Integrated Science Assessment
for Particulate Matter (External Review Draft - October 2021)
Dear Administrator Regan:
The 2021 Clean Air Scientific Advisory Committee (CASAC) Particulate Matter (PM) Review Panel,
hereafter referred to as the Panel, met on October 14, 2021, November 17-19, 2021, December 1-2,
2021, February 25, 2022, and February 28, 2022, to peer review the EPA's Supplement to the 2019
Integrated Science Assessment for Particulate Matter (External Review Draft - October 2021), hereafter
referred to as the Draft ISA Supplement. The Chartered CASAC approved the Panel's report on
February 28, 2022. The CASAC's consensus responses to the agency's charge questions and the
individual review comments from the Panel are enclosed.
The CASAC commends the EPA for returning to its long-standing practice of constituting an ad hoc
panel of experts to complement the expertise of the Chartered CASAC. The CASAC recommends that
the practice of convening a panel of additional experts continue for all future NAAQS reviews because
the give-and-take deliberation and participation of multiple scientific experts, including multiple experts
from all key disciplines needed to conduct a high-quality scientific review, is fundamental to the
Chartered CASAC's ability to provide the highest quality scientific advice. With a fully constituted ad
hoc panel of experts, the CASAC has a depth and breadth of expertise that enables it to fulfill its
mandate to provide advice and recommendations to the EPA.
The CASAC is concerned that the compressed timeframe for this and other recent CASAC reviews has
made it difficult for the CASAC to provide the highest quality review possible. The compressed
timeframe has necessitated simultaneous review of large documents, which is not optimal. The
timeframe does not allow for development of second drafts of documents that incorporate CASAC
advice (should the CASAC recommend second drafts). The CASAC recommends that for future
reviews, the EPA follow a review plan that allows for adequate time for the CASAC to review the
documents and for sufficient time for the EPA to incorporate CASAC advice into second drafts (if
requested by the CASAC). The review plan should also allow sufficient time for CASAC advice on
earlier documents (e.g., Integrated Science Assessment) to also be incorporated into subsequent
documents, (e.g., Risk and Exposure Assessment and Policy Assessment). Documents that are
-------
sequential (e.g., the Integrated Science Assessment and Policy Assessment) should not be developed
simultaneously and presented to the CASAC for simultaneous review. In spite of the compressed
timeframe for this review, the CASAC is confident that it has appropriately completed its task. Once the
EPA addresses the CASAC's comments, the final Integrated Science Assessment (ISA) Supplement and
Policy Assessment (PA) deserve the Administrator's full consideration and are adequate for rulemaking.
Overall, the CASAC finds the Draft ISA Supplement to be a well-written, comprehensive evaluation of
the new scientific information published since the 2019 PM ISA. There are several recommendations for
strengthening and improving the document highlighted below and detailed in the consensus responses.
The scope of the Draft ISA Supplement is limited to health effect categories where the 2019 PM ISA
concluded a causal relationship (i.e., short- and long-term PM2.5 exposure and cardiovascular effects and
mortality). Although this limitation is appropriate for the targeted purpose of the Draft ISA Supplement,
the CASAC notes that this limitation precludes consideration of any new evidence that might change a
causality determination from a likely to be causal relationship to a causal relationship. This limitation
should be explicitly acknowledged. It should also be noted in the Draft ISA Supplement that this
limiting of scope applies only to this document and is not intended to establish a precedent for future
ISAs. A discussion of the rationale for limiting the scope of the document to U.S. and Canadian studies
should also be included. The Draft ISA supplement should also provide additional background and
rationale for the reconsideration of the December 2020 decision to retain the PM National Ambient Air
Quality Standards (NAAQS). Although continued refinements to the current weight-of-evidence (WOE)
causal determination framework are possible, the CASAC unanimously supports the use of the current
WOE causal determination framework, as described in the 2015 Preamble to the ISA, for this review
and strongly believes that this framework should not be replaced without a comprehensive evaluation of
alternatives.
The CASAC finds the summary of health effects to be well-written and thorough, with appropriate
identification, evaluation, and characterization of available scientific evidence, within the stated scope of
the Draft ISA Supplement. The text on "causal modeling methods" should be reworded to clarify these
methods' role in the WOE causality determinations. While recent emergence of studies employing such
methods is important to the Draft ISA Supplement, the CASAC recommends that the term "causal" not
be used to describe these methodologies and to use alternative descriptive language to avoid the
potential misconception that labeling some studies as "causal" carries an implication that more
traditional epidemiologic analysis methods cannot support a causal determination or should receive less
weight in any WOE causality determinations. The CASAC notes the importance of these methodologies
in the Draft ISA Supplement for their ability to reduce some of the uncertainties raised by the previous
CASAC and the previous Administrator in response to the 2019 ISA. It is important to clarify that these
methods are not intended to replace the causality determinations of previous ISAs, but rather have been
recently adopted in service of strengthening the body of evidence for causality determinations.
Regarding study descriptions and findings, more detail would be useful to give context to the results
(e.g., hazard or risk ratios and their confidence intervals, sample size, methods for exposure assessment).
Specifically, conclusions are needed for these studies regarding what was learned, and the ways and
degrees in which the studies strengthen or weaken the state of scientific evidence. There should be
consistent presentation of effect estimates (e.g., relative risk, hazard ratios, odds ratios) and their
confidence intervals (including the pollutant concentration to which they are scaled), descriptions of
exposure measurement approaches used and distributions of PM concentrations, conclusions made from
these findings, and better distinguishing of individual-level and community-level measures of
socioeconomic status (SES) data for the studies presented.
-------
The CAS AC notes that there is a progression going from the 2009 ISA to the 2019 ISA to this Draft ISA
Supplement indicating continued strengthening of the causal health endpoints relationship with PM2.5.
The literature, as it is expanding, continues to show strong associations with health effects, even though
concentrations of PM2.5 in the air have been decreasing over time.
The CASAC recommends using different language when discussing race/ethnicity in the Draft ISA
Supplement. The summaries and conclusions within this document use the terms "White" and "non-
White" as the broadest categories. The Draft ISA Supplement should refer to the group "non-White" as
"People of Color (POC)" or "Communities of Color (COC)," as appropriate. The CASAC recognizes
that the different uses of the terminology in previous EPA documents as well as in the published
literature means that there will be some nuance to how this recommendation is implemented.
The CASAC agrees that recent evidence confirms the causal relationship between PM and visibility
impairment. However, a few clarifications are needed. Section 4 acknowledges variations in humanly
perceivable acceptance of visibility levels. The use of "apparent contrast" instead of "total light
extinction" shows less variability to determine acceptable visual air quality. Values for the cut-off
between unacceptable versus acceptable levels for atmospheric light extinction and non-subjective
contrast measurements should be clarified. Science-based visibility standards warrant additional
research using objective scenarios to quantify visibility improvement.
With increases in the frequency and intensity of wildfires and reductions in sulfur dioxide (SO2) and
oxides of nitrogen (NOx) emissions, mass scattering coefficients for the major PM2.5 components (e.g.,
organic and elemental carbon, sulfate, and nitrate) need to be further examined with more recent data
(e.g., 2015 onward). The "revised" Interagency Monitoring of Protected Visual Environments
(IMPROVE) chemical extinction equation should be included, the Lowenthal and Kumar IMPROVE
equation should be included, and differences among the various IMPROVE equations that use split
component algorithms need to be addressed.
The CASAC appreciates the opportunity to provide advice on the Draft ISA Supplement and looks
forward to the agency's response.
Sincerely,
/s/
Dr. Elizabeth A. (Lianne) Sheppard, Chair
Clean Air Scientific Advisory Committee
Enclosures
-------
NOTICE
This report has been written as part of the activities of the EPA's Clean Air Scientific Advisory
Committee (CASAC), a federal advisory committee independently chartered to provide extramural
scientific information and advice to the Administrator and other officials of the EPA. The CASAC
provides balanced, expert assessment of scientific matters related to issues and problems facing the
agency. This report has not been reviewed for approval by the agency and, hence, the contents of this
report do not represent the views and policies of the EPA, nor of other agencies within the Executive
Branch of the federal government. In addition, any mention of trade names or commercial products does
not constitute a recommendation for use. The CASAC reports are posted on the EPA website at:
https://casac.epa.gov.
1
-------
U.S. Environmental Protection Agency
Clean Air Scientific Advisory Committee
CHAIR
Dr. Elizabeth A. (Lianne) Sheppard, Rohm and Haas Professor in Public Health Sciences, Department
of Environmental & Occupational Health Sciences and Department of Biostatistics, Hans Rosling
Center for Population Health, University of Washington, Seattle, WA
MEMBERS
Dr. Michelle Bell, Mary E. Pinchot Professor of Environmental Health, Yale University School of the
Environment, New Haven, CT
Dr. James Boylan, Assistant Branch Chief, Air Protection Branch, Environmental Protection Division,
Georgia Department of Natural Resources, Atlanta, GA
Dr. Judith C. Chow, Nazir and Mary Ansari Chair in Entrepreneurialism and Science and Research
Professor, Division of Atmospheric Sciences, Desert Research Institute, Reno, NV
Dr. Mark W. Frampton, Professor Emeritus of Medicine, Pulmonary and Critical Care, University of
Rochester Medical Center, Rochester, NY
Dr. Christina H. Fuller, Associate Professor, Department of Population Health Sciences, Georgia State
University School of Public Health, Atlanta, GA
Dr. Alexandra Ponette-Gonzalez, Associate Professor of Geography and the Environment, University
of North Texas, Denton, TX
SCIENCE ADVISORY BOARD STAFF
Mr. Aaron Yeow, Designated Federal Officer, U.S. Environmental Protection Agency, Science
Advisory Board, Washington, DC
11
-------
U.S. Environmental Protection Agency
Clean Air Scientific Advisory Committee
Particulate Matter Review Panel (2021)
CHAIR
Dr. Elizabeth A. (Lianne) Sheppard, Rohm and Haas Professor in Public Health Sciences and
Professor, Department of Environmental & Occupational Health Sciences and Department of
Biostatistics, Hans Rosling Center for Population Health, University of Washington, Seattle, WA
CASAC MEMBERS
Dr. Michelle Bell, Mary E. Pinchot Professor of Environmental Health, Yale University School of the
Environment, New Haven, CT
Dr. James Boylan, Assistant Branch Chief, Air Protection Branch, Environmental Protection Division,
Georgia Department of Natural Resources, Atlanta, GA
Dr. Judith C. Chow, Nazir and Mary Ansari Chair in Entrepreneurialism and Science and Research
Professor, Division of Atmospheric Sciences, Desert Research Institute, Reno, NV
Dr. Mark W. Frampton, Professor Emeritus of Medicine, Pulmonary and Critical Care, University of
Rochester Medical Center, Rochester, NY
Dr. Christina H. Fuller, Associate Professor, Department of Population Health Sciences, Georgia State
University School of Public Health, Atlanta, GA
Dr. Alexandra Ponette-Gonzalez, Associate Professor of Geography and the Environment, University
of North Texas, Denton, TX
CONSULTANTS
Mr. George A. Allen, Senior Scientist, Northeast States for Coordinated Air Use Management
(NESCAUM), Boston, MA
Dr. John R. Balmes, Professor Emeritus, Department of Medicine, University of California, San
Francisco, San Francisco, CA
Dr. Jane Clougherty, Associate Professor, Department of Environmental and Occupational Health,
Dornsife School of Public Health, Drexel University, Philadelphia, PA
Dr. Deborah Cory-Slechta, Professor, Department of Environmental Medicine, School of Medicine
and Dentistry, University of Rochester, Rochester, NY
Dr. Terry Gordon, Professor, Department of Environmental Medicine, New York University School of
Medicine, New York, NY
111
-------
Dr. Michael T. Kleinman, Professor, Department of Medicine, Division of Occupational and
Environmental Medicine, University of California, Irvine, Irvine, CA
Dr. Stephanie Lovinsky-Desir, Assistant Professor, Department of Pediatrics, College of Physicians
and Surgeons, Columbia University, New York, NY
Dr. Jennifer Peel, Professor and Section Head of Epidemiology, Department of Environmental and
Radiological Health Sciences, Colorado State University, Fort Collins, CO
Dr. David Rich, Associate Professor of Epidemiology, Departments of Public Health Sciences,
Medicine, and Environmental Medicine, University of Rochester Medical Center, Rochester, NY
Dr. Jeremy Sarnat, Associate Professor of Environmental Health, Rollins School of Public Health,
Emory University, Atlanta, GA
Dr. Neeta Thakur, Assistant Professor, School of Medicine, University of California at San Francisco,
San Francisco, CA
Dr. Barbara Turpin, Professor and Chair, Environmental Sciences and Engineering, Gillings School of
Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC
Dr. Marc Weisskopf, Cecil K. and Philip Drinker Professor of Environmental Epidemiology and
Physiology, Department of Environmental Health and Department of Epidemiology, Harvard T.H. Chan
School of Public Health, Boston, MA
Dr. Corwin (Cory) Zigler, Professor of Statistics and Data Sciences, Department of Statistics and Data
Sciences, University of Texas at Austin, Austin, TX
SCIENCE ADVISORY BOARD STAFF
Mr. Aaron Yeow, Designated Federal Officer, U.S. Environmental Protection Agency, Science
Advisory Board, Washington, DC
iv
-------
Consensus Responses to Charge Questions on the EPA's
Supplement to the 2019 Integrated Science Assessment for Particulate Matter
(External Review Draft — October 2021)
CASAC Advice on the Process and Timeline for CASAC Reviews
The Clean Air Scientific Advisory Committee (CASAC) commends the Environmental Protection
Agency (EPA) for returning to its long-standing practice of constituting an ad hoc panel of experts to
complement the expertise of the Chartered CASAC. The previous CASAC recognized the need for a
larger review panel, stating in its letter to the Administrator on April 11, 2019: "Additional expertise is
needed for the Clean Air Scientific Advisory Committee (CASAC) to provide a thorough review of the
particulate matter (PM) National Ambient Air Quality Standards (NAAQS) documents. The breadth and
diversity of evidence to be considered exceeds the expertise of the statutory CASAC members, or indeed
of any seven individuals. For example, the Chartered CASAC has found it difficult to achieve consensus
in some areas (summarized below), and to do so likely requires further scientific expertise from, and
discussion with, epidemiologists and additional experts in human clinical studies and toxicology." The
CASAC recommends that the practice of convening a panel of additional experts continue for all future
NAAQS reviews because the give-and-take deliberation and participation of multiple scientific experts,
including multiple experts from all key disciplines needed to conduct a high-quality scientific review, is
fundamental to the Chartered CASAC's ability to provide the highest-quality scientific advice. With a
fully constituted ad hoc panel of experts, the CASAC has a depth and breadth of expertise that enables it
to fulfill its mandate to provide advice and recommendations to the EPA.
The CASAC is concerned that the compressed timeframe for this and other recent CASAC reviews has
made it difficult for the CASAC to provide the highest quality review possible. The compressed
timeframe has necessitated simultaneous review of large documents, which is not optimal. The
timeframe does not allow for development of second drafts of documents that incorporate CASAC
advice (should the CASAC recommend second drafts). The CASAC recommends that for future
reviews, the EPA follow a review plan that allows for adequate time for the CASAC to review the
documents and sufficient time for the EPA to incorporate CASAC advice into second drafts (if
requested by the CASAC). The review plan should also allow sufficient time for CASAC advice on
earlier documents (e.g., Integrated Science Assessment) to also be incorporated into subsequent
documents (e.g., Risk and Exposure Assessment and Policy Assessment). Documents that are sequential
(e.g., the Integrated Science Assessment and Policy Assessment) should not be developed
simultaneously and presented to the CASAC for simultaneous review.
In spite of the compressed timeframe for this review, the CASAC is confident that it was able to
appropriately complete its task. Once the EPA addresses the CASAC's comments, the final Integrated
Science Assessment (ISA) Supplement and Policy Assessment (PA) deserve the Administrator's full
consideration and are adequate for rulemaking.
1
-------
Executive Summary
The Executive Summary is intended to provide a concise synopsis of the key findings and conclusions of
the draft PM Supplement for a broad range of audiences.
a. Please comment on the clarity with which the Executive Summary communicates the key information
from the draft PM Supplement.
b. Please provide recommendations on whether additional information should be added to the Executive
Summary or information that should be left for discussion in the subsequent sections of the draft PM
Supplement.
The Executive Summary is appropriately brief, two pages total. It clearly and concisely explains the
purpose and limited scope of the Draft Supplement to the 2021 ISA for PM, hereafter referred to as the
Draft ISA Supplement, and notes inclusion of recent studies on novel confounding adjustment methods
or accountability analyses, near-ambient experimental studies, disparities in PM2.5 exposure or health
risk by race or socioeconomic status (SES), and effects of PM2.5 exposure on COVID-19 health
outcomes. Passing mention is made of visibility topics, which is appropriate, given the scope and focus
of Draft ISA Supplement. Importantly, the Executive Summary explicitly states that the Draft ISA
Supplement does not include the entire body of literature that supports weight-of-evidence (WOE)
conclusions. It notes the focus of the Draft ISA Supplement on recent studies that support and extend the
causality determinations that were the subject of extensive discussions during the 2019 PM NAAQS
review. The level of detail and information included in the Executive Summary is appropriate for its
intended purpose.
The sentence on page ES-1, lines 21-24, is confusing and unclear, given that the Draft ISA Supplement
attempts to put recent studies into context, and also to draw new WOE conclusions in light of the added
studies. It is not a complete literature review, but is a multidisciplinary evaluation. This sentence could
be reworded to be more specific. Additional recommended changes for clarity in the Draft ISA
Supplement, noted in the other consensus responses, should also be incorporated into the Executive
Summary.
Section 1 - Introduction and Scope
Section 1 consists of an introduction detailing why the draft PM Supplement is being developed along
with the rationale and scope for the topics and studies considered.
Please comment on the clarity of the section, whether the scope is appropriate for the purpose of the
draft PM Supplement, and whether additional information is needed to convey the purpose of the draft
PM Supplement and the basis for the targeted evaluation conducted.
Section 1 is a clear and concise summary of the purpose and scope of the Draft ISA Supplement. It
explains that the scope is limited to health-effect evidence categories that the 2019 PM ISA concluded
had a causal relationship, i.e., cardiovascular effects and mortality associated with short- and long-term
PM2.5 exposure. Although this limitation is appropriate for the targeted purpose of the Draft ISA
Supplement, where consideration is focused on recent literature that could support alternative PM2.5
2
-------
annual and/or daily standards, the CASAC notes that this limitation precludes consideration of any new
evidence that might change a causality determination from "likely to be causal" to "causal" (e.g.,
nervous system effects or respiratory effects). Although this is not expected unless there is substantial
new evidence for those effects, this limitation should be explicitly acknowledged in the introduction.
This section should give some additional background, context, and rationale for the reconsideration of
the December 2020 decision to retain the NAAQS for PM. The CASAC emphasizes the importance of
documenting actions taken by the EPA prior to and during the previous PM NAAQS review, including
the CASAC reviews of the 2019 ISA and 2019 PA.
This background should include a summary of the previous CASAC's consideration of the causal
determination framework, and its recommendation that a "more explicit, systematic, and transparent
process" be used for determining causal relationships. This recommendation resulted in the National
Academies of Sciences, Engineering, and Medicine (NASEM) committee on "Assessing Causality from
a Multidisciplinary Evidence Base for National Ambient Air Quality Standards," which is currently in
deliberations (https://www.nationalacadernies.org/our-work/assessing-causalitv-from-a-
multidisciplinary-evidence-base-for-national-ambient-air-qualitv-standards). Additional detail on this
would inform the reader on the recent developments on this topic since the last review.
The current CASAC notes that the causal determination framework, as documented in the 2015
Preamble to the ISA, is based on WOE and professional judgement. The causal framework has been
developed over many years and has been reviewed and supported by the CASAC during multiple
previous reviews. While continued refinements and alternatives to this WOE framework are currently
being evaluated by the NASEM committee, the CASAC unanimously supports the use of the causal
determination framework for this review and strongly believes that this framework should not be
replaced without a comprehensive evaluation of alternatives.
The CASAC has concerns that constraining the scope of the Draft ISA Supplement to only including
effects with a determination of a causal relationship in the 2019 ISA could set a precedent for future
ISAs. During the CASAC meetings, the EPA clarified that future ISAs would include the full range of
causal classifications. This constraint of the scope should be noted in the introduction.
The scope of the Draft ISA Supplement is limited to U.S. and Canadian epidemiologic studies, and this
needs to be explained in the introduction. Although the EPA provided the rationale for this constraint
during the meeting and clarified that it is only for this Draft ISA Supplement, the CASAC finds this to
be somewhat weak, given that the U.S. and Canadian health care systems and population characteristics
are very different and that the composition of east coast U.S. PM pollution can be very different from
west coast pollution. The Liu et al. (2019) study cited in the Draft ISA Supplement is an example of
PM2.5 health effects coherence across the world. There are also new and relevant European studies on
the health effects of low-level PM2.5 exposures, such as the Health Effects Institute (HEI) ELAPSE
project; see HEI's public comments on the Draft ISA Supplement
(https://casac.epa.gov/ords/sab/f?p=105:19:17351896549913 :::RP.19:P19 ID:962#materials) and the
September 2021 ELAPSE study (Strak et al., 2021).
The Draft ISA Supplement notes the inclusion of several "causal modeling" studies that have appeared
since the 2019 ISA, without specifying exactly what types of modeling strategies this entails. It seems
clear that the EPA is using this term to describe a specific class of established methodologies often
3
-------
called "causal modeling" or "causal inference," that are relatively novel to the purview of the PM ISA.
These methodologies are found particularly in the statistics, epidemiology, and social science literature,
and include instrumental variables, difference in differences, propensity scores, and doubly-robust
additive modeling strategies. These methods are distinct from and should not be conflated with the WOE
framework for causal determinations described in the 2015 Preamble to the ISA. The CAS AC
recommends some introductory comments clarifying why the "causal" label is used for these studies,
even if only for the purpose of contextualizing these studies within the broader epidemiologic literature.
The CASAC is particularly concerned that some will interpret the labeling of some studies as "causal"
as an implication that other epidemiological studies do not support a causal determination or should
receive less weight in any WOE causality determinations. Alternative labeling of these methodologies,
that avoids the word "causal" include "novel confounding adjustment methods," "novel epidemiological
modeling tools," or "novel epidemiologic methods." The use of these terms would be preferable as it
would reduce the chance of conflating these novel tools with the WOE causal determination framework.
Regardless of how they are labeled, the CASAC identifies the importance of these methodologies in the
Draft ISA Supplement for their ability to reduce some of the uncertainties raised by the previous
CASAC and the previous Administrator in response to the 2019 ISA. The CASAC recommends that the
EPA clarify that these methods are not intended to replace the causality determinations of previous
ISAs, but rather have been recently adopted in service of strengthening the body of evidence for
causality determinations.
Regarding the PM causality determinations, the CASAC notes that there is a progression going from the
2009 ISA to the 2019 ISA to this Draft ISA Supplement indicating continued strengthening of the causal
health endpoints relationship with PM2.5. The literature, as it is expanding, continues to show strong
associations with health effects, even though concentrations of PM2.5 in the air have been decreasing
over time.
