\ UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
o WASHINGTON, D C 20460
JUN 18 2010
OFRCE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
SUBJECT: Transmittal of Science Advisory Board Report
FROM: Vanessa T. Vu ^^-*=-^-<- ^—
Director, Science Advisory Board Staff Office (HOOF)
TO: Karen Sheffer
EPA Headquarters Library Repository (3404T)
This is to advise you that the Science Advisory Board, Clean Air Scientific Advisory
Committee, Carbon Monoxide Review Panel, issued a report numbered EPA-CASAC-10-012,
Review of the Risk and Exposure Assessment for the Review of the Carbon Monoxide Primary
National Ambient Air Quality Standards (NAAQS): Second External Review Draft, dated
May 19,2010.
Two copies of the report are attached and a third copy has been sent electronically to
the attention of Ms. Jeannie Turner at turner.ieannie@epa.gov. The report is also available in
electronic format on the Science Advisory Board's web site at hthr/Avww.epa.gov/sab.
If you have any questions regarding this report, please contact the Designated Federal
Officer, Dr. Holly Stallworth directly at (202) 343-9867.
Attachments (2)
Internet Address (URL) • httpV/www epa gov
Recycled/Recyclable • Printed with Vegetable Oil Based Inks on 100% Postconsumer. Process Chlorine Free Recycled Paper
-------
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON D.C. 20460
OFFICE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
May 19,2010
EPA-CASAC-10-012
The Honorable Lisa P. Jackson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, N.W.
Washington, D.C. 20460
Subject: Review of the Risk and Exposure Assessment for the Review of the Carbon
Monoxide Primary National Ambient Air Quality Standards (NAAQS): Second
External Review Draft
Dear Administrator Jackson:
The Clean Air Scientific Advisory Committee (CASAC or Committee) Carbon
Monoxide (CO) NAAQS Review Panel met on March 22-23, 2010 to review EPA's Risk and
Exposure Assessment for the Review of the Carbon Monoxide Primary National Ambient Air
Quality Standards (NAAQS): Second External Review Draft. The Chartered CASAC held a
public teleconference on April 1 9, 201 0, to review and approve the report. This letter provides
CASAC's overall comments and evaluation. We highlight the most important issues which need
to be addressed as the second draft Risk and Exposure Assessment (REA) is finalized.
The Panel expressed appreciation to EPA staff for the major improvements made in the
second draft of the REA. The changes are responsive to the suggestions and concerns expressed
by the Panel in its review of the first draft. Nonetheless, CASAC offers several suggestions and
concerns to be considered as the second draft undergoes final revisions.
The Panel encourages a clearer distinction between the levels set for the CO NAAQS and
the concentrations at which exposures are currently experienced throughout the country. Current
levels of CO are far lower than historic levels and the risk for health effects associated with these
current levels may be minimal or difficult to quantify with certainty. However, the degree of
protection afforded to susceptible populations by the current NAAQS still needs to be considered
by EPA. A greater degree of protection may be warranted.
As mentioned in its review of the first draft, the Panel felt strongly that the focus of the
REA, and the associated Policy Assessment document, should be broader than cardiac ischemia
(coronary artery disease or CAD). The Panel recognizes that compelling evidence comes from
clinical studies demonstrating a relationship between elevated levels of carboxyhemoglobin
-------
(COHb) and a reduced time to the onset of angina. These studies have been at the center of the
evidence used to set the NAAQS for CO. However, there is increasing evidence that CO
increases the frequency and severity of congestive heart failure and enhances susceptibility to
arrhythmias. Consequently, we recommend that a broad set of health outcomes be considered,
beyond cardiac ischemia. The susceptible populations might also include those with pulmonary
disease and the fetus. The Allred et al. (1989) study should be more completely presented.
While a reduction in time to onset of angina is an important and easily interpretable clinical
outcome, this response is subjective. In contrast, the outcome of ST segment depression, as
assessed by a blinded cardiologist, is an objective measure of myocardial ischemia and should
receive greater consideration. Moreover, ST segment depression has been validated, both as an
outcome of inadequate delivery of oxygen to the myocardium and as a risk factor for more
frequent arrhythmias.
The Panel recommends greater clarity regarding the major contributors to COHb:
ambient outdoor exposures, endogenous production of CO within the body, and finally indoor
sources of CO from home cooking, heating, and passive smoking. The relative importance of
these contributors to COHb must be more clearly delineated. The REA should address how
these multiple sources are used in modeling and contribute to variability and uncertainty in
model results.
We are concerned about two aspects of the adequacy of the current CO monitoring
network. First, more sensitive and precise monitors need to be deployed to measure levels that
are less than or equal to 1 ppm. Such monitors are needed to validate CO exposure models.
Second, the approach for siting monitors needs greater consideration. More extensive coverage
may be warranted for areas where concentrations may be more elevated, such as near roadway
locations. The Panel found that in some instances current networks underestimated carbon
monoxide levels near roadways. Such underestimation is a critical issue since populations with
low social economic status (SES) are often overrepresented in those areas. People with SES are
more likely to smoke, a substantial source of CO. In addition, African Americans have a higher
incidence of sickle cell disease, which affects oxygen transport.
In regard to the quantitative risk assessment, the Panel recommends greater clarity in
describing the model that was used, along with information available about its validity. The
profile of COHb in time with varying CO exposures is complex since loading (increased COHb
levels) is much more rapid than unloading of COHb levels as ambient CO levels drop. In an
analysis that acknowledges multiple sources, it is essential to emphasize the increment which is
attributable to ambient CO.
-------
The CASAC and Panel membership are listed in Enclosure A. The Panel's responses to
EPA's charge questions are presented in Enclosure B. Finally, Enclosure C is a compilation of
individual panel member comments. We look forward to the Agency's response and the
successful completion of the CO "NAAQS review.
Sincerely,
/Signed/ /Signed/
Dr. Joseph D. Brain, Chair Dr. Jonathan M. Samet, Chair
CASAC CO Review Panel 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. 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 necessarily 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 of
commercial products does not constitute a recommendation for use. CASAC reports are posted on
the EPA website at http://www.epa.gov/CASAC.
-------
Enclosure A
Rosters of the CASAC CO Panel and CASAC
U.S. Environmental Protection Agency
Clean Air Scientific Advisory Committee
Carbon Monoxide Review Panel
CHAIR
Dr. Joseph Brain, Cecil K. and Philip Drinker Professor of Environmental Physiology,
Department of Environmental Health, Harvard School of Public Health, Harvard University,
Boston, MA
MEMBERS
Dr. Paul Blanc, Professor and Chief, Department of Medicine, Endowed Chair, Occupational
and Environmental Medicine, Division of Occupational and Environmental Medicine, University
of California San Francisco, San Francisco, CA
Dr. Thomas Dahms, Professor and Director, Anesthesiology Research, School of Medicine, St.