The 2019 PM ISA concluded a causal relationship for each of the three non-ecological welfare effects
categories evaluated: visibility effects, climate effects, and materials effects. Given that the scope of the
Draft ISA Supplement includes only those studies most informative in considering potential revisions to
the PM NAAQS, the welfare effects are appropriately limited to visibility impairment.
Section 2 - Overview of Main Conclusions of the 2019 Integrated Science Assessment for
Particulate Matter
To ensure that recent studies are put in the context of the conclusions of the 2019 PMISA the draft PM
Supplement pulls in information verbatim from the 2019 PMISA to orient the audience. Two ways this
was done in the draft PM Supplement is through Section 2 which is the Integrated Synthesis Chapter
(i.e., Chapter 1) of the 2019 PMISA and leading off each health and welfare effects discussion in
Section 3 and 4 with the Summary and Causality Determination from the 2019 PMISA.
Please comment on this approach and whether any additional modifications to the structure of the
document can be made to better integrate evidence evaluated in the draft PM Supplement with
conclusions from the 2019 PMISA.
Section 2 provides context for the recent studies in the Draft ISA Supplement from the conclusions of
the 2019 PM ISA. Although the repetition of this contextual material may make it more difficult to
4
-------
quickly focus on the summary and integration of new material in the Draft ISA Supplement and its
relevance to this reconsideration, Section 2 is still a useful addition.
Chapter 3 summarizes the relevant causal determinations from the 2019 ISA in sections 3.1.1.1, 3.1.2.1,
3.2.1.1, and 3.2.2.1. Each is four to six pages long including tables. Combined, the text and tables
provide clear and concise summaries of the information necessary to provide context for the material on
new studies that follow. The CASAC finds these background summaries useful and helpful, but notes
that the transitions from them to the new study material being presented in each section could be
improved to make the overall section more reader-friendly. For a specific example, Section 3.1.1.2
(Recent U.S. and Canadian Epidemiologic Studies), follows a bolded sentence concluding the discussion
of the causal determination of the 2019 ISA with an entire paragraph that is about biological mechanism
before starting to discuss recent U.S. and Canadian epidemiologic studies in the next paragraph. While
the information in the first paragraph of this section is correctly given, the placement of this paragraph is
an example of a difficult transition.
Section 3 - Evaluation of Recent Health Effects Evidence
Section 3 characterizes the recent health effects evidence that falls within the scope of the draft PM
Supplement.
a. Please comment on the identification, evaluation, and characterization of the available
scientific evidence in Section 3.
Overall, the CASAC finds the summary of health effects to be well-written and thorough, with
appropriate identification, evaluation, and characterization of available scientific evidence, including in
the context of the conclusions from the 2019 PM ISA. A few additional studies are identified in Panel
members' individual comments. The CASAC recommends the following improvements.
Consistent with the CASAC's advice pertaining to this topic in Section 1, the text on "causal modeling
methods" should be reworded to clarify these methods' role in and distinction from the WOE causality
determinations outlined in the 2015 Preamble to the ISA. While recent emergence of studies employing
such methods is important to the Draft ISA Supplement, the CASAC recommends alternative
descriptive language to avoid the potential misconception that labeling some studies as "causal" carries
an implication that more traditional epidemiologic analysis methods cannot support a causal
determination or should receive less weight in any WOE causality determinations. Both traditional and
more novel epidemiologic methods should be employed in service of strengthening the body of
evidence.
Regarding the study descriptions and findings, the evaluation and characterization of some studies could
be improved. There are places where more detail would be useful to give context to the results (e.g.,
hazard ratios, sample size, methods for exposure assessment, whether co-pollutant confounding was
assessed by 2-pollutant or multi-pollutant models). Specifically, conclusions are needed for these studies
regarding what was learned, and the ways and degrees in which the studies strengthen or weaken the
state of scientific evidence. Language on sample size should be rephrased to avoid the misconception
that larger studies are always better, and instead focus on the tradeoffs of larger sample size (e.g.,
statistical power) with smaller studies (e.g., often more detailed individual-level data or exposure
5
-------
assessment). Similarly, wording on multi-city versus single-city studies should acknowledge similar
tradeoffs. More information on uncertainty is warranted, such as differences in studies by location,
differences in concentration-response functions (in particular, potential differences in the shape of the
functions and characterization of uncertainty in the shape reported in different publications). The
discussion of heterogeneity of effect estimates largely focuses on variation in particulate matter
components, which may indicate key sources, but more attention should be given towards distinguishing
the variation in particle composition from effect modification by other individual- and contextual-level
variables (e.g., race/ethnicity, urbanicity, housing stock, air conditioning prevalence) that influence
variation in effect estimates. More attention to both disproportionate exposures and effects will allow a
fuller understanding of observed heterogeneity in risk, especially as it relates to the critically important
issue of environmental justice.
While it is addressed in the 2015 Preamble to the ISA, it would be helpful for publication bias to be
acknowledged in this section, along with a comment that while it may add uncertainty, publication bias
cannot explain these overall scientific findings and is not likely to impact the conclusions.
There are multiple terms utilized to describe races and ethnicities in the United States, which is reflected
by the studies included in the Draft ISA Supplement. Race/ethnicity is a fluid concept that is relevant by
time, country, region, population and government. Therefore, the most useful terminology for the
purpose of protecting public health has changed over time. The CASAC recommends updated language
for discussions of race/ethnicity, such as "People of Color (POC)" or "Communities of Color (COC)"
rather than "non-White," and "indicator" rather than "proxy." The CASAC recognizes that the different
uses of the terminology in previous EPA documents as well as in the published literature means that
there will be some nuance to how this recommendation is implemented.
Additional specific comments:
• The CASAC suggests consideration of different language on susceptibility and sensitivity in the
Draft ISA Supplement, and that the term "vulnerability" be dropped.
• Given the importance to the consideration of lowering regulatory limits of studies that focus on
exposures specifically below current standards, the CASAC suggests adding a new subsection in
Section 3 that specifically groups these studies together and describes/summarizes their findings.
• While a footnote on page 3-1 does indicate that "risk estimates from epidemiologic studies
examining short-term exposures are for a 10-[j,g/m3 increase in 24-hour avg PM2.5 concentrations
and long-term exposures are for a 5-[j,g/m3 increase in annual concentrations," this scaling
approach is not as clear as it might be throughout the section. These PM2.5 scaling factors should
be included in each introduction paragraph of each disease-specific section.
• Although there is a definition of how the Draft ISA Supplement defines short-term versus long-
term exposures in Section 2.1, the difference between what is considered short-term and long-
term exposure also needs to be highlighted in Section 3.
• Section 3.3 should be re-ordered, placing the discussion of at-risk populations first and COVID-
19 last. This is recommended to correspond to the more recent nature of the COVID-19
pandemic and considerations for the ISA.
6
-------
• Additional suggested edits and minor errors requiring correction can be found in the individual
comments from Panel members.
b. Please comment on whether the summary sections in Section 3 appropriately characterize recent
evidence in the context of the conclusions of the 2019 PMISA.
The summary subsections appropriately characterize recent evidence in the context of the causal
determinations of the 2019 PM ISA.
c. Please comment on whether there are any topics or studies that fall within the scope of the draft
PM Supplement that should be added or receive additional discussion in Section 3 or any topics for
which discussion should be shortened or removedfrom Section 3.
The biological plausibility and mechanistic pathways of the associations between both short-term and
long-term exposures to PM2.5 and health outcomes could be better highlighted and developed. While
there is no need to discuss biological plausibility and mechanisms in great detail, as was done in the
2019 PM ISA, it should be made clear that the evidence in support of these concepts remains strong and
these lend confidence to the causal determinations.
Section 3.3.3 (Populations and Lifestages at Potentially Increased Risk of a PM-Related Health Effect)
would benefit from additional considerations regarding SES data. The level of SES data is an important
component in understanding research questions, modeling and interpretation of results. Individual-level
and community-level data differ with regards to their representativeness and accuracy, and the level of
resolution utilized in a study could drastically change interpretation of results. The CASAC suggests that
studies that utilize individual-level SES data be distinguished from those that use community-level data,
including in summaries of findings. Dividing the discussion into individual- and community-level
measures will help the reader digest the data.
The articles by Yitshak-Shade et al., Son et al. (2020), Crouse et al. (2019), Stieb et al. (2020) include
the impact of green space on associations between PM2.5 and mortality. However, there is no discussion
of green space in the Draft ISA Supplement. The EPA could consider removing these references. If
these studies are to be included in the Draft ISA Supplement, then there should be additional assessment
of the linkages between green space and PM2.5 concentrations; green space and health; and quality of
green space by race/ethnicity.
In Section 3.3.3.1., the CASAC encourages the EPA to explicitly state when including interpretation of
results from studies that include race within indices of vulnerability versus those that only include
socioeconomic variables. Combining the interpretations of these studies may lead to conflation of the
influences of socioeconomic status and racism.
Section 4 - Evaluation of Recent Welfare Effects Evidence
Section 4 characterizes the recent welfare effects evidence that falls within the scope of the draft PM
Supplement.
7
-------
a. Please comment on the identification, evaluation, and characterization of the available
scientific evidence in Section 4.
b. Please comment on whether the summary section in Section 4 appropriately characterizes recent
evidence in the context of the conclusions of the 2019 PMISA.
c. Please comment on whether there are any topics or studies that fall within the scope of the draft
PM Supplement that should be added or receive additional discussion in Section 4 or any topics for
which discussion should be shortened or removedfrom Section 4.
The CASAC agrees with the assessment in Section 4 that recent evidence confirms the "causal
relationship" between PM and visibility impairment. However, a few clarifications are needed.
Section 4 acknowledges variations in humanly perceivable acceptance of visibility levels. The use of
"apparent contrast" instead of "total light extinction" shows less variability to determine acceptable
visual air quality. Values for the cut-off for unacceptable versus acceptable levels for atmospheric light
extinction and non-subjective contrast measurements should be clarified. Science-based visibility
standards warrant additional research using objective scenarios to quantify visibility improvement.
With increases in the frequency and intensity of wildfires and reductions in sulfur dioxide (SO2) and
oxides of nitrogen (NOx) emissions, mass scattering coefficients for the major PM2.5 components (e.g.,
organic and elemental carbon, sulfate, and nitrate) need to be further examined with more recent data
(e.g., 2015 onward). The "revised" Interagency Monitoring of Protected Visual Environments
(IMPROVE) chemical extinction equation (Pitchford et al., 2007) that has been recommended by the
EPA and applied to the most recent Regional Haze State Implementation Plans should be included. In
addition, the Lowenthal and Kumar (2016) IMPROVE equation should be added to the document along
with discussion in detail. Differences among the various IMPROVE equations that use split component
algorithms need to be addressed.
Analysis of regional variations between PM2.5 composition, light extinction, and views may provide
some perspective on causes and variations of organic mass in the intermountain west and southwest,
increases in nitrate and ammonium in the central U.S., along with a decline in sulfate/organic ratios and
increasing organosulfate/total sulfate ratios in the southeast (Riva et al., 2019). These changes may alter
optical properties that are not accounted for by the IMPROVE chemical extinction equations.
Recent evidence points to effects of microplastics in PM on climate (e.g., Revell et al., 2021), dry and
wet deposition (Brahney et al., 2020), and ecosystems (e.g., Halle et al., 2020; Sobhani et al., 2022).
These and other emerging PM components require further consideration in future reviews.
Section 5 - Summary and Conclusions
The Summary and Conclusions section (Section 5) provides an overview of the evidence evaluated
in the draft PM Supplement.
Please comment on the level of detail provided within this section and whether revisions
should be made to further summarize recent evidence.
8
-------
Section 5 is a concise 4-page summary of the Draft ISA Supplement. There is a single paragraph
summarizing each of the four major health outcomes for which exposure to PM2.5 was determined to
have a causal relationship in the 2019 ISA (short- and long-term cardiovascular and mortality effects)
The "Additional Considerations" are also summarized. The summary is accurate, with an appropriate
level of detail. The CASAC suggests conclusions be added to the summaries focused on COVID-19
(page 5-3), and visibility effects (page 5-4). Additional recommended changes for clarity in the Draft
ISA Supplement made elsewhere in the consensus responses should be incorporated into this section.
Finally, the CASAC recommends that the EPA incorporate into this section some specific proposed
edits contained in Panel members' individual comments that will further improve balance and clarity.
9
-------
References
Brahney, J., Hallerud, M., Heim, E., Hahnenberger, M., & Sukumaran, S. (2020). Plastic rain in
protected areas of the United States. Science, 3(55(6496), 1257-1260.
http s: //doi. or g/ 10.1126/sci ence. aaz5 819
Di, Q., Wang, Y., Zanobetti, A., Wang, Y., Koutrakis, P., Choirat, C., Dominici, F., & Schwartz, J. D.
(2017). Air pollution and mortality in the Medicare population. New England Journal of Medicine,
376{26), 2513-2522. https://doi.org/10.1056/neimoal7Q2747
Halle, L. L., Palmqvist, A., Kampmann, K., & Khan, F. R. (2020). Ecotoxicology of Micronized Tire
Rubber: Past, present and future considerations. Science of The Total Environment, 706, 135694.
http s: //doi. or g/10.1016/j. scitotenv .2019.135694
Liu, C., Chen, R., Sera, F., Vicedo-Cabrera, A.M., Guo, Y., Tong, S., Coelho, M., Saldiva, P.H.N.,
Lavigne, E., Matus, P., Valdes Ortega, N., Osorio Garcia, S., Pascal, M., Stafoggia, M., Scortichini, M.,
Hashizume, M., Honda, Y., Hurtado-Diaz, M., Cruz, J., Nunes, B., Teixeira, J.P., Kim, H., Tobias, A.,
Iniguez, C., Forsberg, B., Astrom, C., Ragettli, M.S., Guo, Y.L., Chen, B.Y., Bell, M.L., Wright, C.Y.,
Scovronick, N., Garland, R.M., Milojevic, A., Kysely, J., Urban, A., Orru, H., Indermitte, E., Jaakkola,
J.J.K., Ryti, N.R.I., Katsouyanni, K., Analitis, A., Zanobetti, A., Schwartz, J., Chen, J., Wu, T., Cohen,
A., Gasparrini, A., Kan, H. (2019). Ambient particulate air pollution and daily mortality in 652 Cities.
New England Journal of Medicine, 357(8), 705-715. https://doi.org/10.1056/neimoal817364
Lowenthal, D. H., & Kumar, N. (2016). Evaluation of the IMPROVE equation for estimating aerosol
light extinction. Journal of the Air & Waste Management Association, 66(7), 726-737.
https://doi.org/10.1080/10962247.2Q16.1178187
Pitchford, M., Malm, W., Schichtel, B., Kumar, N., Lowenthal, D., & Hand, J. (2007). Revised
algorithm for estimating light extinction from IMPROVE particle speciation data. Journal of the Air &
Waste Management Association, 57(11), 1326-1336. https://doi.Org/10.3155/1047-3289.57.l 1.1326
Revell, L. E., Kuma, P., Le Ru, E. C., Somerville, W. R., & Gaw, S. (2021). Direct radiative effects of
airborne microplastics. Nature, 595(7881), 462-467. https://doi.org/10.1038/s41586-021 -03864-x
Riva, M., Chen, Y., Zhang, Y., Lei, Z., Olson, N.E., Boyer, H.C., Narayan, S., Yee, L.D., Green, H.S.,
Cui, T., Zhang, Z., Baumann, K., Fort, M., Edgerton, E., Budisulistiorini, S.H., Rose, C.A., Ribiero,
I.O., e Oliveira, R.L., dos Santos, E.O., Machada, C.M.D., Szopa, S., Zhao, Y., Alves, E.G., de Sa, S.S.,
Hu, W., Knipping, E.M., Shaw, S.L., Duvoisin, S., de Souza, R.A.F., Palm, B.B., Jimenez, J.L., Glasius,
M., Goldstein, A.H., Pye, H.O.T., Gold, A., Turpin, B.T., Vizuete, W., Martin, S.T., Thortnton, J.A.,
Dutcher, C.S., Ault, A.P., Surratt, J.D., (2019). Increasing isoprene epoxydiol-to-inorganic sulfate
aerosol ratio results in extensive conversion of inorganic sulfate to organosulfur forms: Implications for
aerosol physicochemical properties. Environmental Science & Technology, 53(15), 8686-8694.
https://doi.org/10.1021/acs.est.9b01Q19
10
-------
Sobhani, Z., Panneerselvan, L., Fang, C., Naidu, R., & Megharaj, M. (2022). Chronic and
transgenerational effects of polyethylene microplastics at environmentally relevant concentrations in
earthworms. Environmental Technology & Innovation, 25, 102226.
https://doi.org/10.1016/i .eti.2021.102226
Strak, M., Weinmayr, G., Rodopoulou, S., Chen, J., Hoogh, K. D., Andersen, Z. J., Atkinson, R. K.,
Bauwelinck, M. J., Bekkevold, T. T., Bellander, T. M., Boutron-Ruault, M. U., Brandt, J. U., Cesaroni,
G. U., Concin, H. U., Fecht, D. U., Forastiere, F. U., Gulliver, J. U., Hertel, O. U., Hoffmann, B. U.,
Hvidtfeldt, U. U., Janssen, N. U., Jockel, K. U., J0rgensen, J. U., Ketzel, M. U., Klompmaker, J. U.,
Lager, A. U., Leander, K. U., Liu, S. U., Ljungman, P. U., Magnusson, P. U., Mehta, A. U., Nagel, G.
U., Oftedal, B. U., Pershagen, G. U., Peters, A. U., Raaschou-Nielsen, O. U., Renzi, M. U., Rizzuto, D.
U., Schouw, Y. U., Schramm, S. U., Severi, G. U., Sigsgaard, T. U., S0rensen, M. U., Stafoggia, M. U.,
Tj0nneland, A. U., Verschuren, W. U., Vienneau, D. U., Wolf, K. U., Katsouyanni, K. U., Brunekreef,
B. U., Hoek, G. U., & Samoli, E. (2021). Long term exposure to low level air pollution and mortality in
eight European cohorts within the ELAPSE PROJECT: Pooled analysis. BMJ.
https://doi.org/10.1136/bmi.nl904
11
-------
Appendix A
Individual Comments by the 2021 CASAC Particulate Matter Review Panel Members
on the EPA's Supplement to the 2019 Integrated Science Assessment for Particulate Matter
(External Review Draft — October 2021)
Mr. George A. Allen A-2
Dr. John R. Balmes A-5
Dr. Michelle Bell A-8
Dr. James Boylan A-12
Dr. Judith C. Chow A-16
Dr. Jane Clougherty A-20
Dr. Deborah Cory-Slechta A-22
Dr. Mark W. Frampton A-24
Dr. Christina H. Fuller A-28
Dr. Terry Gordon A-33
Dr. Michael T. Kleinman A-37
Dr. Stephanie Lovinsky-Desir A-40
Dr. Jennifer Peel A-42
Dr. Alexandra Ponette-Gonzalez A-45
Dr. David Rich A-48
Dr. Jeremy Sarnat A-52
Dr. Neeta Thakur A-55
Dr. Barbara Turpin A-58
Dr. Marc Weisskopf A-61
Dr. Corwin Zigler A-64
A-l
-------
Mr. George A. Allen
Charge Question 2 - Section 1 consists of an introduction detailing why the draft PM Supplement is
being developed along with the rationale and scope for the topics and studies considered.
a. Please comment on the clarity of the section, whether the scope is appropriate for the purpose of the
draft PM Supplement, and whether additional information is needed to convey the purpose of the draft
PM Supplement and the basis for the targeted evaluation conducted.
Section 1 is a clear and concise summary of the purpose and scope of the ISA Supplement. This section
explains (1.2.1) that the scope of the document is limited to health effect evidence categories that the
2019 PM ISA concluded had a causal relationship, e.g., limited to short- and long-term PM2.5 exposure
and cardiovascular effects and mortality. This is appropriate for the targeted purpose of the ISA
Supplement, where consideration is focused on recent literature that could support alternative PM2.5
annual and/or daily standards.
The scope of the Supplement is limited to U.S. and Canadian epidemiologic studies. I did not find an
explanation for that in this section; there are new and relevant European study publications on health-
effects of low level PM2.5 such as the HEI ELAPSE project. See HEI's ISA Supplement comments and
the September 2021 ELAPSE study publication (https://www.bmi.com/content/374/bmi.nl904).
The material on welfare effects in this section is brief and limited to visibility impairment, but that is
appropriate given scope of this Supplement and its primary focus on the PM2.5 health-based standards.
Charge Question 3 - To ensure that recent studies are put in the context of the conclusions of the 2019
PM ISA the draft PM Supplement pulls in information verbatim from the 2019 PM ISA to orient the
audience. Two ways this was done in the draft PM Supplement is through Section 2 which is the
Integrated Synthesis Chapter (i.e., Chapter 1) of the 2019 PMISA and leading off each health and
welfare effects discussion in Section 3 and 4 with the Summary and Causality Determination from the
2019 PMISA.
a. Please comment on this approach and whether any additional modifications to the structure of the
document can be made to better integrate evidence evaluated in the draft PM Supplement with
conclusions from the 2019 PMISA.
While providing context of recent studies to the conclusions of the 2019 PM ISA is necessary, I found
the repeated emphasis of the 2019 findings somewhat excessive for an audience who is already familiar
with the previous PM review. The repetition of this contextual material made it more difficult for me to
quickly focus on the summary and integration of new material in the Supplement and its relevance to
this review. This document is a Supplement and, as is stated in the introduction, is not intended to stand
on its own for support of this reconsideration review.
Chapter 3 summarizes the relevant material from the 2019 ISA in sections 3.1.1.1, 3.1.2.1, 3.2.1.1, and
3.2.2.1. Each is four to six pages long including tables. Combined, the text and tables provide a good
A-2
-------
summary of the information necessary to provide context for the material on new studies that follow.
The summaries are clear and concise. Given these summaries, perhaps the 2019 Integrated Synthesis
material in Chapter 2 could be incorporated by reference to help focus the document on new studies.
For the summary of PM2.5 short-term cardiovascular effects, the reference to table 3-4 (total mortality)
on page 3-3 line 8 should be to table 3-1 directly below it.
Charge Question 1 - The Executive Summary is intended to provide a concise synopsis of the key
findings and conclusions of the draft PM Supplement for a broad range of audiences.
a. Please comment on the clarity with which the Executive Summary communicates the key information
from the draft PM Supplement.
b. Please provide recommendations on whether additional information should be added to the Executive
Summary or information that should be left for discussion in the subsequent sections of the draft PM
Supplement.
The Executive Summary is brief, two pages total. It clearly and concisely explains the purpose and
limited scope of the Supplement, and notes inclusion of recent studies on causal modeling methods or
accountability analyses, near-ambient experimental studies, disparities in PM2.5 exposure or health risk
by race or SES, and effects of PM2.5 exposure on Covid-19 health outcomes. Passing mention is made
of visibility topics, which is ok given the focus of this reconsideration on the primary health standards.