Louis University, St. Louis, MO
Dr. Russell R. Dickerson, Professor and Chair, Department of Meteorology, University of
Maryland, College Park, MD
Dr. Laurence Fechter, Senior Career Research Scientist, Department of Veterans Affairs, Loma
Linda VA Medical Center, Loma Linda , CA
Dr. H. Christopher Frey, Professor, Department of Civil, Construction and Environmental
Engineering, College of Engineering, North Carolina State University, Raleigh, NC
Dr. Milan Hazucha, Professor, Department of Medicine. Center for Environmental Medicine,
Asthma and Lung Biology, University of North Carolina - Chapel Hill, Chapel Hill.NC
Dr. Joel Kaufman, Director, Occupational and Environmental Medicine Program, University of
Washington, Seattle, WA
Dr. Michael T. Kleinman, Professor, Department of Medicine, Division of Occupational and
Environmental Medicine, University of California, Irvine, Irvine, CA
Dr. Francine Laden, Professor, Channing Laboratory, Harvard University, Boston, MA
Dr. Arthur Penn, Professor LSU School of Veterinary Medicine, Department of Comparative
Biomedical Sciences, Louisiana State University, Baton Rouge, LA
-------
Dr. Beate Ritz, Professor. Epidemiology, School of Public Health, University of California at
Los Angeles, Los Angeles, CA
Dr. Paul Roberts. Executive Vice President, Sonoma Technology, Inc., Petaluma, CA
Dr. Armistead (Ted) Russell, Professor. Department of Civil and Environmental Engineering,
Georgia Institute of Technology, Atlanta, GA
Dr. Anne Sweeney, Professor of Epidemiology. Department of Epidemiology and Biostatistics,
School of Rural Public Health, Texas A&M Health Science Center, College Station, TX
Dr. Stephen R. Thorn, Professor, Institute for Environmental Medicine, University of
Pennsylvania, Philadelphia, PA
SCIENCE ADVISORY BOARD STAFF
Ms. Kyndall Barry, Designated Federal Officer, 1200 Pennsylvania Avenue, NW, Washington,
DC
-------
U.S. Environmental Protection Agency
Clean Air Scientific Advisory Committee
CHAIR
Dr. Jonathan M. Samet, Professor and Flora L. Thornton Chair, Department of Preventive
Medicine, University of Southern California, Los Angeles, CA
MEMBERS
Dr. Joseph Brain, Cecil K. and Philip Drinker Professor of Environmental Physiology,
Department of Environmental Health, Harvard School of Public Health, Harvard University,
Boston, MA
Dr. H. Christopher Frey, Professor, Department of Civil, Construction and Environmental
Engineering, College of Engineering, North Carolina State University, Raleigh, NC
Dr. Donna Kenski, Data Analysis Director, Lake Michigan Air Directors Consortium,
Rosemont, 1L
Dr. Armistead (Ted) Russell, Professor, Department of Civil and Environmental Engineering,
Georgia Institute of Technology, Atlanta, GA
Dr. Helen Suh, Associate Professor, Department of Environmental Health, School of Public
Health, Harvard University, Boston, MA
Dr. Kathleen Weathers, Senior Scientist, Gary Institute of Ecosystem Studies, Millbrook, NY
SCIENCE ADVISORY BOARD STAFF
Dr. Holly Stallworth, Designated Federal Officer, Washington, DC
-------
Enclosure B
CASAC's Consensus Responses to EPA's Charge Questions
1 Does the Panel find the summary of CO exposure and discussion of ambient CO sources,
exposures, dose, health effects and risk characterization approach to be technically sound,
clearly communicated, and appropriately characterized?
In general, the Panel found Chapter 2 to be well organized, technically sound and a good
conceptual overview of the REA. The chapter provides a sound rationale for the decision to use
COHb level as the internal dose metric for assessing exposure to ambient levels of CO and for
characterizing the potential for health risks in the population of persons with coronary artery
disease (CAD). While some Panel members supported a cautious approach to using
epidemiological data in the risk assessment, overall there were concerns that the current
presentation under-emphasized the epidemiologic findings. Even if the epidemiological data are
not used in the risk assessment, it is important to incorporate them into the discussion of risk.
The epidemiological evidence provides information on non-hypoxia relevant mechanisms and
chronic outcomes that cannot be addressed by relying on COHb levels alone.
The Panel continued to be concerned with EPA's designation of the at-risk population. The
choice of modeling risk for the CAD population needs to be further justified and/or expanded to
include other susceptible populations. Again, the findings of epidemiological studies suggest
several groups to be considered, especially the broader category of cardiovascular disease
(CVD). Since this chapter serves as the introduction to the REA, it should be edited as the
Panel's recommendations are incorporated into subsequent chapters.