Importantly, this Summary explicitly states that the Supplement does not include the entire body of
literature that supports WOE conclusions. It notes the focus of the Supplement on recent studies that
support and extend the causality determinations that were the subject of extensive discussions during the
2019 PM NAAQS review. The level of detail and information included in the Summary is appropriate
for its intended purpose.
Charge Question 6 - The Summary and Conclusions section (Section 5) provides an overview of the
evidence evaluated in the draft PM Supplement.
a. Please comment on the level of detail provided within this section and whether revisions should be
made to further summarize recent evidence.
Section 5 is a concise 4-page summary of the findings from recent studies considered in this
Supplement. There is a single paragraph summarizing each of the four major PM2.5 health outcome
causal categories (short- and long-term cardiovascular and mortality effects). Detail on each is minimal
but sufficient for this overview.
Consider using "causal" relationship for these summaries, because causality is a critical topic of this
reconsideration. Example: "... different statistical approaches and cohorts spanning diverse geographic
locations and populations provide additional support for the «causal» PM2.5-mortality relationship."
A-3
-------
Two additional short paragraphs cover near-ambient experimental studies and Covid-19, and two
additional bullets summarize populations and lifestages risk findings with SES and race. Health risk
disparities for Black populations seems to be de-emphasized relative to higher PM2.5 exposures in
predominantly Black neighborhoods. I'd consider this backwards (see Di et al. 2017b, (NEJM, figure
2a), and worthy of more attention in general.
Visibility effects are bulleted under the Populations and Lifestages heading and should be separate.
A-4
-------
Dr. John R. Balmes
Charge Question 1 - The Executive Summary is intended to provide a concise synopsis of the key
findings and conclusions of the draft PM Supplement for a broad range of audiences.
a. Please comment on the clarity with which the Executive Summary communicates the key information
from the draft PM Supplement.
The Executive Summary clearly and appropriately communicates the information underlying the causal
relationship determinations for short-term and long-term exposures to PM2.5 and cardiovascular effects
and mortality.
Charge Question 2 - Section 1 consists of an introduction detailing why the draft PM Supplement is
being developed along with the rationale and scope for the topics and studies considered.
a. Please comment on the clarity of the section, whether the scope is appropriate for the purpose of the
draft PM Supplement, and whether additional information is needed to convey the purpose of the draft
PM Supplement and the basis for the targeted evaluation conducted.
The scope and clarity of Section 1 are appropriate for the introduction to the organization of the
Supplement.
Charge Question 3 - To ensure that recent studies are put in the context of the conclusions of the 2019
PMISA the draft PM Supplement pulls in information verbatim from the 2019 PM ISA to orient the
audience. Two ways this was done in the draft PM Supplement is through Section 2 which is the
Integrated Synthesis Chapter (i.e., Chapter 1) of the 2019 PMISA and leading off each health and
welfare effects discussion in Section 3 and 4 with the Summary and Causality Determination from the
2019 PMISA.
a. Please comment on this approach and whether any additional modifications to the structure
of the document can be made to better integrate evidence evaluated in the draft PM
Supplement with conclusions from the 2019 PMISA.
My major comment about the approach to the integration of evidence in both the 2019 PM ISA and the
Supplement is the characterization of the relationship of both short-term and long-term exposures to
PM2.5 of the respiratory effects as "likely to be causal." First, my assessment of the epidemiological
evidence is that it supports a causal relationship, even if experimental evidence is weaker than for
cardiovascular effects. Second, I find it somewhat logically inconsistent with the likely causal
determination for respiratory effects that in Chapter 3 of the Supplement respiratory tract inflammation
is used as part of the biological mechanism rationale for cardiovascular effects. I also find this
inconsistency with regard to the discussions of respiratory-specific mortality (used to support the
mortality causal relationship determination) and growth of lung function in children (used to support the
vulnerability of children). In addition, the likely causal relationship determination for lung cancer (and
A-5
-------
lack of evidence for other cancers) also supports the respiratory toxicity of exposures to PM2.5
exposures.
I also think that the discussion of the experimental evidence regarding the carcinogenicity of PM2.5 "as
a whole" is somewhat misleading. While I understand that the NAAQS is based on particle size and that
the epidemiological evidence is largely for PM2.5 rather than for its components, chemical composition
undoubtedly determines carcinogenicity. Discussing experimental studies of PM2.5 carcinogenicity
without comment about the source and characterization of the particulate matter used in the studies gives
the impression that exposure to any type of PM2.5 can increase risk of lung cancer, but I doubt that this
is likely.
Section 2.2.5 on Populations and Lifestages at Potentially Increased Risk of a PM-Related Health Effect
should list all of the populations determined in the 2019 ISA to be vulnerable so as to avoid confusion.
Charge Question 4 - Section 3 characterizes the recent health effects evidence that falls within the scope
of the draft PM Supplement.
a. Please comment on the identification, evaluation, and characterization of the available
scientific evidence in Section 3.
In general, it is my opinion that the characterization and evaluation of the available evidence in Section
3 is appropriate.
b. Please comment on whether the summary sections in Section 3 appropriately characterize
recent evidence in the context of the conclusions of the 2019 PMISA.
It is my opinion that the summary sub-sections of Section 3 appropriately characterize the recent
evidence.
c. Please comment on whether there are any topics or studies that fall within the scope of the
draft PM Supplement that should be added or receive additional discussion in Section 3 or
any topics for which discussion should be shortened or removedfrom Section 3.
I would like to see a discussion of Respiratory Effects, especially given new evidence that wildfire-
specific PM may be more potent than non-wildfire PM for respiratory outcomes.
Charge Question 5 - Section 4 characterizes the recent welfare effects evidence that falls within the
scope of the draft PM Supplement.
a. Please comment on the identification, evaluation, and characterization of the available
scientific evidence in Section 4.
No comments at this time.
A-6
-------
b. Please comment on whether the summary section in Section 4 appropriately characterizes
recent evidence in the context of the conclusions of the 2019 PMISA.
No comments at this time.
c. Please comment on whether there are any topics or studies that fall within the scope of the
draft PM Supplement that should be added or receive additional discussion in Section 4 or
any topics for which discussion should be shortened or removedfrom Section 4.
No comments at this time.
Charge Question 6 - The Summary and Conclusions section (Section 5) provides an overview of the
evidence evaluated in the draft PM Supplement.
a. Please comment on the level of detail provided within this section and whether revisions
should be made to further summarize recent evidence
I think the level of detail is OK.
A-7
-------
Dr. Michelle Bell
Overall, the supplement to the ISA is extremely well-written and thorough. Below are my preliminary
comments.
Page ES-2 and ES-3: The statement that "there may be PM2.5 exposure and health risk disparities by
socioeconomic status (SES), specifically among people of low SES." seems understated given the
current scientific evidence. I appreciate that this bullet point distinguishes between exposure and health
risk disparities for both racial/ethnic minorities and SES.
Executive Summary: Growing evidence indicates that PM chemical structure impacts the health impact
of PM, although the current scientific literature cannot disentangle the various sources, chemical
structures, and components to identify a specific PM characteristic to target, other than by size. This
might be worth noting as a bullet point in the Executive Summary. Although it does not change the
outcome of the resulting recommendation, the growing science in this area may be worth highlighting.
Page 1-1: The definition of "welfare effects" to exclude ecological effects associated with particulate
matter, even if used consistently in this document, differs from how this phrase is commonly applied.
This is described as a footnote on page 1-1. To make sure readers are aware of this definition of welfare
effects, it may be worth moving this footnote to the main text. If possible, it may be useful to have
different wording other than "welfare effects" to note that this is a subset of welfare effects, if
appropriate alternative phrasing can be developed.
Page 1-3 and 1-4: The focus on U.S. and Canadian studies is reasonable but could be justified if there is
room. A focus is different from excluding other studies. Excluding studies from other locations entirely
is questionable is they do provide critical scientific evidence. Similarly, excluding studies without causal
methods of accountability analysis is negating decades of critical literature including many more recent
and valuable studies that. The evaluation of causality based only on these methods is questionable.
Although, it is appreciated that the text notes that other studies exist (end of Section 1.2). This issue
arises throughout the supplement.
Page 2-7: Define CAPS at first use. This is important as CAPS is sometimes used in the air pollution
context to mean criteria air pollutants.
Page 3-1: The exclusion of studies outside of Canada and the United States needs more justification (see
comment above). A focus on these regions could relate to population characteristics and pollution levels
as well as composition. However, excluding all other studies would require more justification than is
provided. This issue is further complicated by the inclusion of studies from outside these countries, such
as Asian and Australian studies in Section 3.
Page 3-1: The disparities of race/ethnicity and socioeconomic status (SES) are examples of
environmental health disparities, not the only ones. This sentence states that they were the only ones
considered. Are they meant to be listed as examples?
A-8
-------
Tables 3-1: The column heading of PM2.5 Concentrations Associated with Effects is unclear and the
footnote is not particularly helpful ("c Describes the PM2.5 concentrations with which the evidence is
substantiated.") This could refer to concentrations for which effects were estimated in situations for
which no effect was estimated for other concentrations, or to merely the levels of PM2.5 for those study
areas and timeframe, meaning the levels for the study, in which case it might mean the average
concentration over the study or incorporate the full range. As a minor point, but one that will aid clarity,
"effects" could be positive or negative and here this column means harmful effects. It is also unclear
why concentrations are missing from some rows where they could be provided.
Page 3-4: As a minor note, the footnote to this table could denote CMAQ as the Community Multiscale
Air Quality model, as some readers will not know this is a modeling framework.
Page 3-9: The sentence "As expected, the city specific estimates were relatively uncertain and
heterogeneous across cities when there were a small number of daily ED visits." does add much. It may
not be needed.
Page 3-10: The description of Figure 3-1 should explain what lag structures were selected as most of
these studies included sensitivity analysis with multiple lag structures. This is particularly important as
results are generally robust to alternative lags. This could link to Section 3.1.1.2.8.
Page 3-11: Minor note: For Figure 3-1, the notes column seems inconsistent in terms of what
information is included as location information is included both there and in the location column.
Page 3-12: When describing a study that was intended to study racial differences, those results could be
presented here.
Pages 3-14, 3-16, and 3-18: The comment regarding lag structure for Figure 3-1 also applies to Figure 3-
2, Figure 3-3, and 3-4.
Page 3-18: The phrasing that large studies have higher counts "potentially providing statistical power
needed to perform stratified analyses." could be better worded. What is meant here is the ability to
detect associations. Small datasets can be stratified, although they are less informative. Such analyses in
large, or small, datasets may be stratified or may be through other methodologies.
Page 3-20: Throughout this document, it would be helpful to note where new findings are consistent
with the earlier literature. Page 3-20 on co-pollutant adjustment is an example, but this is a broader
issue. Section 3.1.1.2.7 nicely describes some studies on co-pollutants, but these are not addressing a
new issue and a reader might misinterpret this as new evidence rather than building on the existing
evidence of co-pollutants as potential confounders for the PM-health associations.
Page 3-20: The joint exposure studies are interesting, but still limited in their ability to disentangle the
complex effects of PM with other pollutants, similar to the efforts to disentangle impacts from various
PM sources. The text here does not adequately describe this issue. Also, it is in a section titled Potential
Copollutant Confounding, although looking at joint effects is not a true confounding study.
Page 3-24: There is no basis for the statement "Uncertainty related to exposure assessment was
generally reduced with consideration of studies 6 included in the 2019 PM ISA that applied hybrid
A-9
-------
exposure assessment techniques that combined land use 7 regression with satellite AOD measurements
and PM2.5 concentrations measured at fixed site monitors." In fact, the opposite is likely true. These
approaches add uncertainty in exposure assessment and add more study areas, and therefor populations.
They do not make exposure estimates more certain better than a monitor measurement.
Page 3-26: Table 3-2: The comment above for Table 3-1 regarding concentration specifications applies
here as well.
Page 3-47: Figure 3-12 may not be needed.
Page 3-49: This section is on mortality with subsections on short and long-term exposure, whereas the
previous sections are on short or long-term exposure with subsections on types of mortality. The overall
structure of this document could be confusing and repetitive.
Page 3-50: The above comment for Table 3-1 regarding concentration specifications applies here as
well. This also applies to subsequent tables with similar format.
Page 3-51: Differences in population characteristics also contribute to the different findings of PM and
health across regions.
Page 3-51: The phrase "exposure factors" is vague. Perhaps this could be more specific or include some
examples to aid readers.
Page 3-52: There are not studies that use "all available PM2.5 data" but studies that use additional
sources of PM2.5 data beyond monitors.
Page 3-53: The discussion of factors contributing to regional heterogeneity on PM-health associations is
missing text on differences in population characteristics, which has been examined in several studies.
Page 3-56: The comment above regarding lag structure for Figure 3-1 also applies to Figure 3-13. In this
case, footnotes for some studies not sensitivity analysis for lag structures, but others do not. An
alternative strategy would be to note, in the text describing this figure, that multiple lag structures were
examined and findings were generally robust, and to explain why the specific lags were selected for
inclusion in the figure.
Page 3-56: The Liu et al. study is best described as "worldwide" as there are some regions that are not
well represented (e.g., Africa).
Page 3-56 to 3-57. Liu et al. did not "establish" the MCC network, but this study is part of the MCC
projects. The MCC existed well before this paper and was not established by Liu. This text needs to be
reworded.
Page 3-59: The comment above for Figure 3-1 regarding lag structure applies here as well.
Page 3-60: The section on effect modification is a bit muddled as it discusses confounding as well. It
also focuses on effect modification by specific community-level factors and ignores many others that
A-10
-------
have been examined, such as community-level demographics beyond those mentioned here. This whole
subsection warrants a closer look at its goal, focus, and structure.
Page 3-63: Figure 3-16 is fine, but it provides information on the magnitude of the central estimate
without any information on its statistical uncertainty. If a map of different effect estimates across
location is used, one that incorporates statistical uncertainty would be more informative.
Page 3-65: See comment above regarding Liu et al., which is not "worldwide". A global multi-city study
would be more accurate.
Page 3-81: The text noting HEI is a odd as the funding sources of other studies are not mentioned.
Page 3-93: Figure 3-25 needs a better title that is more descriptive. This is the estimated loss in life
expectancy for existing levels of PM2.5 compared to a threshold of 2.8 migrogm/m3, which is the
lowest observed level.
Page 3-95: These methods to address confounding have been used in many previous studies, so they
might not be best labelled as novel. This information would be better placed in a section on confounders
rather than its own section on methods.
Page 3-95: The text on temporal trends needs to be clear that this refers to temporal trends in
confounders, not in the association. This applies to text in later pages as well.
Page 3-128: Section 3.3.3.1.1 should note that disparities in exposure are noted by PM2.5 chemical
composition, not just total mass ofPM2.5
Section 3 charge question: Overall, this section well characterizes the identification, evaluation, and
characterization of the available scientific evidence; appropriate characterizes recent evidence in the
context of the conclusions of the 2019 PM ISA.
A-ll
-------
Dr. James Boylan
1. The Executive Summary is intended to provide a concise synopsis of the key findings and conclusions
of the draft PM Supplement for a broad range of audiences.
a. Please comment on the clarity with which the Executive Summary communicates the key information
from the draft PM Supplement.
b. Please provide recommendations on whether additional information should be added to the Executive
Summary or information that should be left for discussion in the subsequent sections of the draft PM
Supplement.
The Executive Summary clearly communicates the key information from the draft PM Supplement. I
have no recommendations for information that should be added or deleted from the Executive Summary.
2. Section 1 consists of an introduction detailing why the draft PM Supplement is being developed along
with the rationale and scope for the topics and studies considered.
a. Please comment on the clarity of the section, whether the scope is appropriate for the purpose of the
draft PM Supplement, and whether additional information is needed to convey the purpose of the draft
PM Supplement and the basis for the targeted evaluation conducted.
In general, the purpose and scope of the draft PM ISA Supplement is clearly presented. It is appropriate
for the draft PM ISA Supplement to focuses on the health effects and the welfare effects where the 2019
PM ISA concluded a causal relationship.
One of the important objectives of the ISA is to make causal determinations that are then used in the
REA and PA documents. In the previous PM review, CASAC questioned the current causal
determination framework which is based on weight-of-evidence and professional judgement leading to
results than can't be replicated by others. Instead, the CASAC recommended that a "more explicit,
systematic, and transparent process" should be used for determining causal relationships. According to
the CASAC letter dated April 11, 2019 to Administrator Wheeler, ".. .the CASAC finds that the Draft
ISA does not present adequate evidence to conclude that there is likely to be a causal relationship
between long-term PM2.5 exposure and nervous system effects; between long-term ultrafine particulate
(UFP) exposure and nervous system effects; or between long-term PM2.5 exposure and cancer." This is
an example of two different groups looking at the same evidence and coming to different
conclusions on the causal relationships. In the final ISA, EPA agreed that the causal relationship
between long-term ultrafine particulate (UFP) exposure and nervous system effects should be changed
from "Likely to be Causal Relationship" to "Suggestive of, but not Sufficient to Infer, a Causal
Relationship". This is an example of the same group looking at the same evidence and coming to a
different conclusion.
The previous CASAC recommendation on the current causal determination framework resulted in the
National Academies of Sciences, Engineering, and Medicine (NASEM) committee on "Assessing
A-12
-------
Causality from a Multidisciplinary Evidence Base for National Ambient Air Quality Standards"
(https://www.nationalacademies.org/our-work/assessing-causalitv-from-a-multidisciplinary-evidence-
base-for-national-ambient-air-qualitv-standards). Additional detail on this would inform the reader on
the recent developments on this topic since the last review.
3. To ensure that recent studies are put in the context of the conclusions of the 2019 PMISA the draft
PM Supplement pulls in information verbatim from the 2019 PMISA to orient the audience. Two ways
this was done in the draft PM Supplement is through Section 2 which is the Integrated Synthesis Chapter
(i.e., Chapter 1) of the 2019 PMISA and leading off each health and welfare effects discussion in
Section 3 and 4 with the Summary and Causality Determination from the 2019 PMISA.
a. Please comment on this approach and whether any additional modifications to the structure of the
document can be made to better integrate evidence evaluated in the draft PM Supplement with
conclusions from the 2019 PMISA.
Pulling information directly from the 2019 PM ISA is a good approach to lay the foundation for the new
information presented in the draft PM ISA Supplement.
5. Section 4 characterizes the recent welfare effects evidence that falls within the scope of the draft PM
Supplement.
a. Please comment on the identification, evaluation, and characterization of the available scientific
evidence in Section 4.
b. Please comment on whether the summary section in Section 4 appropriately characterizes recent
evidence in the context of the conclusions of the 2019 PMISA.
c. Please comment on whether there are any topics or studies that fall within the scope of the draft PM
Supplement that should be added or receive additional discussion in Section 4 or any topics for which
discussion should be shortened or removedfrom Section 4.
The use of contrast rather than total light extinction appears to make the level of acceptable visual air
quality across different locations more consistent.
Pages 4-4 and 4-5 discusses "an unacceptable level of visibility impairment". However, the definition of
"unacceptable" is not discussed. It appears that if 50% or more of the observers rated the visibility
acceptable, then it was classified as "acceptable"; and if less than 50% of the observers rated the
visibility acceptable, then it was classified as "unacceptable". If so, the document should clearly state
this definition.
Page 4-5 states "When the features approximately reach the visual range, corresponding to a contrast
between about -0.03 to -0.05, about 50% of observers rated the image as not acceptable." However,
Page 4-6 states ".. .that visibility preference studies suggest that about 50% of individuals would find
visibility unacceptable if at any time the more distant landscape features nearly disappear, and that this
occurs when these features are near the visual range and have contrast levels of approximately -0.02 to -
0.05". The text should be updated to be consistent since both sentences are discussing the 50%
A-13
-------
unacceptable contrast level based on Figure 4-2. To me, it appears that -0.03 is more appropriate than -
0.02.
Page 4-6 states "Further, an acceptability level of 90% would require contrast levels to remain above a
level of about -0.01." That statement does not seem accurate. To me, it appears that an acceptability
level of 90% would require contrast levels to remain below a level of about -0.05 or -0.07.
Page 4-7 mentions the "original" IMPROVE equation, "revised" IMPROVE equation, and the
"modified" IMPROVE equation. The differences between the "revised" IMPROVE equation and the
"modified" IMPROVE equation should be discussed in more detail. The document states:
"This algorithm was modified with the goal of reducing bias that had been observed in
applications of the original IMPROVE equation by splitting major PM components
between small and large size modes in recognition that atmospheric PM generally follows
a bimodal size distribution (Pitchford et al., 2007). This approach has been referred to as
the revised IMPROVE equation (U.S. EPA, 2019) or the split component algorithm
(Prenni et al., 2019). However, by the time of publication of the 2019 PM ISA, new
studies had concluded that the modified IMPROVE equation had not been generally
successful in decreasing the bias in atmospheric extinction estimates associated with the
original equation (U.S. EPA, 2019)."
Based on this statement, the document appears to imply that the "revised" IMPROVE equation is the
same as the "modified" IMPROVE equation, which is not correct.
The "original" IMPROVE algorithm (Equation 13-6) and the "modified" original IMPROVE algorithm
(Equation 13-7) are presented in the 2019 PM ISA on pages 13-11 and 13-12, respectively. These
equations tend to underestimate the highest light scattering values and overestimate the lowest values at
IMPROVE monitors throughout the U.S. To resolve these biases, a "revised" IMPROVE equation was
developed (Pitchford et al., 2007) that divides PM components into small and large particle sizes with
separate mass scattering efficiencies and hygroscopic growth functions for each size. The "revised"
IMPROVE equation both reduced bias at the lowest and highest scattering values and improved the
accuracy of the reconstructed bext. However, the "revised" IMPROVE equation is not presented in the
2019 PM ISA or the draft PM ISA Supplement. Here is the "revised" IMPROVE equation:
bext = 2.2 x fs(RH) x [Small Sulfate] + 4.8 x fL(RH) x [Large Sulfate]
+ 2.4 x fs(RH) x [Small Nitrate] + 5.1 x fL(RH) x [Large Nitrate]
+ 2.8 x [Small Organic Mass] + 6.1 x [Large Organic Mass]
+ 10 x [Elemental Carbon]
+ 1 x [Fine Soil]
+ 1.7 x fss(RH) x [Sea Salt]
+ 0.6 x [Coarse Mass]
+ Rayleigh Scattering (site specific)
+ 0.33 x [N02 (ppb)]
This equation should be added to the document and discussed in detail. This equation has been the
preferred IMPROVE equation for the past 15 years and is being used by all states in their most recent
A-14
-------
Regional Haze State Implementation Plans (for the second implementation period). In addition, use of
this equation is recommended by EPA in their "Modeling Guidance for Demonstrating Air Quality
Goals for Ozone, PM2.5, and Regional Haze" (November 29, 2018), pages 146-148.
In addition, the Leventhol and Kumar IMPROVE equation (2016) should be added to the document and
discussed in detail. Both the "revised" IMPROVE equation and the Leventhol and Kumar IMPROVE
equation are utilized in the draft PM PA document.