2 Does the Panel find the considerations of current ambient carbon monoxide monitoring
data, including specifically the data for the monitors included in this draft of the assessment,
and the discussion of the extent to which near roadway concentrations are represented to be
technically sound, clearly communicated, and appropriately characterized?
The discussion is technically sound, clearly communicated, and appropriately characterized. The
treatment of the CO monitoring data, the description of the extent that these monitors represent
near-roadway concentrations, and the data used in this iteration of the assessment are improved
over the first external draft REA. The current monitoring network does not represent near-
roadway concentrations very accurately, which is now well-documented in the REA. The
increased discussion of NCore and measurement characteristics is useful and appropriately
placed.
Should there be additional monitoring for indoor and in-vehicle exposures? Representative
monitoring to evaluate emissions inventories or models may be different from monitoring to
assess exposure.
3 In recognition ofCASAC comments of first draft REA, this draft REA is expanded from the
previous assessment in a number of ways (summarized in section I 3 of the draft document)
The assessment study areas are in the Denver and Los Angeles study areas We are
-------
interested in eliciting the views of the Panel on the usefulness of this approach in informing
our review the NAAQSfor CO What are the Panel members' views on the following aspects
in which the assessment has been expanded from the previous draft7
a An important change of this assessment from the first draft is the expansion of each of the
modeling domains to mclude a greater number of ambient monitors used as input to
APEX. Additionally, this draft assessment employs an algorithm that adjusts for temporal
and spatial heterogeneity in ambient concentrations across each study areas.
An important improvement in this assessment from the first draft is the expansion of each
modeling domain to include a greater number of ambient monitors for input to APEX.
Additionally, this draft assessment employs an algorithm that adjusts for temporal and spatial
heterogeneity in ambient concentrations across each study area. While use of a larger number of
ambient air monitors may have improved exposure assessment, we are not convinced that spatial
heterogeneity driven by proximity to sources can be adequately captured by the current ambient
monitoring network. In fact, exposures outdoor, in homes and in workplaces near roadways
might be underestimated.
It would be helpful to describe with greater clarity how data from overlapping districts, zones
and areas were treated for input into the APEX model. What approach was used to avoid
duplication of input data from overlapping zones? For Los Angeles, was one of the areas
designated as a dominant source or was each area considered separately in the assessment?
These are all questions that should be addressed in future analyses.
In generating simulated individuals, demographic variables should include socioeconomic status
(SES) and race if possible. These variables will impact other APEX modules, particularly
COHb. If not, future data collection efforts should provide sufficient coverage.
We find the tables of exposure values vs. estimated number and percentage of CHD persons
affected at specific CO concentrations to be very instructive. Tables showing estimated COHb
levels vs. number of people with CHD and persons/days are similarly instructive. Additional
calculations and tabulation of endogenous COHb level using APEX and additional plots
reflecting contribution of endogenous COHb to total COHb are illustrative.
b The current draft assessment also include an increase in the number of
microenvironmenls modeled over that in the first draft (from two to eight) and improved
the representation of variability in estimated microenvironmental concentrations,
including in-vehicles
The Panel has no major issues with this approach.
c This draft assessment has implemented the mass-balance model for estimating
concentrations in indoor microenvironments
We consider the selection of the mass-balance model for indoor air to be appropriate.
-------
4. Does the Panel view the results of the draft exposure analyses to be technically
sound,clearly communicated, and appropriately characterized7
The Panel questions EPA's assumption that it has captured spatial heterogeneity in homes and
workplaces near busy roadways. We agree with the summary of findings (p. A-5) that the
current physiology file data is obsolete and may even be incorrect for some variables. While
some variables were already updated, others such as race. SES, total hemoglobin (THb) and
DLco should be either added or replaced in the input module. With these qualifications, the
Panel answers affirmatively to question 4.
5 Does the Panel find the derivation and presentation of the modeling approach as a whole
(chapters 4 and 5) to be technically sound, clearly communicated, and appropriately
characterized?