Page 4-8 states "The large and rapid change in mass scattering efficiencies during atmospheric aging
presents a challenge for accurately estimating atmospheric light extinction based on constant mass
scattering coefficients, as in the IMPROVE equation." However, it should be noted that the "revised"
IMPROVE equation and the Leventhol and Kumar IMPROVE equation divides sulfate, nitrate, and
organic mass PM components into small and large particle sizes with separate mass scattering
efficiencies for each size.
Page 4-8 discusses "mass scattering efficiencies for wildland fire smoke" but does not discuss the
composition of wildland fire smoke. The document should add a breakdown of typical wildland fire
smoke by PM components (e.g., organic carbon, elemental carbon, sulfate, nitrate, etc.).
A-15
-------
Dr. Judith C. Chow
Section 4 characterizes the recent welfare effects evidence that falls within the scope of the draft PM
Supplement.
a. Please comment on the identification, evaluation, and characterization of the available scientific
evidence in Section 4.
b. Please comment on whether the summary section in Section 4 appropriately characterizes recent
evidence in the context of the conclusions of the 2019 PMISA.
c. Please comment on whether there are any topics or studies that fall within the scope of the draft PM
Supplement that should be added or receive additional discussion in Section 4 or any topics for which
discussion should be shortened or removedfrom Section 4.
A better integration may be needed to connect Section 2.3 on "Welfare Effects" and Section 4 on
"Evaluation of Recent Welfare Effects Evidence". Both sections emphasize visibility impairment, but
not climate change, material damage, or ecosystem degradation.
Section 2.3.1 on "Visibility Impairment" acknowledged the changes in PM2.5 composition by region and
season that have affected the apportionment of light extinction among PM2.5 species. It highlights the
steep decline in sulfate of-4.6-6.1% per year in both rural and urban areas from 2002-2012 (U.S.EPA,
2019). However, this record is nearly a decade old, not representative of current status. Section 4.1 on
"Summary of Evidence for Visibility Effects from 2019 Integrated Science Assessment for Particulate
Matter" further emphasizes the importance of ammonium sulfate in particle light scattering. However,
neither Section 2.3.1 nor Section 4 adequately provide overall perspectives on relationships between
PM2.5 composition and light extinction as well as their seasonal and annual changes over the last decade.
It also contains incorrect statements. For example, Section 2.3.1 states that"... although PM2.5 sulfate is
still responsible for more light extinction than any other single species..(Lines 21-22, page 2-38) and
Section 4.1 emphasizes "... ammonium sulfate has historically accounted for a larger fraction of PM2.5
mass than other components..(Lines 24-25, page 4-1). These statements are not entirely true, they are
too generalized as they don't apply to the western U.S. where organic mass accounts for a large fraction
of PM2.5 mass as illustrated in Figures 13-4 and 13-5 (page 13-25 and 13-26) of the ISA (U.S.EPA,
2019). Overall, more detailed analysis of recent PM2.5 speciation data and their association with light
extinction are needed.
Section 4.2.1 on "Visibility Preference and Light Extinction" is based on Malm et al. (2019) that
summarizes relationships between public acceptability and atmospheric light extinction in four U.S.
(Washington, D.C., Denver, CO, Phoenix, AZ, and Grand Canyon, AZ) and two Canada (Chilliwack,
B.C., and Abbotsford, BC) locations. Although the acceptability levels among the studies are more
consistent when plotted against apparent contrast of distant features that are most sensitive to haze (e.g.,
Malm et al., 2019), it appears that public perception of visibility impairment or the importance of scenic
views is qualitative and judgmental with large uncertainties (Smith, 2013). A science-based visibility
standard warrants additional research using objective scenarios that quantify visibility improvements
A-16
-------
(e.g., haziness index or deciview) by relating changes in light extinction to human perceived variations
or perceptible changes.
With respect to recent advances in visibility monitoring and assessment, Section 4.2.2 on "Visibility
Monitoring and Assessment" briefly summarizes average atmospheric extinction reduction in the U.S. of
1.8% per year for the period of 1990-2018 and 2.8% per year for the period of 2002-2018. This is based
on reconstruction of total light extinction estimated from IMPROVE network speciated PM
concentrations by Hand et al. (2020). Although seasonal and spatial patterns of visibility impairment are
documented in the 2019 ISA (Chapter 13.2.4, U.S. EPA 2019), the analysis is based on older (2005-
2008 and 2011-2014) data that does not represent new information for the most recent period (e.g.,
2015-2018 or 2016-2019). Hand et al. (2019) find increasing organic mass (OM) to organic carbon (OC)
ratios across the IMPROVE network after 2011, highest during summer in the east and not necessarily
influenced by particle bound water. Information on spatial interpolation of average monthly
reconstructed light extinction coefficient (bext, Mm"1) by major chemical component for each region for
the most recent period will provide some perspective on overall changes, especially on increases in OM
in the mountain west and southwest and increases in nitrate and ammonium in the central U.S., since the
last review (2011-2014).
Section 4 would benefit from additional weight-of-evidence analyses and methods to monitor secondary
PM NAAQS indicators of visibility (Pitchford, 2010). Based on linear regression, So et al. (2015)
estimated time-resolved hourly total light extinction using continuous hourly PM2.5, NO2, relative
humidity, and historical monthly averaged aerosol chemical composition at four monitoring sites in the
Lower Frazer Valley of British Columbia, Canada. The Pitchford (2010) approach may warrant a revisit.
Case studies can be conducted at selected sites to evaluate if the hybrid modeling approach can provide
relatively accurate and time-resolved light extinction estimate in regions with sparse visibility
monitoring, thereby extending the spatial coverage of the nationwide visibility monitoring network.
Section 4 calls for evaluation of recent welfare effect evidence, but only addresses the visibility effects
without discussing climate, ecosystem, and material effects. Brief statements are made on Section 2.3.2
on "Climate Effects" and Section 2.3.3 on "Materials Effects", but only the 2019 ISA is cited, without
providing any new insights. Recent reviews on environmental and health impacts of air pollution (e.g.,
Manisalidis et al, 2020) and the physical science basis of climate change (IPCC, 2021) provide
background on non-visibility effects. Recent evidence points to effects of microplastics in PM on
climate (e.g. Revell et al., 2021), dry and wet deposition (Brahney et al., 2020), and ecosystems (e.g.
Halle et al., 2020; Sobhani et al., 2022). These and other emerging PM components require further
consideration.
References
Brahney, J., Hallerud, M., Heim, E., Hahnenberger, M., Sukumaran, S., (2020). Plastic rain in protected
areas of the United States. Science, 368, 1257-1260. 10.1126/science.aaz5819.
Halle, L.L., Palmqvist, A., Kampmann, K., Khan, F.R., (2020). Ecotoxicology of micronized tire rubber:
Past, present and future considerations. Science of the Total Environment, 706,
10.1016/j.scitotenv.2019.135694.
A-17
-------
Hand, J.L., Gill, T.E., Schichtel, B.A., (2019). Urban and rural coarse aerosol mass across the United
States: Spatial and seasonal variability and long-term trends. Atmospheric Environment, 218,
10.1016/j.atmosenv.2019.117025.
Hand, J.L., Prenni, A.J., Schichtel, B.A., Malm, W.C., Chow, J.C., (2019). Trends in remote PM2.5
residual mass across the United States: Implications for aerosol mass reconstruction in the IMPROVE
network. Atmospheric Environment, 203, 141-152. 10.1016/j.atmosenv.2019.01.049.
Hand, J.L., Prenni, A.J., Copeland, S., Schichtel, B.A., Malm, W.C., (2020). Thirty years of the Clean
Air Act Amendments: Impacts on haze in remote regions of the United States (1990-2018). Atmospheric
Environment, 243, 10.1016/j.atmosenv.2020.117865.
IPCC, 2021: Summary for Policymakers. In: Climate Change 2021: The Physical Science Basis.
Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on
Climate Change [Masson- Delmotte, V., P. Zhai, A. Pirani, S.L., Connors, C. Pean, S. Berger, N. Caud,
Y. Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthewss, T.K. May cock,
T. Waterfield, O. Yelekci, R. Yu and B. Zhou (eds.)]. Cambridge University Press. In Press.
Malm, WC; Schichtel, B; Molenar, J; Prenni, A; Peters, M. (2019). Which visibility indicators best
represent a 41 population's preference for a level of visual air quality? J Air Waste Manag Assoc, 69:
145-161.
Manisalidis, I., Stavropoulou, E., Stavropoulos, A., Bezirtzoglou, E. (2020). Environmental and Health
Impacts of Air Pollution: A Review. Frontiers in Public Health, 8(14).
Pitchford, M. (2010). Memorandum to PM NAAQS Review Docket (EPA-HQ-OAR-2007- 0492).
Assessment of the Use of Speciated PM2.5 Mass-Calculated Light Extinction as a Secondary PM
NAAQS Indicator of Visibility. November 17, 2010. Docket ID No. EPA-HQ-OAR-2007-0492. Office
of Air Quality Planning and Standards Research Triangle Park, NC.
Revell, L.E., Kuma, P., Le Ru, E.C., Somerville, W.R.C., Gaw, S., (2021). Direct radiative effects of
airborne microplastics. Nature, 598, 462-467. 10.1038/s41586-021-03864-x.
Smith, A.E., (2013). An evaluation of the robustness of the visual air quality "preference study" method.
Journal of the Air & Waste Management Association, 63, 405-417.
So, R., Vingarzan, R., Jones, K., Pitchford, M. (2015). Modeling of time-resolved light extinction and its
applications to visibility management in the Lower Fraser Valley of British Columbia, Canada. J Air
Waste Manag Assoc, 65:6, 707-720.
Sobhani, Z., Panneerselvan, L., Fang, C., Naidu, R., Megharaj, M., (2022). Chronic and
transgenerational effects of polyethylene microplastics at environmentally relevant concentrations in
earthworms. Environmental Technology and Innovation, 25, 10.1016/j.eti.2021.102226.
A-18
-------
U.S. EPA (U.S. Environmental Protection Agency). (2019). Integrated Science Assessment (ISA) for
particulate 16 matter (final report, Dec 2019) [EPA Report], (EPA/600/R-19/188). Washington, DC. 17
https://cfpub.epa.gOv/ncea/i sa/recordisplay.cfm?deid=347534.
A-19
-------
Dr. Jane Clougherty
Overall, I found the ISA supplement to be an impressive, thorough review of the relevant literature. It is
generally well-written, through very repetitive, and the structure (perhaps unavoidable) can be very
challenging to follow.
My preliminary comments focus on Section 3. (I have no specific comments on Sections 1 and 2, except
that they are very repetitive).
Perhaps also worth noting that publication bias is likely present, though probably unlikely to explain
observed concentration-response functions.
The heading for section 3.1.2.4 should refer to cardiovascular effects, not mortality.
Section 3.2: Mortality
3.2.1: Short-Term PM2.5 Exposure
I agree with the broad assessment that the bulk of the epidemiologic evidence supports a positive
association between PM2.5 and mortality (p. 3-56 ISA Table 11-1), though associations vary
substantially, as would be expected with between-study variation in methods, spatial differences in PM
composition and population susceptibility, etc. (geographic heterogeneity).
I'm a bit hesitant on the issue of co-pollutant adjustment, as many of the recent studies examining this
confounding are based in Europe (where local traffic emissions contain more PM2.5 due to high diesel
prevalence) and Asia (where PM levels are generally much higher). It is noted on p. 3-60 that only one
multi-city US study investigated co-pollutant confounding (Lavigne et al, 2018). My concern in that
many of the larger US studies have leaned on larger-scale regional models for PM2.5 at 1 km2 resolution
or larger, which is appropriate to the spatial scale of variation for PM2.5, but either not thoroughly
adjusted for NO2 or other local emissions indicators, or not done so at the much finer spatial scales at
which local sources vary.
Table 3-4 section on biological plausibility seems to lean on epidemiologic evidence, rather than
toxicology or mechanistic studies, to establish biologic plausibility. Likewise, on p. 3-75 (first para)
there is reference to epidemiologic studies for cardiovascular and respiratory morbidity lending
"biologic plausibility" to epidemiologic evidence for mortality, rather than simply corroboration.
There is an emphasis on multi-city studies, which makes sense from a policy standpoint, and some
epidemiologic considerations (larger sample size and generalizability), but raises issues in considering
effect modification and co-pollutant confounding, specifically:
(1) It is a slightly awkward definition of effect modification that includes between-city variation,
because inter-urban variation may be due to differences in exposures (sources, composition) or
characterization thereof (e.g., modeling error), rather than factors that may actually alter
exposure-health relationships (e.g., population characteristics).
A-20
-------
(2) Multi-city studies can suffer from imperfect capture of exposure variation within each city (if
using common model everywhere & not fully accounting for composition/ source differences),
or mis-specification of co-pollutant effects (if not captured with fine spatial resolution within
each city). Compromises are often made for consistency in exposure modeling across multiple
cities, which may lead to within-city mis-characterization of exposures or C-R relationships. This
is not to say multi-city studies are not valid, but single-city studies may, in many cases, provide a
cleaner base of comparison for effect modification analyses.
3.2.2: Long-Term PM2.5 Exposure
In the overall review of updated studies (p. 3-78), as elsewhere, it appears taken for granted that larger
sample size (and multi-city study) implies a higher-quality study, though there are often important trade-
offs between sample size and accuracy/ resolution in exposure assignment.
3.3.3. Populations and lifestages
p. 3-128: "these additional studies provide further evidence that lower SES communities are exposed to
higher concentrations of PM2.5 compared to higher SES communities." Please insert "on average," or
similar. This statement is certainly true on average, though there important nuances to this relationship
(eg, better transportation and therefore higher PM in some higher-SES communities) that lend to non-
linearities & settings where exposures are higher in wealthier central parts of some US and European
cities.
A good example of the issue raised above re: co-pollutant adjustment is demonstrated on p.3-94: "PM2.5
-mortality associations are null in models with ozone (HR = 1.00 [95% CI: 1.00, 1.01]) and oxidants
(HR = 1.00 [95% CI: 0.99, 1.01]), and attenuated in a model with N02 (HR = 1.01 [95% CI: 1.00,
1.02])." It is challenging to interpret this result in the absence of information on the scales of
measurement for each. This is important because most of the other multi-pollutant studies cited report
that adjustment does not measurably alter PM-mortality associations. What is different here? Is there
perhaps better measurement for NO2 simply by measuring it at a finer scale? If not, how should we
reconcile this result with others?
Excellent that disparities in both exposures and C-R relationships are addressed separately and clearly.
Consider removing Canadian studies from this section - different social & economic context, context of
health disparities very different, different patterns of historical discrimination by race and ethnic group,
universal access to healthcare and education alter interpretability of SES indicators for US regulatory
context.
Example where larger (geographic) studies may be harder to interpret, in that SES indices don't transfer
well across space/ social settings, or across urban-rural gradients. Different costs of living, material
assets (car ownership).
Consider removing references to greenspace, as a determinant of PM and exposures, differential
distributions by race and SES (also quality, access) - multiple pathways linking greenspace to health
simultaneously. Probably complicates more than adds.
A-21
-------
Dr. Deborah Cory-Slechta
Charge Question 4 - Section 3 characterizes the recent health effects evidence that falls within the scope
of the draft PM Supplement.
a. Please comment on the identification, evaluation, and characterization of the available
scientific evidence in Section 3.
In my view, the scientific evidence presented in Section 3 is amply identified, evaluated and
characterized. It is clearly stated as to the studies included in the Draft Supplement in terms of
those most informative with respect to potential revisions to the PM NAAQS and the extent to
which these findings compare to the scientific conclusions reached in the 2019 PM ISA. In
addition, studies were included that had the potential to mitigate potential confounds and further
inform the strength of the relationships/associations. In addition, study weaknesses are noted as
are inconsistent results.
b. Please comment on whether the summary sections in Section 3 appropriately characterize recent
evidence in the context of the conclusions of the 2019 PMISA.
The conclusions reached in the draft supplement are appropriately related to the 2019 PM ISA.
Specifically, evidence that is consistent with the 2019 PM ISA conclusions is noted, and evidence
that is inconsistent or not consistent is also noted. Summaries, particularly the figures that are
included, are appropriate as well. For this reviewer, the fact of the overwhelming number of positive
associations, whether technically significant or not, in spite of all of the differences between the
studies is particularly compelling. Further, differences in strength of the association be expected
given that contaminant profiles of the PM2.5 is going to differ by geography, climate, weather, etc.
c. Please comment on whether there are any topics or studies that fall within the scope of the draft
PM Supplement that should be added or receive additional discussion in Section 3 or any topics for
which discussion should be shortened or removedfrom Section 3.
The focus on cardiovascular and mortality endpoints is appropriate given the intent of the Supplement to
evaluate the data that has been reported since the 2019 ISA as these were both listed as causal,
relationships between other endpoints and PM2.5 are not yet as clear nor does new published data since
the cutoff for the 2019 ISA contribute sufficiently to change the conclusions with respect to those
endpoints. In my understanding, the draft PM supplement includes all of the relevant studies that have
been published. It might be useful for various reasons to include a summary table in the supplement
specifically detailing the US and Canadian studies being relied on and details of those studies. In
addition, the notes in the figures, e.g., updates of Figures from the 2019 PM ISA might be include
sample sizes/age, etc. rather than just the name of the cohort that the study was based on.
One topic that does come to mind, although not necessarily related to the current document or its
ultimate purpose and which may be included in the 2019 PM ISA is the fact that exposure to air
pollution is lifelong, beginning in utero. Obviously, this cannot be accommodated in terms of data or
A-22
-------
specific calculations but may be an important reminder with respect to the problem itself, given that
right now we're not even focused on lifetime exposures.
A-23
-------
Dr. Mark W. Frampton
Charge Question 1 - Executive Summary
a. Please comment on the clarity with which the Executive Summary communicates the key information
from the draft PM Supplement.
In general, the Executive Summary clearly communicates the key findings of the Supplement.
Specific comments:
Page ES-1, line 21. This sentence is somewhat confusing and unclear. "This Supplement to the 2019 PM
ISA is not intended to represent the full multidisciplinary evaluation of evidence that results in the
formation of weight-of-evidence conclusions, but instead is an assessment that puts the results of recent
studies in the context of the scientific conclusions (i.e., causality determinations) presented within the
2019 PM ISA." It seems that both parts of this sentence are true; the Supplement attempts to put recent
studies into context, and also to draw new weight of evidence conclusions in light of the added studies.
Suggest rewording this sentence.
b. Please provide recommendations on whether additional information should be added to the Executive
Summary or information that should be left for discussion in the subsequent sections of the draft PM
Supplement.
No suggestions.
Charge Question 2 - Introduction
a. Please comment on the clarity of the section, whether the scope is appropriate for the purpose of the
draft PM Supplement, and whether additional information is needed to convey the purpose of the draft
PM Supplement and the basis for the targeted evaluation conducted.
Page 1-3, Health Effects. This section needs rewording. It states that,
".. .for these health effect categories the recent studies evaluated are limited to: .. .Epidemiologic
studies that employed causal modeling methods or conducted accountability analyses...". This statement
is not true; these types of studies were included in the Supplement, but it was not "limited to.
Perhaps this bullet point should be moved to "Key Scientific Topics".
The Introduction states, "Therefore, within this Supplement the focus is only on the health effects
evidence where the 2019 PM ISA concluded a causal relationship. " This precludes consideration that
new evidence might move a determination from likely to causal for another health effect, such as
nervous system effects, respiratory effects, or cancer. It also precludes consideration that the causality
judgement may change in this review, even if the overall evidence base has not changed substantially. In
other words, opinions/judgements of the EPA authors, and of the current CASAC, may differ from those
A-24
-------
of the previous review, even with a similar evidence base, such that a health effect considered to be
"likely" in the previous review would be considered causal in this review. This approach does not allow
for that, although admittedly the possibility seems unlikely.
The Introduction should provide additional background on the reasons and rationale for the
reconsideration. It should also list the key CASAC findings in its review of the draft ISA, as detailed in
the CASAC letter to the Administrator of April 11, 2019, and the EPA responses. Specifically in this
regard, the CASAC letter states ".. .the CASAC finds that the Draft ISA does not present adequate
evidence to conclude that there is likely to be a causal association between long-term PM2.5 exposure
and nervous system effects; between long-term UFP exposure and nervous system effects; or between
long-term PM2.5 exposure and cancer." In the final ISA, EPA did change UFP and nervous system
effects to "suggestive", but did not accept CASAC's recommendations on causality determinations for
PM2.5 and nervous system effects and cancer. The justifications for those decisions should be stated.
The justification for limiting studies to those in the US and Canada should be stated in the Supplement.
It appears that considered studies were not always limited to US and Canada. See Figure 3-13, for
example.
"Cardiometabolic disease" should be defined as it is used in this document.
Charge Question 3 - To ensure that recent studies are put in the context of the conclusions of the 2019
PMISA the draft PM Supplement pulls in information verbatim from the 2019 PMISA to orient the
audience. Two ways this was done in the draft PM Supplement is through Section 2 which is the
Integrated Synthesis Chapter (i.e., Chapter 1) of the 2019 PMISA and leading off each health and
welfare effects discussion in Section 3 and 4 with the Summary and Causality Determination from the
2019 PMISA.
a. Please comment on this approach and whether any additional modifications to the structure of the
document can be made to better integrate evidence evaluated in the draft PM Supplement with
conclusions from the 2019 PMISA.
The current structure, providing the Integrated Synthesis Chapter from the 2019 ISA, and leading off
each effects section with the Summary and Causality Determination from the 2019 ISA, is effective and
helps integrate the new findings with those in 2019.
One minor structural change is suggested:
In Section 3.1, Cardiovascular Effects, the sub-sections start with summaries of the key findings in the
2019 ISA, as noted above, and this is helpful. However, Section 3.1.1.2, "Recent U.S. and Canadian
Epidemiologic Studies", begins not with new findings as expected, but with descriptions of biologically
plausible mechanisms described in the 2019 ISA. Then the next paragraph begins with the new studies.
This first paragraph is out of place; it could be eliminated, or worked into the summary of the 2019 ISA.
The same is true for sections 3.1.2.2, 3.2.1.2, and 3.2.2.2.
A-25
-------
Charge Question 4 - Section 3
a. Please comment on the identification, evaluation, and characterization of the available scientific
evidence in Section 3.
In general, Section 3 is clearly and concisely presented, and the conclusions are well-supported by the
evidence.
Specific comments:
Many of the figures are difficult to read, and need more vertical space. Fig 3-23 in particular does not
display well.
Page 3-28: "However, a recent toxicological study adds to similar evidence from the 2009 PM ISA.
The reference should be provided here, and it should be clarified whether this is a new study not covered
in the 2019 ISA. If this refers to Lippmann 2013, that study was reviewed in the 2019 ISA, and is not
really "recent".
Page 3-31, describing findings from Bai, et al. 2019: "A stronger association was observed in the highest
tertile of (>38.97 ppm) Ox concentrations (HR: 1.12 [95% CI: 1.09, 1.15])." Suggest clarifying what
association is being referred to here. If this is from Table 5 of the paper, the data are incorrect. The HR
for incident AMI and PM2.5, in the highest tertile of Ox, was 1.08 (1.06, 1.10).
Similarly, with regard to Section 3.1.2.2.4, page 3-37, Bai et al. Table 5 indicates the HR for incident
CHF in the highest tertile was 1.08 (1.07, 1.09), rather than the 1.12 given in the text of the Supplement.
Page 3-39, 2nd paragraph of Section 3.1.2.2.8: This paragraph needs revising; it is often unclear what the
various HRs refer to.
Page 3-57, subsection "Other nonaccidental mortality". This discussion of sudden death and the
Rappazzo et al. study is confusing. This section appears to deal specifically with out of hospital cardiac
arrest, so the section title should reflect that. Also, the OR CIs in the Rappazzo study include 1.0, so,
although the ORs are positive, they are statistically non-significant, and this should be so indicated.
Section 3.3.1, page 3-121. The Hemmingsen et al. human study was reviewed in the 2019 ISA, but only
in the context of cancer, with regard to the negative findings on DNA damage and blood mononuclear
cell gene expression. The positive findings on vascular and cardiac function, reported in a separate
publication, were not reviewed in 2019, even though the publication date was 2015. Perhaps that should
be mentioned in the Supplement for clarity.
b. Please comment on whether the summary sections in Section 3 appropriately characterize recent
evidence in the context of the conclusions of the 2019 PMISA.
The summary sections do appropriately characterize the new evidence, and are very helpful.
c. Please comment on whether there are any topics or studies that fall within the scope of the draft PM
Supplement that should be added or receive additional discussion in Section 3 or any topics for which
discussion should be shortened or removedfrom Section 3.
A-26
-------
No additional comments.
Charge Question 6 - Summary and Conclusions section (Section 5)
a. Please comment on the level of detail provided within this section and whether revisions should be
made to further summarize recent evidence.
This section provides a concise and accurate summary of the findings of the Supplement, with an
appropriate level of detail.
Specific comments:
Page 5-2, first paragraph under Cardiovascular Effects, Short Term PM2.5 Exposure: "In addition, these
studies report evidence that continues to indicate an immediate effect of PM2.5 on cardiovascular-
related outcomes primarily within the first few days after exposure..An "immediate" effect
somewhat contradicts the latter part of the sentence, "within the first few days". Immediate suggests
minutes or at most hours. Section 2.2.2.2 defines the lags as follows: ".. .immediate (e.g., lag 0-1 days),
delayed (e.g., lag 2-5 days), or prolonged (e.g., lag 0-5 days)...". Suggest specifying that the evidence
predominantly supports immediate or slightly delayed effects.
Page 5-3, Mortality, Long Term PM2.5 Exposure. The following sentence seems contradictory, and
needs further clarification; underlining added. "The assessment of the C-R relationship continues to
generally support a linear, no-threshold relationship with certainty down to 4 ug/m3. However, some
uncertainties remain about the shape of the C-R curve at relatively low PM2.5 concentrations (<8
Ug/m3)...".
A-27
-------
Dr. Christina H. Fuller
Comments on Section 3.3 Key Scientific Topics
a. Please comment on the identification, evaluation, and characterization of the available
scientific evidence in Section 3.
b. Please comment on whether the summary sections in Section 3 appropriately characterize recent
evidence in the context of the conclusions of the 2019 PMISA.
c. Please comment on whether there are any topics or studies that fall within the scope of the draft PM
Supplement that should be added or receive additional discussion in Section 3 or any topics for which
discussion should be shortened or removedfrom Section 3.
The Key Scientific Topics covered within section 3.3 include the following: (a) recent experimental
studies conducted at near-ambient concentrations; (b) effects at the ambient concentrations reported in
epidemiologic studies (Section 3.3.1); (c) the role of PM2.5 exposure on COVID-19 infection and death
(Section 3.3.2) and (d) an evaluation of studies that examine PM2.5 exposure and health risk disparities
among racial and ethnic groups and socioeconomic status (SES) (Section 3.3.3).
Recommendations:
The summaries and conclusions within this document use the term White and non-White as the broadest
categories. I recommend the Supplement refer to the group non-White as People of Color (POC) or
Communities of Color (COC), as appropriate. There are multiple terms utilized to describe the span of
races and ethnicities in the United States, which is reflected by the studies included in this Supplement.
Race/ethnicity is a fluid concept that is relevant by time, country, region, population and government.
Therefore, the most useful terminology for the purpose of protecting public health has changed over
time.
There is a lack of consistency in the manner in which statistical significance is noted in this section.
Therefore, I recommend that the same level of detail be provided for each study. Include the statistical
significance of the findings for studies included in this section so that readers can best interpret the
results and implications on the conclusions presented therein. Statements of statistical significance
should utilize the effect estimate and 95% confidence intervals.
I recommend that exposure measurement be described for each study included in the 2019 Draft ISA
Supplement. A key point in critically evaluating each study is knowing the details of exposure
measurement. Validity and interpretation of results is linked closely to the methods and accuracy of
PM2.5 measurement and exposure assessment.
I recommend that studies which evaluated autocorrelation between PM2.5 and race/ethnicity (or SES) be
noted in the Supplement. The most robust epidemiologic studies examine autocorrelation and employ
statistical adjustments when necessary.
A-28
-------
The level of SES data (e.g. individual vs community) is an important component in evaluating research
questions, models and results. I recommend that Section 3.3.3 (Populations and Lifestages at Potentially
Increased Risk of a PM-Related Health Effect) include greater discussion related to this.
Specific comments regarding recommended changes:
Section 3.3.1 Recent Experimental Studies at Near-Ambient Concentrations:
Page 3-121, Lines 28 - 31: Align the terms "nonfiltered" and "unfiltered" air, by selecting only one term
for these sentences.
Page 3-122, Line 22: State the directions of effect for men and women here, in addition to writing that
they are in opposite directions.
Section 3.3.2 PM2.5 Exposure and COVID-19 Infection and Death:
This section could use expansion, especially since COVID-19 and PM2.5 disproportionately impact
populations of color (POC).
There is some evidence that supports a relationship between PM2.5 exposure and COVID-19 incidence,
severity and mortality. Although mentioned in this section, the language on PM2.5 increasing
susceptibility should be stressed further. This relationship may partially explain the disparate exposures
of lower-income and people of color communities (covered in 3.3.3.1.2 and 3.3.3.2.1) who have been hit
hardest by COVID in terms of cases, severity and mortality in that population.
3.3.1.2 Long-term PM2.5 Exposure:
Page 3-124, Line 11: Specify the exact range of the study. From X date to June 18, 2020.
Page 3-124, Line 22 - 24: There is an incongruence between the PM2.5 measurements and outcomes
assessment. Include the authors' explanation for the acceptability of these choices.
Page 3-125, Line 2: Include the start date for the study.
3.3.3 Populations and Lifestages at Potentially Increased Risk of a PM-Related Health Effect
The level of SES data is an important component in understanding research questions, modeling and
interpretation of results. Individual-level and community-level data differ with regards to their accuracy
and interpretations. I recommend dividing the discussion of the studies according to individual-level and
community-level measures of SES and race/ethnicity. Pay particular attention to the comparison of
studies that use different community-level indicators such as census block-group, census tract, zip code
and county.
3.3.3.1 Socioeconomic Status
Page 3-127, line 24: Add the term "populations" before "having".
A-29
-------
Page 3-127, line 27: It is more accurate to describe educational attainment as an "indicator" or
"measure" of SES compared to a "proxy". SES is complicated construct that cannot me estimated by
any single item.
3.3.3.1.1 Exposure Disparity
Page 3-128, lines 8-19: As an example of the added detail on PM2.5 exposure assessment, the source of
the PM2.5 estimation is necessary to evaluate this study by Lee (2019) and compare to that of Rosofsky
et al (2018). Although this information is contained in the linked articles a brief mention here would be
useful.
Page 3-128, lines 26-27: I assume the groups here are mutually exclusive, however, it is not clear from
the description. Please clarify.
Page 3-129, line 8: Similar to the definition/clarification of composite PM provided in Chapter 2,1
suggest including a definition of composite here as it pertains to measures of SES.
Page 3-129, line 37: The study by Weaver et al (2019) is in need of a summary sentence to bring
together an interpretation of the findings. Although the Clusters are defined it is difficult to identify a
pattern (or not) in the findings. Clarify in one or a few sentences.
Figure 3-30: The figure is full of information and overall is a good presentation. However, it could be
improved by including horizontal lines separating the measures of SES and increasing the font size. I
would change the heading of the third column to Comparison groups instead of Reference Group and list
the referent group first.
3.3.3.1.2 Health Risk Disparity
Here again it is advisable to separate the studies further based on individual or community level
indicators of SES.
Long-term PM2.5 Exposure
Page 3-131, Line 30: Was income also measured at the zip code level? Clarify this point.
Page 3-132, Lines 1-4: Provide detail on the study type and spatial resolution of data for Zhang et al
2021. Since this is a Canadian study you can describe based on its U.S> equivalent, such as zip code,
census block-group, etc.
Page 3-133, Line 3: Add cardio metabolic mortality and cardiovascular mortality here, because those
endpoints show a similar differential between low greenspace, low deprivation and low greenspace, high
deprivation groups.
Page 3-137: After the final paragraph add a summary sentence or two about the findings from Weaver et
al (2019).
A-30
-------
3.3.3.2 Race/Ethni city
Page 3-138, line 4: "The 2009 PM ISA observed little evidence for increased PIVh.s-related risk by race
and some evidence of increased risk by Hispanic ethnicity." Is it the conclusion of the 2009 PM ISA that
this was due to a lack of studies examining these associations or that there was sufficient research to
draw this conclusion? Given that there has been exponential growth in studies examining race/ethnicity
and air pollution in recent years answering this question has implications for the current review of
research.
Long-Term PM2.5 Exposure
Page 3-146, line 8: The statement that, "Since moving was essentially random" is not accurate given the
research on redlining and segregation. Although these terms were utilized by the study authors, I do not
find that they are appropriate here. I suggest replacing the existing sentence with this one: Utilizing data
from the subpopulation of Medicare enrollees that moved, the authors were able to examine changes in
PM2.5 exposure.
Suggestions:
Consider removing Canadian studies from this section because their findings may not be generalized to
the U.S. Canada differs from the U.S. with regards to its social & economic context; health disparities;
historical discrimination by race and ethnic group; and universal access to healthcare and education.
These differences alter the interpretability of SES indicators for US regulatory context.
The articles by Yitshak-Shade et al, Son et al (2020), Crouse et al (2019), Stieb et al (2020) all include
the impact of green space on associations between PM2.5 and mortality. However, there is no discussion
of green space in the Supplement. The inclusion of these articles require added information regarding
the linkages between green space and PM2.5 concentrations; green space and health; and quality of green
space by race/ethnicity.
Charge Question 6
Comments on Section 5 (Summary/Conclusions)
Overall this section is well written and provides an accurate and succinct summary of Sections 1-4 of the
Draft Supplement to the 2019 PM ISA.
Recommendations:
The summaries and conclusions within this document use the term White and non-White as the broadest
categories. I recommend the Supplement refer to the group non-White as People of Color (POC) or
Communities of Color (COC), as appropriate. There are multiple terms utilized to describe the span of
races and ethnicities in the United States, which is reflected by the studies included in this Supplement.
Race/ethnicity is a fluid concept that is relevant by time, country, region, population and government.
A-31
-------
Therefore, the most useful terminology for the purpose of protecting public health has changed over
time.
There are some areas where conclusions need to be added, which I have noted in the specific comments
below.
Specific comments regarding recommended changes:
Line 26-27: I recommend altering, "specifically Black individuals, and low SES individuals" to
"specifically Black and low-SES populations". The use of the term populations emphasizes that the
identified vulnerabilities are linked to population health and also that the evidence shown refers to
population studies.
Cardiovascular Effects -> Long-term PM2.5 Exposure:
Line 25:"more diverse populations" is a term not specific enough for its use here. In most settings in the
United State "diverse populations" implies diversity in race/ethnicity, which is not the intended meaning
here. Replace this expression with "general population" or "population without preexisting conditions".
Additional Considerations Regarding the Health Effects of PM2.5 COVID-19 Infection and Death:
Unlike most other summary paragraphs in this section, the subsection focused on COVID-19 does not
have a conclusion. Add a statement(s) that draw a conclusion from the data summarized.
Populations and Lifestages at Potentially Increased Risk of a PM-Related Health Effect
Socioeconomic Status:
This summary paragraphs needs to distinguish between individual-level and community-level indicators
of SES. (Please see my comments to Section 3 for a more broad discussion.) Present the differences in
findings from each of these types of indicators and the implications on their interpretation and
comparison.
Visibility Effects:
Similar to the subsection on COVID-19, this paragraph does not have a conclusion. The paragraph
discusses methods and methodological improvements utilized in the included studies. State what the
compiled evidence reveals to be the effect of PM2.5 on contrast and light extinction and the level of
confidence of this conclusion.
A-32
-------
Dr. Terry Gordon
Charge Question 2 - Section 1 consists of an introduction detailing why the draft PM Supplement is
being developed along with the rationale and scope for the topics and studies considered.
a. Please comment on the clarity of the section, whether the scope is appropriate for the purpose of the
draft PM Supplement, and whether additional information is needed to convey the purpose of the draft
PM Supplement and the basis for the targeted evaluation conducted.
The scope of the introduction for the Supplement were very well written and justified the purpose and
need for the Supplemental PM ISA. In particular, the focus on the health effects associations that were
'causal' versus 'likely causal' was very appropriate.
Minor Comment: In the Executive Summary, the bulleted list is a bit unclear. The order made this
confusing - a bullet on short-term morbidity, then long-term mortality, and then a separate para going
back to short-term mortality could be rearranged.
Line 4, page 2-2 - PM3 must be a typo.
Charge Question 3 - To ensure that recent studies are put in the context of the conclusions of the 2019
PM ISA the draft PM Supplement pulls in information verbatim from the 2019 PM ISA to orient the
audience. Two ways this was done in the draft PM Supplement is through Section 2 which is the
Integrated Synthesis Chapter (i.e., Chapter 1) of the 2019 PMISA and leading off each health and
welfare effects discussion in Section 3 and 4 with the Summary and Causality Determination from the
2019 PMISA.
a. Please comment on this approach and whether any additional modifications to the structure
of the document can be made to better integrate evidence evaluated in the draft PM
Supplement with conclusions from the 2019 PMISA.
I believe the approach to include the Integrated Synthesis Chapter of the 2019 PM ISA was very
appropriate for starting off this Supplemental PM ISA. The approach was warranted and the pathway for
reviewing the supplement was very clear. Starting with the Summary and Causality Determination from
the 2019 PM ISA was warranted in terms of clarity and efficiency.
Charge Question 4 - Section 3 characterizes the recent health effects evidence that falls within the scope
of the draft PM Supplement.
Section 3.1 Cardiovascular Disease
a. Please comment on the identification, evaluation, and characterization of the available
scientific evidence in Section 3.1.
A-33
-------
The appropriate summaries and tables are presented in an efficient manner. I agree with the evaluation
and characterization of the short-term and long-term evidence linking CVD to PM2.5 exposures. The
Supplemental ISA clearly bolsters the PM ISA evidence that the association of CVD disease (and
mortality) is key to the evaluation of the protection of the current PM NAAQS.
b. Please comment on whether the summary sections in Section 3.1 appropriately characterize
recent evidence in the context of the conclusions of the 2019 PMISA.
The summary section is well written and identifies and summarizes the recent evidence that supplements
the conclusions of the 2019 PM ISA. The Relative Risk figures are important summaries of the studies
(and the color coding made the inter-ISA Review comparisons much easier/efficient).
c. Please comment on whether there are any topics or studies that fall within the scope of the
draft PM Supplement that should be added or receive additional discussion in Section 3.1 or
any topics for which discussion should be shortened or removedfrom Section 3.1.
In general, the scope covered by Section 3.1 was very well written. For example, the Summary for the
association between short-term PM2.5 exposure and cardiovascular effects was concise and on target. As
far as potential topics for shortening/removal, the somewhat long description of the weak or 'null'
association between short-term PM2.5 and stroke was surprising in its detail and could perhaps be
shortened. Conversely, the description of recent studies on aggregated cardiovascular endpoints lacks a
conclusion and could thus be expanded. Similarly, the section of long-term PM2.5 and hypertension
describes a moderate amount of evidence for altered blood pressure in post-menopausal women, yet
there is no conclusion statement(s) for that section and the potential susceptibility for this sub-population
is rarely mentioned elsewhere.
As far as additional topics or studies in regards to the short-term effects of PM2.5, the study by LC Chen
(Lippmann, 2013 HEI report which was reviewed in the 2019 PM ISA) demonstrated that relatively
short-term changes in ambient particle sources (e.g., 72 hours or less) could affect cardiovascular
endpoints in mice exposed to concentrated ambient PM.
Section 3.2 Mortality
a. Please comment on the identification, evaluation, and characterization of the available
scientific evidence in Section 3.2.
The mortality section was very well written and clearly evaluated the causal evidence for the association
between PM exposure and mortality (cardiovascular-related and respiratory-related). In particular, for a
non-epidemiologist, the explanation and evaluation of the confounding concerns and uncertainty factors
were clearly written. Table 2-2 and 3-1 were particularly useful.
b. Please comment on whether the summary sections in Section 3.2 appropriately characterize
recent evidence in the context of the conclusions of the 2019 PMISA.
The summary section is well written and identifies and summarizes the recent evidence that supplements
the conclusions of the 2019 PM ISA. The one area that could be clarified is the susceptible population
discussions - while the increased exposure concentrations vs. inherent susceptibility responses was
initially explained, attribution to one or the other reasoning was less clear in some subsections.
A-34
-------
c. Please comment on whether there are any topics or studies that fall within the scope of the
draft PM Supplement that should be added or receive additional discussion in Section 3.2 or
any topics for which discussion should be shortened or removedfrom Section 3.2.
No suggestions for additional topics or studies.
Minor comments
Table 3-2 - should the final column on PM2.5 concentrations have more data for the different sections?
Line 7, page 3-37 - Is a word missing (increase?).
The title for 3.1.2.4 states 'mortality' but should be cardiovascular effects/disease.
Defining descriptors for the C-R curve at low concentrations is needed. Linear is obvious but not so for
some of the others.
Section 3.3 Key Scientific Topics
a. Please comment on the identification, evaluation, and characterization of the available
scientific evidence in Section 3.3.
As mentioned above, the disparities discussion needs to be tightened in respect to exposure differences
or innate susceptibility differences.
The near ambient clinical studies provide strong evidence for short-term effects, but this evidence didn't
seem to carry over to the PA.
I would suggest shortening the covid section -it's worth discussing but does the evidence for
associations with ambient PM2.5 warrant the several pages of discussion?
Minor Comments:
Figure 3-36 is unclear. Perhaps the figure legend can be expanded. Also, its title is confusing: PM2.5
cause by?
b. Please comment on whether the summary sections in Section 3.3 appropriately characterize
recent evidence in the context of the conclusions of the 2019 PMISA.
The summary sections are clear and do an excellent job in characterizing the conclusions of the recent
evidence.
c. Please comment on whether there are any topics or studies that fall within the scope of the
draft PM Supplement that should be added or receive additional discussion in Section 3.3 or
any topics for which discussion should be shortened or removedfrom Section 3.3.
A-35
-------
The covid section is meant to stand on its own but given the immune altering effects of PM2.5 exposure,
it is puzzling why this supporting immune-PM2.5 interactions is not included.
A-36
-------
Dr. Michael T. Kleinman
Charge Question 1 - The Executive Summary is intended to provide a concise synopsis of the key
findings and conclusions of the draft PM Supplement for a broad range of audiences.
a. Please comment on the clarity with which the Executive Summary communicates the key
information from the draft PM Supplement.
• Clearly sets up the purpose of the supplemental analyses with respect to the extension
of the 2019 document.
• Provides an excellent summary of the 2019 findings and provides a summary of
additional supporting information from recent studies to bolster the finding of causal
effects of PM2.5 for both cardiovascular effects and mortality.
• Provides a brief summary of the additional support for the finding that there are
PM2.5 exposure and health disparities by race, ethnicity and social economic status.
b. Please provide recommendations on whether additional information should be added to the
Executive Summary or information that should be left for discussion in the subsequent
sections of the draft PM Supplement.
• Perhaps add a final statement the analyses and findings from the 32019 ISA were,
with few exceptions, strengthened by the addition of the analyses of the recently
published studies.
• The Box table on 2-1 (Overall conclusions of the 2019 ISA) could be copied to the
end of the Exec Summary with the addition of a column showing the corresponding
findings from the recent studies.
Charge Question 2 - Section 1 consists of an introduction detailing why the draft PM Supplement is
being developed along with the rationale and scope for the topics and studies considered.
a. Please comment on the clarity of the section, whether the scope is appropriate for the
purpose of the draft PM Supplement, and whether additional information is needed to convey
the purpose of the draft PM Supplement and the basis for the targeted evaluation conducted.
• Effectively sets up the rationale, scope and organization of the supplement.
A-37
-------
Charge Question 3 - To ensure that recent studies are put in the context of the conclusions of the 2019
PMISA the draft PM Supplement pulls in information verbatim from the 2019 PMISA to orient the
audience. Two ways this was done in the draft PM Supplement is through Section 2 which is the
Integrated Synthesis Chapter (i.e., Chapter 1) of the 2019 PMISA and leading off each health and
welfare effects discussion in Section 3 and 4 with the Summary and Causality Determination from the
2019 PMISA.
a. Please comment on this approach and whether any additional modifications to the structure
of the document can be made to better integrate evidence evaluated in the draft PM
Supplement with conclusions from the 2019 PMISA.
• There seems to be a fair amount of redundancy between Sections 2 and 3.
• Could Section 2 be shortened?
Charge Question 4 - Section 3 characterizes the recent health effects evidence that falls within the scope
of the draft PM Supplement.
a. Please comment on the identification, evaluation, and characterization of the available
scientific evidence in Section 3.
• It would make it easier for readers if there was more parallelism between sections 2
and 3.
• Putting the findings of section 3 into a table similar to the box on 2-1 would be
difficult but could highlight the important findings in this section.
b. Please comment on whether the summary sections in Section 3 appropriately characterize
recent evidence in the context of the conclusions of the 2019 PMISA.
• The review of the recent literature is quite comprehensive and the approach to
analyzing and integrating the information is appropriate.
c. Please comment on whether there are any topics or studies that fall within the scope of the
draft PM Supplement that should be added or receive additional discussion in Section 3 or
any topics for which discussion should be shortened or removedfrom Section 3.
• There could be more emphasis on sex as a biological variable especially with respect
to cardiovascular effects of PM.
Charge Question 5 - Section 4 characterizes the recent welfare effects evidence that falls within the
scope of the draft PM Supplement.
a. Please comment on the identification, evaluation, and characterization of the available
scientific evidence in Section 4.
A-38
-------
• Appropriate
b. Please comment on whether the summary section in Section 4 appropriately characterizes
recent evidence in the context of the conclusions of the 2019 PMISA.