The data added to the ambient source inputs for the exposure modeling reflect commendable
responsiveness to the feedback we provided in the previous review of the initial REA draft. The
modeling appears to be technically sound.
The subject matter is complex and highly challenging to communicate clearly, particularly the
material in Chapter 4. The presentation of this material seems to be aimed at the exposure
modeling community, which makes it difficult for others to readily grasp. Nonetheless, the
detailed description of the APEX model provides a helpful snapshot of the extensive nature of
the model. The derivation and presentation of the modeling approach as a whole are well
presented. Moreover, the application of the "CHAD" database in modeling the physiological
changes of simulated residents during their daily lives appears to be appropriately handled. This
approach has been applied and vetted for other regulated air pollutants. Nonetheless, the
coupling of the non-linear CFK with the CHAD in the APEX model would be more convincing
if this approach had been validated with actual field study measurements of delivered dose. We
understand, however, that such validation is not possible on practical grounds. Moreover,
previous approaches are less sophisticated from a modeling point of view and also lack field
validation for the same reasons of feasibility.
Appendix C contains some helpful illustrations of the variability in time spent for a given
individual in locations/activities throughout the year. It might be helpful, additionally, to have
similar illustrations of:
1. Estimated %COHb levels for an individual during a day with exposures that were near
current criteria (maximum allowable) levels of atmospheric CO, and
2. An illustration of the potential variability in peak levels of %COHb throughout the year.
Indeed, inclusion of such illustrative scenarios in the text, rather than in the Appendix, should be
considered. This material could assist the reader in understanding the variability in the modeled
levels of CO exposure.
6 Does the Panel find the derivation and presentation of the COHb estimates (Chapters 5
and 6) to be technically sound, clearly communicated, and appropriately characterized7
10
-------
The derivation and presentation of the %COHb estimates are clearly communicated and
characterized. However, the final %COHb estimates, as presented in the REA, are potentially
problematic. This is related, in part, to their derivation and the basic assumptions made in
arriving at these values, not the computational operations of the APEX system. Modeling
%COHb, as a biological marker of ambient CO exposure, presents particular challenges
stemming from the various sources of exposure, each of which can potentially contribute to the
final %COHb estimate. These sources include: endogenous production; ambient air pollution;
and indoor air sources not accounted for by ambient pollution tracking indoors (e.g., combustion
byproducts from home heating or cooking and secondhand smoking-associated CO). Extensive
modeling in the REA appropriately deals with scenarios of ambient CO contributions to the
indoor exposure environments. It was only well into Chapter 6 that modeling including "Internal
Sources" of COHb was introduced. The Panel further takes issue with the confusing word
choice since this does not mean internal in the sense of endogenous metabolism. As shown in
Tables 6-15 and 6-16, excluding these indoor sources, modeled %COHb values (i.e., from
ambient exposures levels) are similar between the 2000 model that also provides the indoor
("internal") estimates and the current APEX model. Key is that inclusion of the indoor sources
of exposure drives up exposure such that five percent of the population hits a 3% COHb level
and roughly two percent of the population reaches a 4.0% COHb level.
The Panel appreciated the additional attention given to endogenous CO production (again, not
the "Internal" metric above). There was concern that this discussion could lead to confusion,
because the modeled levels of endogenous %COHb are quite a bit lower than population means
for non-smokers. This is because the actual data for the non-smoking population reflect the sum
of ambient exposure, indoor exposure, and endogenous CO production. It might be useful to
clarify these distinctions explicitly. Beyond issues of presentation, the modeled distribution of
endogenous %COHb values seems too narrow. It appears to be based on "normal" healthy
population estimates of endogenous production (albeit with a variety of activity levels).
Literature demonstrating elevated endogenous %COHb values in certain subpopulations (for
example, in persons with sickle cell disease) may be difficult to account for in these models, but
an attempt to address them is warranted. At a minimum, the REA should directly acknowledge
this limitation of the model estimates.