• Information is suitably presented
c. Please comment on whether there are any topics or studies that fall within the scope of the
draft PM Supplement that should be added or receive additional discussion in Section 4 or
any topics for which discussion should be shortened or removedfrom Section 4.
• Climate effects are important factors in the increased numbers and severity of wild
fires and a more thorough discussion could be helpful. Localized drought conditions
that, for example, create reduced humidity conditions in some areas and increased
humidity conditions in others can imbalance the relationship of PM to visibility.
Charge Question 6 - The Summary and Conclusions section (Section 5) provides an overview of the
evidence evaluated in the draft PM Supplement.
a. Please comment on the level of detail provided within this section and whether revisions
should be made to further summarize recent evidence.
• It should be made clear that the focus of this supplement was directed at outcomes
that were identified as causally related to PM and that new respiratory and
neurotoxicity studies were not included.
• Future analyses should re-examine the causal relationships between PM exposures
and respiratory disease, degenerative nervous system effects, birth outcomes and
other outcomes.
A-39
-------
Dr. Stephanie Lovinsky-Desir
General Comments:
In Section 3.3.3.1 that describes disparities in exposure and risk based on socioeconomic status (SES),
there are several studies that were included that used composite indexes to identify the "at risk"
populations. However, many of those indexes are not limited to SES factors, and also include race
and/or ethnicity in their index (e.g. Canadian Marginal Index and Social Vulnerability Index). It does not
seem appropriate to include these studies in the section dedicated to SES since race and ethnicity are not
indicators of SES and are more likely functioning as indicators of racial segregation in the context of
these indexes. I would suggest removing these studies from this section and including a separate section
for studies that evaluate SES, race, and ethnicity together to emphasize the distinction that race, and
ethnicity are not proxies for SES, but rather they are highly correlated with SES because of systemic
racism. The Jorgenson 2020 study mentioned on page 3-147 would be another study to include is a
section dedicated to the combined effects of race and SES.
The supplement may benefit from an update on the literature regarding PM2.5 exposure in children who
are a vulnerable population. Examples of studies that may be considered for inclusion in the supplement
include the following:
1. Strosnider HM, Chang HH, Darrow LA, Liu Y, Vaidyanathan A, Strickland MJ. Age-specific
associations of ozone and fine particulate matter with respiratory emergency department visits in
the United States. Am JRespir Crit Care Med 2019;199:882-890.
2. Cserbik D, Chen Jiu-Chiuan, McConnell Rob, Berhane K, Sowell ER, Schwartz J, Hackman DA,
Kan E, Fan CC, Herting MM. Fine particulate matter exposure during childhood relates to
hemispheric-specific differences in brain structure. Environment International 2020; 143:
1059332.
3. Kim KN, Kim S, Lim YH, Song IG, Hong YC. Effects of short-term fine particulate matter
exposure on acute respiratory infection in children. International Journal of Hygiene and
Environmental Health 2020; 229: 113571.
4. Miao Liu, Wenting Guo, Yunyao Cai, Huihua Yang, Wenze Li, Liangle Yang, Xuefeng Lai, Qin
Fang, Lin Ma, Rui Zhu, Xiaomin Zhang. Personal exposure to fine particulate matter and renal
function in children: A panel study. Environmental Pollution 2020; 266(2): 115129.
5. Shan Liu, Qingyu Huang, Yan Wu, Yi Song, Wei Dong, Mengtian Chu, Di Yang, Xi Zhang, Jie
Zhang, Chen Chen, Bin Zhao, Heqing Shen, Xinbiao Guo, Furong Deng, Metabolic linkages
between indoor negative air ions, particulate matter and cardiorespiratory function: A
randomized, double-blind crossover study among children. Environment International 2020;
138:105663.
Specific Comments:
Page 3-121, lines 27-31: it would be helpful to briefly contextualize why a reduction in heart rate
variability is clinically meaningful.
A-40
-------
Page 3-127, lines 1-3: There is a reference to the Preamble to the ISA that notes a special emphasis
placed on studies that compare responses to a 'reference population'. The term 'reference population' is
often applied to a White population and thus using this terminology has the potential to perpetuate bias.
For the purposes of this ISA supplement, it may be more appropriate to rephrase this sentence to state
"studies that compare between different populations."
Page 3-144, line 18: 'regard less' should be change to regardless.
Page 3-145, line 3: Please clarify that this sentence is referring to IRD groups since there was a
difference observed across RRS groups and the figure that is referenced at the end of the sentence
includes both IRD and RRS metrics.
Page 3-147, lines 5-11: The brief mention of the Wang 2020 and Son 2020 studies do not include a
mention of PM2.5.
Page 3-149, line 8: the Wang 2020 study is mentioned again without being placed in the context of
PM2.5 exposure.
Minor Comments:
Overall, I find that the figure titles/legends are not very informative, often making them difficult to
interpret. It would be helpful if there were more details included in the form of a figure legend to
emphasize the key findings that are being highlighted in the figures.
A-41
-------
Dr. Jennifer Peel
General Comments about the draft ISA document:
• The document is a robust evaluation of the literature, with consideration of coherence from several
lines of evidence. The document includes a clearly defined scope and is a clear and transparent.
• Consider more explicitly justifying the scope for the supplement.
• Consider clarifying terms and description of populations at risk in the document, particularly when
using the term factor. For example, children is not really describing a factor; younger age (and even
better, specific age groups) would be clearer, in addition to including the comparison group (younger
ages compared to adults, or to older adults?). The same suggestion holds for race as a factor.
• In the Executive Summary section discussing the evidence for long term exposure and mortality, it
may be helpful to clearly explain how studies evaluating life expectancy differ from other studies
evaluating long term exposure and mortality.
• I suggest reducing the use of abbreviations as much as possible. Common abbreviations such as PM
and ED and CVD may be fine, but others may not be necessary nor helpful in such a document.
• The Figures used throughout the document are very helpful for delineating the evidence since the
2019 document and since the 2009 document.
• It would be helpful to repeat frequently that the unit increase for results presented are per 10 ug/m3
unless otherwise stated.
• This suggestion is likely outside of the defined scope and is a recommendation for future ISAs;
however, the interaction between extreme heat and PM could be considered in the sections
evaluating populations at increased risk (with the exception of the consideration of season)
(particularly for short term mortality)
Comments on sections outside of assigned charge questions (draft ISA document):
• Page 3-21: The section describing Zhang et al. 2018 and Wang et al. 2018 could be clarified,
including a clear description of what the analysis and results adds to the body of evidence.
• Page 3-23: It may be helpful here to clarify the assumptions of the IPW modeling, including how the
assumptions were evaluated and if those assumptions were met.
• The sections referencing mortality within Section 3.1 are appropriately brief and refer to more
details in the Section 3.2
• Section 3.1.1.2.6 (short term exposure, CVD mortality): This paragraph lacks references and may be
incomplete?
• Section 3.1.2.4 May have an incorrect title? Should be cardiovascular effects (not mortality)?
• Section 3.3.1 and Section 5, Experimental Studies at Near-Ambient PM2.5 Concentrations, (page 5-
3); consider including the following publications:
o Cole-Hunter T, Dhingra R, Fedak KM, Good N, L'Orange CL, Luckasen G, Mehaffy J,
Walker E, Wilson A, Balmes J, Brook RD, Clark ML, Devlin RB, Volckens J, Peel JL.
Short-term differences in cardiac function following controlled exposure to cookstove air
pollution: the subclincal tests on volunteers exposure to smoke (SToVES) study. Env Int.
2021
A-42
-------
o Walker ES, Fedak KM, Good N, Balmes J, Brook RD, Clark ML, Cole-Hunter T, Dinenno F,
Devlin R, L'Orange C, Luckasen G, Mehaffy J, Shelton R, Wilson A, Volckens J, Peel JL.
Acute Differences in Pulse Wave Velocity, Augmentation Index, and Central Pulse Pressure
Following Controlled Exposures to Cookstove Air Pollution in the Subclinical Tests of
Volunteers Exposed to Smoke (SToVES) Study. Environmental Research, 2020.
https://doi.Org/10.1016/i.envres.2019.108831
o Fedak KM, Good N, Walker ES, Balmes J, Brook RD, Clark ML, Cole-Hunter T, Devlin R,
L'Orange C, Luckasen G, Mehaffy J, Shelton R, Wilson A, Volckens J, Peel JL. Acute
changes in lung function following controlled exposure to cookstove air pollution in the
subclinical tests of volunteers exposed to smoke (STOVES) study. Inhalation Toxicology.
2020.
o Fedak KM. Good N, Walker ES, Balmes J, Brook RD, Clark ML, Cole-Hunter T, Devlin R,
L'Orange C, Luckasen G, Mehaffy J, Shelton R, Wilson A, Volckens J, Peel JL. Acute
effects on blood pressure following controlled exposure to cookstove air pollution in the
SToVES study. J Am Heart Assoc. 2019. 8:e012246.
https://doi.org/10.1161/JAHA.119.012246
Charge Question 4 Section 3.2 Mortality (Draft ISA supplement)
• Section 3.2, including sections on short term and long term exposure to PM2.5, are clear in scope and
include relevant studies in the identified time frame (January 2018 - March 2021).
• The evidence from the previous 2019 PM ISA is clearly articulated, and the more recent studies are
accurately described and added to the context of the evidence that was in the 2019 PM ISA.
• Page 3-57: Line 16 refers to Section 3.1.1.2.4; this seems to not be the correct section; should be
Section 3.2.1.2.4 (page 3-61).
• Page 3-61: The description of the Lavigne et al. 2018 study, including an evaluation of effect
modification by oxidant gases, could be clarified by adding information about the definition and
assessment of oxidant gases and perhaps some information about how to interpret this evidence in
the context of PM and health.
• Page 3-65: Figure 3-17 is a little confusing given that the y-axis is percentage change (from Liu t al.
2019); this figure description could clarify the axis.
• Page 3-66, Figure 3-18: The statement about evidence of effects down to 5 ug/m3 could be clarified
by adding justification for this statement. Is this statement based on the lower confidence interval
going below 0 around 5ug/m3, or something else? Is this statement true based on Figure 3-17 as
well?
• Page 3-66, Figure 3-18: And it may be helpful to explain why the uncertainty goes to 0 at the lowest
concentrations?
• Section 3.2.1.3: The description of the Wei et al. 2020 and Wei et al. 2021 studies is almost 2 pages
long; this section could be shortened, including the most relevant information.
• The lack of visible confidence intervals for the Di et al. 2017 results in Figure 3-19 is understandable
given the very large sample size and resulting lack of sampling variability (with virtually the entire
population sampled). Is this the case for other point estimates in this figure with no visible CIs? It
maybe be helpful to explain this.
A-43
-------
• Page 3-79: in the description of the Lefler et al. 2019 study, it may be helpful to explain briefly the
purpose of the spatial decomposition approach, and also clarify what we learn from the result of the
regionally sourced PM being closer to the primary exposure.
• Page 3-103, Figure 3-27: It would be helpful to explain the uncertainty bands here, and why they go
to 0 at lower concentrations.
• Section 3.2.2.2.7: The recent studies (e.g., Pope et al. 2019 and Pinault et al. 2017) have somewhat
contradicting results for the shape of the concentration-response curve at lower concentrations. It
may be helpful in this section to provide more details on the evidence from the 2019 document,
including the evidence of certainty down to 4ug/3 (and explain how the cut point of 4 was
determined).
A-44
-------
Dr. Alexandra Ponette-Gonzalez
Section 4 characterizes the recent welfare effects evidence that falls within the scope of the draft PM
Supplement.
a. Please comment on the identification, evaluation, and characterization of the available scientific
evidence in Section 4.
Overall, section 4 is concise and well written. Section 4 clearly describes the most recent evidence
regarding the relationship between particulate matter and visibility impairment, the non-ecological
welfare effect category considered in the PM ISA supplement.
Section 4.1 provides a clear and succinct summary of the evidence for visibility effects detailed in the
2019 PM ISA.
Section 4.2.1 highlights the lack of visibility preference studies in the US for the period 2009-2019. This
section presents findings from one new visibility preference study by Malm et al. (2019), which
demonstrates how choice of metric (contrast vs. visual range) influences variation in the level of
visibility considered to be acceptable, with contrast resulting in less variation in acceptability levels
among participants.
• On page 4-3, there is a brief description of how visibility preference studies are conducted.
Respondents are shown a series of photographs with differing visibility conditions and asked to
rate the quality of the scene and whether the scene is considered acceptable or unacceptable in
terms of visual air quality. A brief description of how the data are analyzed would be useful to
aid in the interpretation of the text and figures that follow.
• Specifically, it is unclear what the terms "acceptable" and "unacceptable" mean in the context of
visibility preference studies. To better understand Figures 4-1 and 4-2, the split point (e.g., 50%)
between "acceptable" and "unacceptable" in terms of visibility should be reported in the text.
• The term contrast is defined on page 4-3 as the "sharpness with which an object can be
distinguished from another object or background". Additional text could be added here to clarify
that scenes with no apparent contrast have a value of 0 (as plotted in Figure 4-2), and that as
contrast increases values become increasingly negative.
• Page 4-3 states that visibility preference studies use "similar protocols", which is inconsistent
with the PA assessment. On pages 5-25 and 5-26, the PA assessment indicates that the few
existing visibility studies employ different methods and that those methods have been applied
inconsistently. The text in the PM ISA Supplement should therefore be edited for consistency
with the draft PA assessment.
• Figure 4-2 does not include the WASH site (it is also not included in Malm et al. 2019) and
therefore "WASH" should be omitted from the caption.
A-45
-------
• The Figure 4-2 caption states that "an acceptability level of 90% would require contrast levels to
remain above a level of about -0.01". Per the figure and the citation, this should state that the
contrast would need to remain above -0.1.
Section 4.2.2 on advancements in visibility monitoring and assessment describes how PM composition
is changing in the US and how these changes have affected light extinction estimates.
• Given the paucity of new studies since the 2019 PM ISA, a brief description of advances in the
use of photographic images for quantifying atmospheric extinction is warranted.
• This section notes that the relative contribution of biomass burning, dust, and international
transport to visibility impairment is increasing. Given the importance of changing PM
composition, it would be useful to note the growing body of knowledge on airborne
microplastics (Brahney et al. 2020, 2021) and the nascent literature on microplastic effects on
light scattering and absorption (Revell et al. 2021). For instance, Brahney et al. (2020) found that
2.5 to 5% (on average, 4%) of identifiable dust particles in a subset of atmospheric deposition
samples were synthetic polymers. Brahney et al. (2021) and Revell et al. (2021) were published
just after the period which was the focus of the review for the supplement.
• Page 4-8, it may be important to mention that decreases in SO2 and NOx emissions have
coincided with increasing PM emissions from wildland fires as well as dust in some US regions,
such as the Great Plains (Lambert et al. 2020).
• The correlations mentioned between the results of image processing and measured atmospheric
extinction in hazy atmospheres were for which sites in the US? Were these relationships only for
the western US?
b. Please comment on whether the summary section in Section 4 appropriately characterizes recent
evidence in the context of the conclusions of the 2019 PMISA.
Section 4.3 provides a good overall summary of the recent evidence.
• Accounting for the distance between the observer and landscape feature results in less variability
in reported acceptability levels for visibility. It would be good to state the range of distances for
the sites in the text.
• A simple table similar to that on page 2-40 could be added to show how conclusions from the
2019 PM ISA (in one column) compare with new information presented in the Supplement.
c. Please comment on whether there are any topics or studies that fall within the scope of the draft PM
Supplement that should be added or receive additional discussion in Section 4 or any topics for which
discussion should be shortened or removedfrom Section 4.
• As mentioned above, in the context of changing PM composition, additional references are
needed on rising dust emissions in some US regions and on airborne microplastics, a component
of PM for which we know little regarding visibility impairment.
A-46
-------
Minor edits
• Page 4-9, line 5 should read "additionally" and line 8 should read "PM ISA have addressed"
References
Brahney, J., Hallerud, M., Heim, E., Hahnenberger, M., & Sukumaran, S. (2020). Plastic rain in
protected areas of the United States. Science, 368(6496), 1257-1260.
Brahney, J., Mahowald, N., Prank, M., Cornwell, G., Klimont, Z., Matsui, H., & Prather, K. A. (2021).
Constraining the atmospheric limb of the plastic cycle. Proceedings of the National Academy of
Sciences, 118(16).
Lambert, A., Hallar, A. G., Garcia, M., Strong, C., Andrews, E., & Hand, J. L. (2020). Dust impacts of
rapid agricultural expansion on the great plains. Geophysical Research Letters, 47(20),
e2020GL090347.
Malm, W. C., Schichtel, B., Molenar, J., Prenni, A., & Peters, M. (2019). Which visibility indicators
best represent a population's preference for a level of visual air quality?. Journal of the Air & Waste
Management Association, 69(2), 145-161.
Revell, L. E., Kuma, P., Le Ru, E. C., Somerville, W. R., & Gaw, S. (2021). Direct radiative effects of
airborne microplastics. Nature, 598(7881), 462-467.
A-47
-------
Dr. David Rich
Section 3 characterizes the recent health effects evidence that falls within the scope of the draft PM
Supplement.
a. Please comment on the identification, evaluation, and characterization of the available
scientific evidence in Section 3.
b. Please comment on whether the summary sections in Section 3 appropriately characterize recent
evidence in the context of the conclusions of the 2019 PMISA.
c. Please comment on whether there are any topics or studies that fall within the scope of the draft PM
Supplement that should be added or receive additional discussion in Section 3 or any topics for which
discussion should be shortened or removedfrom Section 3.
Overall, Chapter 3 of the ISA supplement is well written, and includes appropriate studies published
since the previous ISA. A few papers that could/should be added are provided below. The document
summarizes evidence for each study appropriately, but there are several sections where modification or
additional information is requested to provide greater clarity. These are described below as well.
1. Throughout each disease category reviewed in Section 3, the text descriptions of each study's
findings do not present the incremental PM2.5 concentration to which each effect estimate is
scaled. Based on the figures, these appear to all be scaled to a 10 |ig/m3 increase in PM2.5
concentration. However, a statement providing this increment and stating that all study findings
provided are scaled to this increment needs to be added to each section. Without such a
statement, results cannot be compared across studies.
2. Consistent language could be used for each statement of findings across all sections to improve
clarity. In some sections (e.g., Page 3-17, lines 17-19; Page 3-19, line 19-20), an effect estimate
is described as a "positive association" or just "association", while in other sections, effect
estimates are described as the risk of disease X associated with each 10 |ig/m3 increase in PM2.5
concentration in the previous 3 days. To improve clarity, the specific effect measure for the study
should be provided, and "positive association" used only when summarizing evidence across
studies for a specific disease section.
3. Page 3-9, lines 28-36 - Discussion of Liu et al (2020) and interpretation of the effect estimates
should be changed. The main OR presented is 1.03 (95% CI = 0.96, 1.12), which is too imprecise
to say that it supports anything but a null association. Further, the text currently states that effects
increased within tertiles of 'long-term NO2 concentrations", but again these effects are imprecise.
I suggest that this be rewritten to state that there appeared to be an increased odds of MI
hospitalization associated with each 10 |ig/m3 increase in PM2.5 concentration in lag days 0-2, but
only among the highest tertile of long-term NO2 concentration.
4. Page 3-9, line 38 - Another study by our group is also relevant here and should be added. Wang
et al (2019) also examined the rate of ST elevation myocardial infarction associated with several
PM markers including PM2.5 and ultrafine particles, as well as several gaseous pollutants. This
study is discussed in the Accountability Studies section, but should be included here as well.
Wang et al (2019) examined whether associations between each pollutant and the rate of STEMI
A-48
-------
changed following a series of air quality policies and an economic recession. However, it
provides estimates of the rate of ST elevation myocardial infarction associated with each
interquartile range increase in PM2.5, ultrafine particle, and other particle measures in the
previous few hours and days, and should be included in this section and Figure 3-1 as well.
WangM, et al. Triggering of ST-elevation myocardial infarction by particulate air pollution in
Monroe County, New York; before, during, and after multiple air quality policies and economic
changes. Environmental Health 2019; 18(1) :82.
5. Page 3-9 and 3-10; Page 3-12 and 3-13; Page 3-13 line 33; Page 3-17 - In each of these
locations, a study by our group (Zhang et al, 2018) is missing. It is described on Page 3-21 lines
24-33 as an accountability study. However, Zhang et al (2018) is also relevant to these sections
and should be added to each section's text and to each summary figure for each section. The
study estimated associations between 1 to 7 day average ambient PM2.5 concentrations and the
rate of hospitalizations for ischemic heart disease, heart failure, arrhythmia, cerebrovascular
disease, and other CVD category hospitalizations among NY adult residents from 2005-2016.
Further, it provides such estimates separately for 3 time periods (2005-2007, 2008-2013, 2014-
2016) with progressively lower average PM2.5 concentrations, and also separately for the 3
upstate NY sites and the 3 New York City sites (also with different average PM2.5
concentrations). These findings should be included in the text and summary figures for these
sections as well. The corresponding publication describing trends in PM2.5 and other pollutants
during the study could also be cited (Squizzato et al 2018).
Zhang W, et al. Triggering of cardiovascular hospital admissions by fine particle concentrations
in New York State: before, during, and after implementation of multiple environmental policies
and a recession. Environmental Pollution 2018;242(Pt B):1404-1416.
Squizzato S, et al. PM2.sand gaseous pollutants in New York State during 2005-2016: spatial
variability, temporal trends, and economic influences. Atmospheric Environment 2018;183:209-
224.
6. Page 3-13 lines 16-20 -1 suggest you add a conclusion statement as to whether these studies are
or are not consistent with the causal conclusion in the 2019 ISA, similar to that on page 3-15
lines 35-37 that states "Overall, these studies support and extend the limited evidence in the
2019 PMISA, reporting positive associations between short-term PM2.5 exposure and HF". For
all studies presented in the supplement, this kind of a conclusion statement would provide a
clearer interpretation of the recent evidence for that disease group.
7. Page 3-17, line 26-27 - Not all of the 4 studies described support an association between PM2.5
and arrhythmia. Only 3 do. What do you mean by the statement that these studies "extend" the
findings of the 2019 PM ISA? Please clarify.
8. Page 3-19 - Section 3.1.1.2.6 - Text is lacking detail to support the conclusions made in this
section. Please add references and or descriptions of the studies/findings that support all of these
conclusions. For example, please provide the references and findings from the studies on which
the following conclusions were made, or the locations in the ISA Supplement text where they are
discussed in more detail:
A-49
-------
a. Lines 24-25 - "recent studies indicate that associations between short term PM2.5
exposure and cardiovascular mortality are relatively unchanged in co-pollutant models".
b. Line 27 - "factors that have been shown to vary between cities and regions of the US,
such as housing characteristics, have been shown to explain some of the city-to-city and
regional variability observed in PM2.5 mortality associations in multi-city epidemiologic
studies".
c. Line 30-32 - "the concentration response relationship between short-term PM2.5 exposure
and mortality further supports a linear relationship, with less confidence in the shape at
concentrations below 5 |ig/m3",
9. Page 3-20 line 13-19 - You provide an effect estimate and 95% CI for deSouza et al (2021), but
not the other studies that reported null associations. Please add the effect estimates and
confidence intervals for these other studies as well, to justify the conclusions you make. These
are needed for both studies finding associations and those that do not.
10. Page 3-20 line 11 - You say that these studies "expand upon the overall assessment of potential
confounding", but what conclusion do you make outside of just that they expand the assessment?
Do they strengthen the association/conclusions made in the 2019 ISA or do they weaken them?
Please provide such a conclusion for these studies.
11. Page 3-20 line 21 - What lag periods/days define "immediate", "delayed", and "prolonged"?
This is provided elsewhere in the ISA Supplement, but should be added here for clarity.
12. Page 3-22 - line 4-5 - The text describing Zhang et al (2018) findings is provided before the
statement: "Overall, across the endpoints examined, there were notable differences (i.e.,
reductions in hospital admissions) after policies were implemented compared to before. " This
should be removed from the supplement, as this was a descriptive statement only, and not the
purpose or a full analysis of the study/paper. In Zhang et al (2018), overall across the endpoints
examined, the main conclusion of the study is that the relative rate (rate of CVD and cause
specific CVD hospitalizations associated with each specified unit increase in PM2.5 concentration
in the previous 1 to 7 days) increased after the policies were implemented and the economic
recession occurred (which was inconsistent with our a priori hypotheses) suggesting the same
dose/concentration of PM2.5 was associated with a greater rate of CVD hospitalizations (perhaps
increased PM toxicity due to changes in PM composition).
Chapter 3 - Cardiovascular effects - long term PM2.5
1. Again, throughout this whole section, the concentration increase to which each effect estimate is
scaled needs to be provided to the reader. As provided in the figures, these appear to be scaled to
a 5 |ig/m3 increase. This increment needs to be provided at the beginning of each section, or
provided for each study result description.
2. For each study described in this section, say what time frame defined "long term", as that is not
consistently provided in each section.
3. Page 3-28 - Throughout this whole section, it would help with clarity if you could provide the
main effect estimates and confidence intervals for the studies on which you state there was an
association or there were no associations between cardiovascular mortality and PM2.5
concentrations. This would allow a comparison of effect size (i.e. what is the rate/risk/odds of
mortality associated with each 5 |ig/m3 increase in PM2.5 concentration in the previous 1 year).
Again, please indicate to what increment these effects are scaled, and what time period in each
A-50
-------
study defines "long term". This is not required, but would help greatly and allow a more
quantitative comparison of study findings over time rather than just a comparison of
"association" or "no association".
4. Page 3-29, lines 22-31. - Details on several studies on which conclusions are made are not
provided and need to be added here. Please provide references for the studies you describe, and
the main effect estimates on which the conclusions in the paragraph are made. Alternatively,
provide the section(s) in the ISA Supplement where they are provided.
5. Page 3-31 lines 10-11 - Please provide the other tertile effect estimates if you are arguing that
the highest tertile effects are "stronger".
6. Page 3-32, line 18 - What is ML? Do you mean MI?
7. Page 3-33, line 4-6 - Please provide the effect estimate and 95% confidence interval for Miller et
al, so that it matches the quantitative results presentation of the new studies reviewed below.
Thus, the reader can make a quantitative comparison of the size of any effect estimates between
new studies and studies presented in previous ISA's, not just a comparison of "no association" or
"association".
8. Page 3-36, line 3 - Please provide the effect estimate and 95% confidence interval for the study
with the conclusion of "no association with cIMT" to be consistent with the positive association
with CAC presented above it.
9. Page 3-37, line 5-6 - "... was observed among MESA participants." Provide the reference for this
study
10. Page 3-37, line 15 - Please provide effect estimates for other tertiles, not just the highest. Was
the deciding factor determining whether there was a positive association based on the effect
estimate or whether the effect was statistically significant? The text needs to provide all the
tertile effect estimates to allow the reader to judge that.
11. Page 3-38, line 12 - What effect estimate is this 1.17? Is this a risk ratio? odds ratio?
12. Page 3-39, line 12 - Please provide the effect estimate and 95% confidence interval for the "null
association" of Wang et al (2020).
13. Page 3-39, line 25 - Provide the main effect for Shin et al (2019), to allow the reader to compare
that to the effects when adjusting for NO2 and O3 that are provided
A-51
-------
Dr. Jeremy Sarnat
Charge Question 4 - Section 3 (Health Effects)
3.2. Mortality
I want to commend EPA staff for their diligence and careful preparation of the supplement to the PM
ISA. The supplemental material serves as a critically useful update to the state-of-the-science regarding
PM and health. I am comfortable with most of the interpretations and general conclusions made within
this chapter (and within this ISA supplement, in general).
• Broadly, the additional, recent epidemiologic findings related to mortality are exceedingly important.
In preparing the 2019 ISA, the relatively few studies on long-term exposure and mortality cast, in
particular, provided relatively limited observational evidence of causal associations at levels below
the annual PM NAAQS. The current supplement addresses this shortcoming, showing excess
mortality at concentrations below, in some locations substantially, the current annual NAAQS levels.
There is also notable, added attention in the supplement to issues of confounding, which I'll address
in the points below.
• The discussion of specific exposure factors as a source of heterogeneity in observed short- and long-
term PM mortality risk is important and probably deserves more attention than given in the
supplement. EPA notes that the housing stock and commuting, as well as land use and traffic) may
explain some heterogeneity, but that these differences 'cannot be attributed to one factor' (3-53).
This seems like an accurate conclusion which should be examined a bit more fully. Clearly, some of
the factors mentioned are truly exposure-related (e.g., housing stock and indoor exposure to ambient
PM), while others are more closely associated with differential emissions and chemical composition.
I'd recommend including some language on why these factors are associated with observed
heterogeneity, maybe in the discussion of the very interesting Baxter et al (2019) results (3-63).
Maybe reference to Section 3.4 in the 2019 ISA is needed?
• Generally, I found the inclusion of alternative methods for assessing potential confounding to be an
improvement over the previous versions of the ISA (Section 3.2.1.2.3), where there was close to sole
reliance on the findings from multipollutant modeling. The causal models and results from
accountability studies, in particular, provide added confidence that the observational findings related
to PM are not unduly confounded by PM co-pollutants.
• While the treatment of confounding in this draft is an improvement over previous discussions in past
ISA's, I continue to struggle with the weight multipollutant models are given when ascertaining the
presence of confounding, and the relatively simple manner that the results are discussed.
Generally, I feel that the ISA parses extremely small differences in coefficients or confidence intervals
around coefficients as meaningful, where I view these differences as likely by-products of residual
errors in the exposure assignment approach. This does not imply that the results don't have value, they
most certainly do in the context of a wider body of similar findings, but the observed differences from
individual studies should be interpreted narrowly.
I would also recommend that language be included in the supplement regarding the utility AND
limitations of multipollutant modelling as a means of ascertaining confounding. These limitations
A-52
-------
necessitated, in part, the development of alternative approaches. As currently written, coefficient
stability or consistently is still presented as a 'gold standard' for addressing this issue - which it is not.
Similarly, while there is ample discussion of the specific difference among the causal models, there is
very little discussion on the limits of the various approaches or the relative benefits of one model over
another.
• There is a question of overall ISA scope, for each of the criteria pollutants, which comes up when
reviewing the multipollutant models. As with previous ISA's, the supplemental material on PM
multipollutant modeling is still solely focused on controlling for confounding (i.e., looking for
changes in coefficients while including other pollutants), but not on co-pollutant effects or
synergistic effects which might be expressed through observed joint effects or effect modification. I
appreciate that the scope of the PM ISA is to determine PM health effects, but the discussion
involving the multipollutant modeling findings begs, in my opinion, some mention of mixtures and
co-pollutant exposures. A related issue concerns specification of the multipollutant models. A key
area of uncertainty is whether epidemiologic models more properly designed to assess the effects of
pollutant mixtures, either in a joint effects or effect modification setting, that may include interaction
terms among the pollutants, are more efficient and provide better fits to the C-R relationship than
models with two, independent pollutant terms. As long as a multipollutant ISA does not currently
exist, there should be space to address these issues in the individual pollutant ISA's.
• A related source of uncertainty regarding specification of the co-pollutant models is the potential
non-linearity of associations between PM and its co-pollutants. The use of linear expressions, within
a co-pollutant setting, to control for confounding of non-linearly correlated co-pollutants could lead
to imprecision and/or bias; an appearance of effects associated with PM, where they do not exist.
• There have been several recent epidemiologic studies examining out-of-hospital cardiac arrest and
PM, mainly from wildfires (Section 3.2.1.2.1; 3-57). In addition to several recent non-US studies
(from Japan, Israel, Europe), a US study that can be cited is:
o Jones, Caitlin G., et al. "Out-of-Hospital Cardiac Arrests and Wildfire-Related Particulate
Matter During 2015-2017 California Wildfires." Journal of the American Heart Association
9.8 (2020): e014125. Authors found OHCA-biomass smoke association, strongest on lag day
2 (OR = 1.70; CI= 1.18-2.13).
• Figure 3-17. Please present both curves on the same scale to aid in interpretability.
• A theoretical question related to the shape of C-R curves (for mainly long-term exposure and
mortality is whether we might expect to see differential measurement error at lower observed PM
concentrations. For studies based primarily on measured estimates of population exposure I could
hypothesize why differential error may exist and lead to differences in the shape of the curve along
its full observed range. For studies using modeling or hybrid approaches, I have a harder time
thinking of specific sources of error, but imagine that they too may exist. I raise this point only to
draw attention to the role of measurement error as a potential driver of the shape of this function and
as a note of caution in overinterpretation of observed supralinear or superlinear trends.
3.3. Key Scientific Topics
Section 3.3 represents a very important addition to the 2019 ISA and the EPA deserves tremendous
credit for presenting more expansive identification of 'at-risk' to include both traditional definitions
involving biological susceptibility, as well as those exposed to elevated PM due to social and economic
A-53
-------
disparities. In the 2019 ISA, an explicit discussion of environmental justice was largely absent, born
disproportionately among Black and Hispanic communities, which is rightly interpreted as a key factor
leading to disparities in PM risk within the current supplement.
A-54
-------
Dr. Neeta Thakur
Section 3.3 Key Scientific Topics
a. Please comment on the identification, evaluation, and characterization of the available scientific
evidence in Section 3.
b. Please comment on whether the summary sections in Section 3 appropriately characterize
recent evidence in the context of the conclusions of the 2019 PMISA.
c. Please comment on whether there are any topics or studies that fall within the scope of the
draft PM Supplement that should be added or receive additional discussion in Section 3 or
any topics for which discussion should be shortened or removedfrom Section 3.
Overall assessment:
I found the document incredible comprehensive yet disorienting. In section 3, there is a lack of
consistency on how studies are discussed, specifically describing the study populations included to
better contextualize results, for example a study that includes a predominantly higher SES, younger
population, results are not comparable to a study in an older, Medicare population. The study population
differences are only discussed in detail when the study includes an analysis across populations, such as
on page 3-31, In 30-38 discussing results from Elliot, Hart, and Weaver. This will also be helpful in
defining what an "adequate margin of safety is," in that, studies that include or are predominantly in
what has been established as "vulnerable groups" that these studies should perhaps be lifted in the
review - especially in light of the stated focus on equity and environmental justice.
Section 3.3.1 (studies at near ambient concentrations):
The two exposure studies, increase mechanistic plausibility, particularly for cardiovascular effect. I did
have issue with some of the details of how the studies are presented which effect interpretability (both
towards the null or towards effect) that I have included in specific comments that follow.
Page 3-121 In 19-31: For the 2015 Hemmingsen study it would be important to highlight that this was
conducted in NON-smokers in addition to obese individuals given the focus on vasomotor function
(endothelium mediated & non-mediated) and heart rate variability response to short-term PM2.5
exposure. In addition, the way the results are currently summarized, it is unclear which arm (exposed to
filtered air vs. non-filtered air) had decreased vasomotor function after exposure. It would be important
to highlight that this is in the non-filtered air arm as evidence of a potential biologic mechanism.
Page 3-122: Wyatt 2020 exposure study, could likely be summarized more succinctly with attention to
significant results. Specific comments regarding the presentation of this study follow.
Page 3-122, In 9-10: Results for pulmonary function measures need to be clarified. As written, the
change in values (e.g. 1.2% decrease in FEV1/FVC) implies that this is a 1.2% decrease from the pre-
measurements, while in fact it is 1.2% difference in FEV1/FVC from those in the unexposed group as
the authors conducted a mean difference analysis controlling for baseline (pre) measures rather than a
A-55
-------
repeated measures analysis. For PEF and FEV1 it would be important to make this distinction as well
given that these measures are often written as %predicted and the use of %difference without annotation
mistakenly inflates the findings.
Page 3-122, In 12-18: Similar comment to above, need to clarify that these are mean differences between
the exposed and unexposed group. The Supplement file includes these raw results were it is more clear
that these are mean differences and not percent changes.
Page 3-122, In 37-39: Could consider adding that the approach to the analysis, mean difference vs.
repeated measures analysis, was also a limitation to the study.
Section 3.3.1 (studies of associations between PM2.5 with COVID19):
For this section, I think the ordering of the section should be reconsidered, with the at-risk section
proceeding the covid 19 section given that this latter section is brand new. Otherwise, I don't have much
to add regarding the assessment of the data for COVID 19 and PM2.5 and agree with the conclusion that
this area needs further study.
Page 3-126, In 10: Should "However" be changed to "In addition," as these factors likely co-occur with
regions with higher PM2.5 exposure and, thus, confound the findings of the mostly ecological studies
presented (which is highlighted in the paragraph below with the inclusion of the critiques by Bourdrel et
al. and Villeneuve and Goldberg).
Section 3.3.1 (assessment across at-risk populations):
For this last section, At-risk populations, I struggle with the way the new data re health effects of PM2.5
across SES strata. When consider as single indicators, the results are mixed, but as SES is a complex
construct, this approach simplifies SES and may lead to mis-classification and varied results based on
what indicator is used. However, pretty consistently, the included studies that use a composite measure
of socioeconomic status, rather than a single indicator (such as income) that there is both
disproportionate burden of exposure (Figure 3-30, Tanzer 2019, Lee 2020, Weaver 2019, Han 2020) and
health effects, particularly for long-term effects, including disease-related mortality (Zhang 2021, Bevan
2021, Wyatt 2020b [particularly panel D of Fig 3-34]). Defining what composite measure of SES is for
the purpose of this ISA supplement would be helpful for comparability. For example, several of the
composite measures also include race/ethnicity and other vulnerability factors, such as comorbid
conditions. These composite measures that include these other vulnerability factors highlight that the
facets of SES are complex and interactive. This may or may not be appropriate depending on the goal of
distilling effect - is it looking for single risk factors, versus vulnerability factors that co-occur or are
highly correlated with one another. Some discussion of how to consider the results from these studies
versus those that look at single indicators should be included.
Re the included studies on race/ethnicity, have concerns about use the term "non-white" populations as
opposed to explicitly stating comparator group(s). I also struggle with requiring studies to have to have a
comparative population, this excludes a number of studies that may have relevance.
A-56
-------
Page 3-144, Short-term exposure studies and mortality general comment: This is more a comment on the
limitation of interpretation of the studies. Short-term exposure studies would stand to benefit from
considering the annual PM2.5 exposure in their model to better understand whether it is the higher
chronic exposure to PM2.5 that increases susceptibility to short-term increases in PM2.5 or the change
in exposure itself. Understanding how differing levels of chronic exposure modify response to short-
term changes would be useful when defining an "adequate margin of safety". Would consider
highlighting this as a potential limitation to interpreting these studies and an area for future research.
Specific Comments:
Page 3-129, In 24-25: Would help to add how clusters were determined, e.g., "neighborhood clusters,
derived by Ward's hierarchical clustering of 11 census-derived socioeconomic variables, located in
within three counties..
3.3.3.2 Race/Ethni city
Page 3-140, In 20-39. Appreciate the EPA staff inclusion of studies that demonstrate that the overall
decline in PM2.5 exposure is largely driven by decrease exposure in White populations but that there
has been an increase in exposure for BIPOC groups and that exposure is inequitable distributed when
considering populations generating vs. those exposed.
A-57
-------
Dr. Barbara Turpin
Charge Question 2 - Section 1 consists of an introduction detailing why the draft PM Supplement is
being developed along with the rationale and scope for the topics and studies considered.
a. Please comment on the clarity of the section, whether the scope is appropriate for the purpose of
the draft PM Supplement, and whether additional information is needed to convey the purpose of
the draft PM Supplement and the basis for the targeted evaluation conducted.
Section 1 was generally clear. However, I strongly suggest that the EPA clarify the rationale for the
scope of the ISA Supplement. The ISA Supplement is limited to providing new information concerning
effects that were determined to be causal in the original ISA. This means that new data that may change
the "causality assessment" for nervous system effects (for example) is not considered.
Charge Question 3 - To ensure that recent studies are put in the context of the conclusions of the 2019
PMISA the draft PM Supplement pulls in information verbatim from the 2019 PM ISA to orient the
audience. Two ways this was done in the draft PM Supplement is through Section 2 which is the
Integrated Synthesis Chapter (i.e., Chapter 1) of the 2019 PMISA and leading off each health and
welfare effects discussion in Section 3 and 4 with the Summary and Causality Determination from the
2019 PMISA.
a. Please comment on this approach and whether any additional modifications to the structure of
the document can be made to better integrate evidence evaluated in the draft PM Supplement
with conclusions from the 2019 PMISA.
This approach makes sense. It was quite helpful for the ISA Supplement to begin with the rational,
followed by the Integrated Synthesis Chapter from the 2019 PM ISA.
Charge Question 4 - Section 3 characterizes the recent health effects evidence that falls within the scope
of the draft PM Supplement.
a. Please comment on the identification, evaluation, and characterization of the available scientific
evidence in Section 3.
b. Please comment on whether the summary sections in Section 3 appropriately characterize recent
evidence in the context of the conclusions of the 2019 PMISA.
The strengths and limitations of various types of evidence were clearly presented. The tables summarize
findings from last review and demonstrate coherence across epidemiologic, animal toxicology and
controlled human exposure studies, supporting a causal relationship with cardiovascular effects. The
addition of new studies demonstrate consistent results while applying advanced methods for addressing
confounding, strengthening the case for causality.
A-58
-------
c. Please comment on whether there are any topics or studies that fall within the scope of the draft
PM Supplement that should be added or receive additional discussion in Section 3 or any topics
for which discussion should be shortened or removedfrom Section 3.
There is considerable evidence calling into question the adequacy of the short term PM2.5 standard.
This includes three epidemiologic studies with analyses restricted to 24-hr concentrations below 25
ug/m3 (Table 3-10). But the integrated science assessment, risk assessment and policy assessment do not
provide the analyses needed to evaluate alternative levels and forms of the short term standard.
Regarding the section on effects by race and socioeconomic status, it is worth noting that the words
"vulnerability," "sensitivity," and "susceptibility" are defined differently across the fields of research
covered in the report. Consider using the words "intrinsic factor" and "extrinsic factor" instead.
COVID-19 material would be better placed at the end of the section where it is presented.
People of color or communities of color is a more appropriate descriptor than non-white.
Charge Question 5 - Section 4 characterizes the recent welfare effects evidence that falls within the
scope of the draft PM Supplement.
a. Please comment on the identification, evaluation, and characterization of the available scientific
evidence in Section 4.
Identification, evaluation and characterization of available evidence in Section 4 is satisfactory.
However, please also consider the following:
Section 4 of the ISA supplement notes that changes in PM2.5 composition are resulting in an increasing
"closure" gap between light extinction and light extinction predicted from particle composition.
(Revised prediction methods address this.) One convincing reason, which is noted in the ISA
Supplement, is the increase in the contribution of wildfire PM2.5. Another possible contributor is worth
considering. Riva et al (2019) argues that as the sulfate/organic ratio decreases, the fraction of sulfate
present as organosulfate increases. Organosulfate and inorganic sulfate have different optical properties
and hygroscopicity which are not accounted for in the IMPROVE light extinction model. The impact of
organosulfates on the closure gap will be most important in the southeast.
Riva, M., Chen, Y., Zhang, Y., Lei, Z., Olson, N. E., Boyer, H. C., ... & Surratt, J. D. (2019).
Increasing isoprene epoxydiol-to-inorganic sulfate aerosol ratio results in extensive conversion of
inorganic sulfate to organosulfur forms: implications for aerosol physicochemical properties.
Environmental science & technology, 53(15), 8682-8694.
b. Please comment on whether the summary section in Section 4 appropriately characterizes recent
evidence in the context of the conclusions of the 2019 PMISA.
A-59
-------
I agree with the assessment in Section 4 of the ISA Supplement that recent evidence confirms the "well-
established relationship between PM and visibility" and evidence is sufficient to conclude this
relationship is causal.
Charge Question 6 - The Summary and Conclusions section (Section 5) provides an overview of the
evidence evaluated in the draft PM Supplement.
a. Please comment on the level of detail provided within this section and whether revisions should
be made to further summarize recent evidence.
Level of details seems appropriate.
Please change "non-white" to "people of color."
A-60
-------
Dr. Marc Weisskopf
Charge Question 4 - Section 3 characterizes the recent health effects evidence that falls within the scope
of the draft PM Supplement.
a. Please comment on the identification, evaluation, and characterization of the available scientific
evidence in Section 3.
Overall the draft PM supplement has captured and characterized the new literature well. However, given
the importance of studies that focus on lower levels of exposure—in particular levels below current
standards—for making decisions on limit setting, it would be helpful to have a section that groups these
papers and describes their findings.
See also specific comments below for some minor issues.
b. Please comment on whether the summary sections in Section 3 appropriately characterize recent
evidence in the context of the conclusions of the 2019 PMISA.
The summary in general does a good job of characterizing recent evidence. I have a couple of broad
comments, though:
1) Where adjustment for co-pollutants is discussed, though (3-48,11. 8-10, 26-27) it is not clear
whether co-pollutant models are just two pollutant models or multi-pollutant models. I would
think multi-pollutant models would be stronger evidence for a specific effect of PM2.5, although
the issue that co-adjusting for a pollutant that reflects something in PM2.5 (e.g. N02 as a marker
of traffic pollution that is part of PM2.5) is complicated as it then changes the interpretation of
the PM2.5 estimate. Adjustment for co-pollutants that are not reflecting components of PM2.5
(e.g. secondary gases) doesn't have that problem.
2) I am a little uncomfortable with the use of language like "causal modelling methods" to refer
to the specific set of studies it is currently referring to. Standard studies (most of the ones
referenced in this ISA and past ones) should be considered causal modelling methods, under the
assumption that control of biases (e.g. by adjustment for confounders) has been accomplished.
The studies that that language now refers to take different approaches to avoid or rule out biases,
and for certain types of bias these can be considered advanced or more robust than the other
literature. But these perhaps more advanced methods also have their own assumptions and also
sometimes have issues they cannot address. So I would favor replacing "causal modeling
methods" to describe this literature with something else—maybe "advanced causal modeling
methods", "modern causal modeling methods", or the like? The robustness of findings to the
different approaches is perhaps of most relevance here. Focusing on describing the kinds of
threats to validity that the newer methods are better at avoiding would be better than just
referring them as causal modeling methods.