Further confusion may be introduced through the random subset estimations, given that the
central tendency (mean) of this random sub-sample seems to differ from the larger modeling
estimate (apparently a chance observation). The narrower distribution is produced by the limited
intra-person variation since most of the data points are derived from multiple runs on a relatively
small subset - (this should not explain the shift in the mean). It may even be relevant to
acknowledge that certain groups at risk for higher endogenous production systematically may be
more likely to have higher ambient scenarios (e.g., persons with sickle cell disease, low SES, and
those living/working near a major roadway). The lack of transparency in the endogenous
production model as applied may also contribute to confusion. The description of what
endogenous rates were used in the model is unclear and the information in Table B-3 on page B-
20 is poorly labeled. Despite these limitations, the material on endogenous production of CO
and its contribution to the overall %COHb in combination with ambient levels of CO is very
informative and indeed necessary. In summary, the Panel is concerned that there is no modeling
11
-------
of %COHb that covers indoor non- ambient sources and endogenous CO production, as well as
ambient exposure.
The staff's presentation at the meeting included additional analyses of the contribution of
ambient CO to the COHb% levels. The Panel found the analyses presented during the meeting
represented an improvement of the discussions in the REA itself. Although the data were
apparently computationally intensive to develop, some Panel members believe that the additional
analyses could be an extremely useful avenue for further development of the standard. The
specific increase in %COHb over the subjects' pre-exposure or filtered air exposure was the
focus of the influential Allred et al. papers and not on the magnitude of the final %COHb. It
may well be that the ambient-attributable increment in %COHb is the most directly analogous
dose for consideration in a risk assessment. Using incremental %COHb as the metric for
ambient-attributable dose could simplify the risk assessment because the issue of "overlap" with
the endogenous range of final COHb% (let alone indoor non ambient contributions) would be
parsed-out. This treatment could still delineate the other sources of exposure as above-noted and
would address the Panel's concerns with regard to the continuity of the CO exposure-response in
experimental studies. The presentation of incremental exposure data and estimated %COHb
combined and separately would allow policy makers to consider ambient exposure alone, as well
as in the context of other sources of exposure.
7 In the Panel's view, to what extent does the modified assessment approach employed in
this second draft assessment provide results that meaningfully inform the EPA's
consideration of the public health implications of the current standards
The modified assessment approach has two key components: estimated exposure and estimated
at-risk (susceptible) population. In terms of exposure, it should also be noted that the additional
information included in the current modification does not relieve uncertainties from the use of a
relatively limited data set (i.e., two case studies in Denver and Los Angeles), even with the
additional monitoring data. The REA could be improved by showing the impact, or lack of
impact, on dose variability that resulted from the inclusion of data from more monitoring sites.
Data resolution from the two case studies and inclusion of data from more monitoring sites will
be particularly relevant for national extrapolations. The Panel has provided additional
suggestions to strengthen the exposure component in response to other charge questions. Such
improvements could serve to better inform the EPA's consideration of the public health
implications of the current CO standard.
There are serious potential data uncertainties in the estimates of the "at risk" population, many of
which might lead to a systematic underestimation of the public health impact of CO exposure.
The REA continues to rely singularly on the National Health Interview Survey (NHIS) data to
provide a population estimate of persons at risk. The "at risk" population has been narrowly
defined as self-reported coronary artery disease, which was the Panel's primary point of
contention and critique of the first draft REA. The revision now includes an estimate of
"undiagnosed" disease that represents approximately 40% in addition to the base population.
This is an important acknowledgment of one aspect of systematic underestimation, although the
American Heart Association source of the mathematical value used is far from convincing. It is
12
-------
recommended that EPA incorporate a female>male differential to address the probable sex-based
gap in CAD diagnosis.