A-61
-------
c. Please comment on whether there are any topics or studies that fall within the scope of the draft PM
Supplement that should be added or receive additional discussion in Section 3 or any topics for which
discussion should be shortened or removedfrom Section 3.
Other than perhaps clarifying the two points above, I think what is there is good.
Specific comments:
1) P. 3-3,1.8: Should be Table 3-1 I believe
2) P. 3-6,11. 30-32: Much of the evidence in the 2019 ISA seems to be from just two-pollutant
models. But such results are tougher to work through to inform lack of confounding than if
multiple co-pollutants were in the model together (e.g. co-adjusting for 03 could still be
confounded by N02).
a. Same two pollutant issue in 3.1.1.2.6
b. 3.1.1.2.7,11 13-19. This is limited additional evidence against confounding by co-
pollutants and still has the issue of only 2-pollutant models.
3) 3.1.1.2.2,1.7-8: personal variables like sex, obesity, smoking cannot really confound a relation
with an estimate of ambient PM2.5 (see Weisskopf and Webster, Epidemiology, 2017) as they
do not contribute to that exposure metric (perhaps outside of SES factors driving residence
location. So only to the extent they vary by location even after whatever SES/residential
adjustment has been done). I would not downgrade any study results because of lack of control
for such personal factors.
a. This issue shows up in several other places in the document as well.
4) 3.1.1.2.3 Should there be some mention in summary of newer literature seeming to show more in
vulnerable pops? No mention now.
5) 3.1.1.2.5: It's not clear to me what the overall conclusion is from what is presented in this
section.
6) 3.1.1.2.6,11. 31-32: What is the literature being referred to for this statement about linearity in the
C-R function?
7) 3.1.1.3,1.14: See overall comment about using the term "causal methods".
8) 3.1.1.3, 3-21,11.30-33 (related to Zhang et al., 2018): Need to indicate that the results indicated
are for the 0-6 day lag average.
a. I don't think the description of this study captures the results well. What is described
seems to focus more on the 0-6 day average, but the results seem a little different for
shorter lags and those seem to be what the authors point to more. The main result seems
to be that the per IQR change in outcomes doesn't change much over the different
periods, except possibly for IHD and MI for which effect sizes seem to get larger in the
later periods (although there is variability by lag period and part of NY). While differing
composition of the PM could be behind this, it could also result from a steeper C-R at
lower exposure levels (although the authors don't comment on this).
9) 3.1.1.3, 3-22 (related to Wang et al., 2019)
a. 11.10-12: Should indicate that the authors' hypothesis was based on the thinking that
secondary species were more important and these went down more in the after period.
b. LI. 22-28: This description of the findings is written as if the results are suggesting some
changes. Instead these results all seem rather null and indeed are interpreted that way by
the authors. I think that interpretation, rather than associations differ by time periods,
needs to come across.
A-62
-------
10) 3-22,11. 33-35 (and top of 3-23): This is not really more causal an approach than typical
adjustment, just a slightly different way of controlling for potential confounders (with maybe the
exception of positivity as they apparently checked the PS distribution for cases and controls for
this).
11) 3-23,1.16 (Qiu et al., 2020): the phrase "with the caveat that the authors had the resources to
obtain all the 16 potential unmeasured confounders" is not correct. The authors state that a key
assumption is no unmeasured confounding and note that they *don't* have the resources to get
all unmeasured confounders. However, they assume one of the most critical is temperature and
their findings are robust to different treatments of temperature. Variables that do not vary over
time are eliminated as possible confounders by the case-crossover design.
12) 3-34,11. 16-18: Why is "associations" used here? Is this meant to indicate less certainty about
causality?
13)Figure 3-6: Should the results among diabetics be shown separately here for the Hart et al. 2015
paper?
14)3.1.2.2.6, 3-38,1.12: Where does the 1.17; 95%CI: 1.10-1.22 come from in the paper? I see 1.13
(1.08, 1.17). And need to indicate that it is an HR and what the unit is—for the 1.13 it is an IQR
(3.98 (J,g/m3) higher level.
15)3-39,11. 27-28: Not clear which HR refer to which pollutant (same for 11. 30-321).
16) 3-69,11. 3-4: A more important aspect of the negative control exposure approach is that if no
association is seen with the negative control exposure (which the authors report, indicated in 11.
11-12), then that indicates no confounding by any measured or unmeasured variables that affect
both the exposure of interest and the negative control exposure. This should be stated as well.
17) 3-69,11. 8-11: With the IV approach I don't see how the effects can necessarily be attributed to
PM2.5—they suggest an effect of the pollution, but can't distinguish which pollutants affected
by the IV account for the effect.
18) 3-95,1.27:1 believe the suggestion is that the non-null associations are confounded. Also not
clear to me that the coefficients are not interpretable if the whole point is that each PM2.5 metric
should be giving the effect of PM2.5.
19) 3-99,11. 7-9: As stated just above, I think this approach is very dependent on the exposure-
covariate relations in the two populations. Not clear to me how strong the indirect adjustment
results really are.
20) 3-108, table 3-7 Wu et al., 2021 results: HR are for a 10ug/m3 increase in PM2.5 (not decrease
as stated), (also fix at 3-116,1.1)
21) 3-110, table 3-7 Schwartz et al., 2018b results: need to add "less" to the effect estimate (0.89 less
years)
22) Table 3-7: DID studies can be quite powerful. Would be good in future to consider ones that also
tried to address co-pollutants that could also be affected by the changes (e.g. policy, plant
closure) if that is a possibility.
A-63
-------
Dr. Corwin Zigler
The Draft Supplement to the 2019 PM ISA (henceforth, the "Draft Supplement") includes ample
discussion of recent studies employing so-called "causal modeling methods" to evaluate the health
effects of PM. The emphasis on causal modeling studies that have appeared since the the 2019 PM ISA
is important and appropriate, and I commend EPA staff on their navigation of this evolving area. The
emergence of these studies as new evidence propelling the reconsideration of the PM standard seems
motivated by at least two connected threads; 1) these methodologies, while established across the
domains of statistics, epidemiology, medicine, computer science, and the social sciences, represent a set
of novel approaches for estimating the health effects of PM exposure with potential to reduce
uncertainties in the current body of evidence, and 2) they are particularly well suited to address many of
the uncertainties highlighted by members of the previous CASAC and the previous Administrator in the
2020 decision to retain the PM2.5 standards.
I believe that the inclusion of these "causal modeling" studies that have appeared since the 2019 PM
ISA support and strengthen the EPA's and the current CASAC's conclusions regarding PM exposure
and mortality. It is essential that the ISA process adopt a modern stance on "causal inference" in the
context of air pollution epidemiology and keep up with this fast-evolving area. However, no operational
definition of "causal modeling methods" is offered in the Draft Supplement, making it difficult to
evaluate how exactly the emergence of these methods contributes to the existing body of evidence,
particularly for readers without a detailed understanding of the methodological technicalities. I have
several comments and concerns related to clarifying the role of these methods in the Draft Supplement.
General Comments
1. The role of "causal modeling methods" in the weight of evidence causality determinations,
a. While the Draft Supplement does not provide any operational definition of "causal
modeling methods," it seems clear that EPA is using this term to describe a specific class
of established methodologies found across literature in statistics, epidemiology, computer
science, medicine, and the social sciences, but relatively new to the body of evidence on
PM health effects. As a means to distinguish a certain class of statistical or epidemiologic
analysis approaches, iden- tifying certain methods as "causal inference" (or similar) does
serve a purpose, as it connotes the use of some specific concepts and analysis tools such
as propensity scores, explicit mod- eling of potential (or counterfactual) outcomes,
double robustness, or the features of quasi- experimental designs. However, when read in
the context of the PM ISA, I am very concerned that the outright labeling of some studies
as "causal" could confuse the purpose of these studies amid the weight of evidence
causality determinations laid out in the 2015 ISA Preamble. For example, readers of the
PM ISA less acquainted with these areas of statistical and epidemio- logic methods
research might inappropriately surmise that methods that are not described as "causal
modeling" should somehow be judged as less relevant for the causality determinations
when considering the weight of evidence when, in fact, both "causal modeling" and more
tra- ditional epidemiologic methods should be employed in service of establishing the
entire body of evidence. In my view, the recent emergence of studies employing these
types of method- ologies for PM health effects supports and strengthens the EPA's
A-64
-------
conclusions through reducing uncertainties around some common threats to validity of
observational epidemiology, most notably, concerns around confounding bias. However,
the importance of these studies should not be misconstrued as detracting from the
importance of high quality but more traditionally- conducted epidemiologic analyses,
which should also contribute substantial weight to the body of evidence.
2. Justify the importance of specific "causal modeling approaches" according to the threats or
uncertainties they purport to mitigate.
a. What makes the "causal modeling methods" described in the Draft Supplement important
is their specific features for clarifying, mitigating, or even resolving common threats to
causal validity in observational epidemiology. In this regard, these methods are no
different from a variety of other types of analysis strategies that appear in the body of
evidence that do not receive the "causal" label. An illustrative example is the discussion
in Section 3.2.2.2.6 of the Draft Supplement on Novel Methods to Address Potential
Confounding. This section outlines how a series of papers dating back to at least 2007
attempt to assess whether there is evidence of unmeasured confounding in the
relationship between long-term PM exposure and mortality. Why, then, are these methods
not described as "causal" for the purposes of interpreting them within the body of
evidence? It would seem to me that the distinction here is that these methods outlined in
Section 3.2.2.2.6 emerge from a different corner of the statistics and epidemiologic
literature, but this strikes me as secondary to the important point that these methodologies
are, much like "causal modeling methods," specifically targeted to reduce uncertainty
around specific types of confounding. My view is that the novel methodologies that have
emerged since the 2019 PM ISA should be discussed in terms of the specific threats to
validity they are designed to address and the specific uncertainties they reduce (or not) in
the body of evi- dence. The fact that these methodologies have been described as "causal
inference" or "causal modeling" in some corners of the literature is important, but, in my
view, secondary to the primary goals of the PM ISA. As written, I believe opportunities
remain in the Draft Supplement to more explicitly distinguish the features of causal
modeling studies that render them important for clarifying and resolving common threats
to validity of epidemiologic studies. In many cases, the potential benefits of the different
types of causal methodologies are never made explicit beyond vague appeals to "the
counterfactual framework." There is an expository literature on the role of causal
inference in air pollution epidemiology consisting of (at least) [Carone et al., 2020,
Dominici and Zigler, 2017, Zigler et al., 2018, Zigler and Dominici, 2014] that may be
useful in framing the discussion in the Draft Supplement. In short, I do not believe it is
nearly as important for EPA to litigate whether any individual study should or should not
be called "causal" as it is for EPA to identify the general quality and conduct of a study
and the specific threats to validity any novel methodology purports to address.
3. Consider alternative labeling of what are currently described as "causal modeling studies"
a. The routine use of the word "causal" in the Draft Supplement must accommodate the
varied scientific and colloquial uses of the word. My opinion is that the most emphasized
and repeated use of the word "causal" in the context of the ISA should be reserved for the
causality deter- minations as laid out in the 2015 ISA Preamble. The fact that the word
"causal" is also used to classify the methods currently described as "causal modeling" is a
reality of the colloquial use of the term and how the statistics and epidemiologic literature
has evolved. In highlighting the importance of novel "causal modeling" for its role in the
A-65
-------
weight of evidence "causality determinations," EPA should consider an alternative
labeling of the specific methodologies to avoid confusion. That is, EPA should be careful
not to elevate the importance of "causal modeling" at the cost of suggesting that the
weight of evidence causality determinations should be abandoned or de-emphasized.
Labeling some studies as "causal modeling" as though they existed in an entirely
different modeling paradigm may serve to perpetuate a dichotomy that studies are
"causal" or otherwise "not causal," which I do not believe is an appropriate organization
of the body of evidence (nor do I believe it was EPA intent to imply so). I suggest that the
Draft Supplement first establish that the methods currently described as "causal
modeling" are grown from a long tradition known in the statistical and epidemiologic
literature as "causal inference," but otherwise refer to them as "novel epidemiologic
analysis strategies," and ingest them into this and future ISAs according to which threats
to validity they are designed to ad- dress (much as the papers described in Section
3.2.2.2.6 are now described). This may help to avoid confusion regarding the role these
methods play in the weight of evidence causality determinations.
Specific Comments
Comments on the Executive Summary
1. The executive summary clearly states that the studies forming the basis of the evaluation within the
supplement include U.S. and Canadian epidemiologic studies that employed "causal modeling methods
or conducted accountability analyses." Absent any further detail on what distinguishes the evidence
from these studies relative to epidemiologic studies that do not use "causal modeling methods," the
importance of this distinction is unclear in the executive summary. I agree with the apparent assertion
that the studies labeled as causal modeling add important evidence to what was available at the time of
the 2019 PM ISA, but detect room in the Executive Summary for more specific description of why these
studies add important information to the body of evidence. Perhaps it would suffice to state that these
studies rely on novel analysis methodologies designed specifically to resolve common threats to validity
in observational epidemiology, and defer details to other sections beyond the executive summary. Page
ES-2 does specify that these studies "used a variety of statistical methods to control for confounding
bias" which I find to be a much more helpful statement alluding to their potential utility over studies
available at the time of the 2019 PM ISA.
2. Given the Draft Supplement's role in the reconsideration of the December 2020 decision to retain the
PM standard, I found it puzzling that the Draft Supplement did not provide any context as to the stated
rationale for the December 2020 decision (as provided, for example, in the 2021 Draft Policy
Assessment). I do not have strong opinion as to what material belongs in the ISA vs. the Policy
Assessment, but I do believe that some of the science that has emerged since the 2019 ISA can
specifically address some of the uncertainties highlighted by the previous Administrator and the
previous CASAC, and would have found a discussion of this helpful in contextualizing the Draft
Supplement.
A-66
-------
Comments on Section 1
1. Section 1.2.2 describes specific criteria for the types of studies considered for evaluation in the
Supplement, including "epidemiologic studies that employed causal modeling methods." A more
specific description of how studies were judged to meet this criterion would help clarify the role
of these studies in the Draft Supplement. To be clear, I have no concerns about which studies are
included under this criterion; later sections of the Supplement provide more detail and leave me
confident that the EPA has correctly identified the importance of the studies judged to satisfy this
criterion. But the criterion as stated is open for interpretation and difficult to define, which may
lead to unnecessary confusion. I suggest some indication of why these particular causal analysis
studies were chosen, which I presume entails more than the mere fact that the study authors
described their methodologies as "causal."
Comments on Section 3
1. In reference to evidence described in the 2019 PM ISA, page 3-21 states that "However, the
body of evidence that supported this causality determination did not include any epidemiologic
studies that conducted accountability analyses or employed causal modeling methods" (with a
similar statement on page 3-46). Absent definition of "causal modeling methods," it is difficult to
tell what exactly is implied by a body of evidence that does not include such studies. Taken at
face value, this could call into question the causal determination in the 2019 ISA, which I do not
believe it should nor do I believe it is EPA intent to imply so.
2. Page3-23 describes, in relation to the Qiu et al. [2020] study, that "several assumptions used by
the authors when applying the IPW methods that are important to recognize." The explicit
description of the series of sensitivity analyses to unmeasured confounding is important and
worth noting. But the following two assumptions - positivity and consistency - are stated without
any comment as to whether they are expected to hold in the Qiu et al. [2020] study or whether
the authors provided any assessment of their validity. This is one example of a possible missed
opportunity for the Draft Supplement to specify what threats to validity might be clarified (or
resolved) with methods such as IPW. Information around the positivity assumption strikes me as
particularly relevant, with many causal inference methods offering a framework to assess threats
to validity that might manifest as a violation of this assumption in a way that may not be
available with more traditional epidemiologic approaches.
3. Line ~9 of page 3-67 attempts to describe what causal modeling methods seek to do. This may
be the closest the Draft Supplement comes to a definition of causal modeling. The general point
about mimicking a randomized experiment through study design and statistical methods is a
good one, but the description of how GPS methods do this is not particularly illuminating, nor
am I sure it is correct. In particular, the statement that "the probability of being exposed is the
same as the probability of being unexposed and the exposure can be considered as 'random'" is
incorrect and/or requires clarification.
4. Section 3.2.1.3 on causal modeling and accountability studies of the PM-mortality relationship is
an example of where there is emphasis on many implementation details of causal modeling
methods (in this case, Wei et al. [2020] and Wei et al. [2021a]) but comparatively little on what
features of these studies contribute to the body of evidence or what threats to validity these
methods are meant to resolve. As written, I believe much of the mechanical details of these
studies could be omitted. As a point of contrast, I found the description of how the methods in
A-67
-------
Schwartz et al. [2018] are designed to protect against certain threats to validity much more
helpful in this context.
5. Section 3.2.2.2 provides enormous amounts of important information on the surprisingly vast
array of recent studies investigating long-term exposures and mortality. The detail on different
cohorts, exposure assignments, and investigation at different levels, is well laid out in relation to
information available in the 2019 PM ISA.
6. Section 3.2.2.2.6 on Novel Methods to Address Potential Confounding focuses on a
methodological area that decomposes the PM-mortality relationship into spatial and temporal
component. Many of these studies are included in Section 11.2.2.4 of the 2019 PM ISA (titled
Studies with Analyses that Inform Causal Inference), but the language of the Draft Supplement
seems to not label them as "causal modeling." I am not convinced this labeling distinction is
strictly necessary, but at a minimum it risks introducing confusion as to what the EPA considers
a "causal modeling" study. More importantly, this indicates to me that the Draft Supplement (and
previous ISAs) may already contain a framework for describing how certain "causal modeling"
studies reduce uncertainties in the body of available evidence since, regardless of whether the
studies cited in this section are labeled "causal modeling," there is clear description of their
particular features for addressing potential confounding.
7. The discussion of the study in Wu et al. [2020], including in the summary information in Table
3-7, includes explicit mention of covariate balance. This is an excellent example of an analysis
feature of this type of methodology that is important in order to assess the ability to adjust for
confounding, and distinguishes this class of methodology from other more traditional (e.g.,
regression) modeling strategies. In this sense, it is an opportunity to emphasize the ability of one
type of "causal modeling" to protect against the threat of confounding and thus add to the
existing body of evidence. I do not believe it was mentioned with respect to any of the other
cited "causal modeling" studies that use propensity scores.
8. Page 3-117 makes reference (when discussing Wei et al. [2021b]) to the assumption that "the
coun- terfactual framework is valid." Without full context, it is a) not clear what exactly it means
for the framework to be valid and b) not clear why this statement wouldn't also apply to all other
methods described as "causal modeling." I am not convinced that this assumption needs to be
stated at all, but if it is it should be accompanied by the appropriate context to discern what it
means.
9. In Section 3.2.2.4, the final paragraph describes the "causal modeling" and accountability
studies, noting that they used different statistical approaches in a causal modeling framework,
and that they collectively provide additional support for the consistent positive associations. I
agree with the statement, but note this as another opportunity where more detail about why these
studies provide new or different evidence would help place these studies in proper context.
References
Marco Carone, Francesca Dominici, and Lianne Sheppard. In pursuit of evidence in air pollution
epidemi- ology: the role of causally driven data science. Epidemiology, 31(1): 1-6, 2020. Publisher:
LWW.
A-68
-------
Francesca Dominici and Corwin Zigler. Best Practices for Gauging Evidence of Causality in Air
Pollution Epidemiology. American Journal of Epidemiology, 2017. doi: 10.1093/aje/kwx307. URL
https://academic.oup.com/aje/article/doi/10.1093/aje/kwx307/4105921/BEST-PRAC
Xinye Qiu, Yaguang Wei, Yan Wang, Qian Di, Tamar Sofer, Yara Abu Awad, and Joel Schwartz. In-
verse probability weighted distributed lag effects of short-term exposure to PM2.5 and ozone on CVD
hospitalizations in New England Medicare participants - Exploring the causal effects. Environmental
Research, 182:109095, March 2020. ISSN 1096-0953. doi: 10.1016/j.envres.2019.109095.
Joel Schwartz, Kelvin Fong, and Antonella Zanobetti. A National Multicity Analy- sis of the Causal
Effect of Local Pollution, N02, and PM2.5 on Mortality. Envi- ronmental Health Perspectives,
126(8):087004, 2018. doi: 10.1289/EHP2732. URL https://ehp.niehs.nih.gov/doi/10.1289/ehp2732.
Publisher: Environmental Health Perspectives.
Yaguang Wei, Yan Wang, Xiao Wu, Qian Di, Liuhua Shi, Petros Koutrakis, Antonella Zanobetti,
Francesca Dominici, and Joel D Schwartz. Causal Effects of Air Pollution on Mortality Rate in
Massachusetts. American Journal of Epidemiology, 189(11): 1316-1323, November 2020. ISSN 0002-
9262. doi: 10.1093/aje/kwaa098. URL https://doi.org/10.1093/aje/kwaa098.
Yaguang Wei, Brent Coull, Petros Koutrakis, Jiabei Yang, Longxiang Li, Antonella Zanobetti, and Joel
Schwartz. Assessing additive effects of air pollutants on mortality rate in Massachusetts. Environmen-
tal Health, 20(1): 19, February 2021a. ISSN 1476-069X. doi: 10.1186/sl2940-021-00704-3. URL
https ://doi. org/10.1186/s 12940-021 -00704-3.
Yaguang Wei, Mahdieh Danesh Yazdi, Qian Di, Weeberb J. Requia, Francesca Dominici, Antonella
Zanobetti, and Joel Schwartz. Emulating causal dose-response relations between air pollutants and mor-
tality in the Medicare population. Environmental Health, 20(1):53, May 2021b. ISSN 1476-069X. doi:
10.1186/s 12940-021 -00742-x. URL https://doi.org/10.1186/s 12940-021 -00742-x.
X. Wu, D. Braun, J. Schwartz, M. A. Kioumourtzoglou, and F. Dominici. Evalu- ating the impact of
long-term exposure to fine particulate matter on mortality among the elderly. Science Advances, July
2020. doi: 10.1126/sciadv.aba5692. URL https://www.science.org/doi/abs/10.1126/sciadv.aba5692.
Publisher: American Association for the Advancement of Science.
Corwin M. Zigler, Christine Choirat, and Francesca Dominici. Impact of National Ambient Air Quality
Standards Nonattainment Designations on Particulate Pollution and Health. Epidemiology (Cambridge,
Mass.), 29(2): 165-174, 2018. ISSN 1531-5487. doi: 10.1097/EDE.0000000000000777.
Corwin Matthew Zigler and Francesca Dominici. Point: Clarifying Policy Evidence With Potential-
Outcomes Thinking—Beyond Exposure-Response Estimation in Air Pollution Epidemiology. Ameri-
can Journal of Epidemiology, 180(12): 1133-1140, December 2014. ISSN 0002-9262, 1476-6256. doi:
10.1093/aje/kwu263. URL http://aje.oxfordjournals.org/content/180/12/1133.
A-69
------- |