Data from NHANES and the Behavioral Risk Survey are easily accessible and will generally
support the NHlS-based, restricted subset of susceptible persons based on a CAD definition.
However, the narrowly-defined CAD prevalence estimate, to the exclusion of all others with
cardiac disease, misinterprets the ISA (particularly Tables 4-9, 5-10, 5-11 and Figures 5-5 though
5-7). Revisiting the NH1S, the prevalence rates for "all heart disease" are considerably greater
than those of narrowly defined CAD. It is very likely that most, if not all, of these persons are at
increased risk for adverse cardiovascular effects from ambient CO in ways that cannot be
distinguished epidemiologically from the CAD subset. It is certainly appropriate for the REA to
present estimates, as indeed it did, using a narrow CAD definition of "at risk" to inform an EPA
public health assessment. However, this approach alone is not sufficient. Much effort is spent
on multiple scenarios of exposure, while falling short in the critical area of defining alternate
measures of the vulnerable population. As a consequence of the approach suggested above, a
population more broadly defined with cardiovascular disease is likely to overlap to a meaningful
extent with adults with chronic obstructive lung disease (through shared risk factors), which may
be another at risk group for adverse CO exposure effects. This is not a determining factor,
however, in the rationale for applying a more broadly-defined cardiac disease definition in
modeling the at-risk population.
8 What are the views of the Panel regarding the adequacy of the assessment of uncertainty
and variability? To what extent have sources of uncertainty been identified and the
implications for the risk characterization been characterized? To what extent has
variability adequately described and represented7
In general, the incorporation of more monitors in each area, more microenvironments, and
variability in various variables within APEX better addresses variability and thus general
uncertainty in exposure and dose. However, the use of the power 0.621 in equations 4-11 and 4-
22, reduces the CO concentration at an outdoor location, relative to the nearest central monitor,
and thus possibly reduces the number of occurrences of the highest CO concentrations.
In addition, all three contributors to COHb (ambient, endogenous production, and indoor
sources, excluding smoking) should be considered in modeling and as contributing to variability
and uncertainty in model results.
13
-------
Enclosure C
Review Comments from the CASAC CO Panel on the Second Draft Risk and
Exposure Assessment to Support the Review of the Carbon Monoxide Primary
NAAQS
Comments received;
Dr. Paul Blanc 15
Dr. Thomas Dahms 18
Dr. Russell Dickerson 21
Dr. Milan Hazucha 23
Dr. Francine Laden 29
Dr. Arthur Penn 30
Dr. BeateRitz 31
Dr. Paul T. Roberts 32
Dr. Anne Sweeney 34
Dr. Stephen Thorn 35
14
-------
Dr. Paul Blanc
5 Does the Panel find the derivation and presentation of the modeling approach as a whole
(Chapters 4 and 5) to be technically sound, clearly communicated, and appropriately
characterized7
Insofar as the increase inputs to the ambient source inputs to the models, this revision was quite
responsive to the input that it received in review of its initial draft and is to be commended. This
aspect of the approach appears to be technically sounds. This subject matter is complex and
highly challenging to communicate clearly, in particular the material in Chapter 4..
6 Does the Panel find the derivation and presentation of the COHb estimates (Chapters 5
and 6) to be technically sound, clearly communicated, and appropriately characterized?
The final COHb estimates in 6 are problem-ridden. This is related in part to the derivation
(including under this the basic assumptions made, not simply the mathematical operations of the
APEX system). The presentation magnifies certain issues by potentially obscuring points
presumed to be manifest but that may be to be more explicit.
Modeling COHb, as a biological marker of ambient air pollution exposure, presents a particular
challenge because there are 4 principal domains of exposure, each which can potentially
contribute to the end concentration measured. These 4 domains are: endogenous production;
ambient air pollution; active cigarette smoking; and supplemental sources of carbon monoxide
beyond these three. This fourth domain, for most persons, is drive by indoor air exposure to CO
from combustion byproducts from home heating or cooking and secondhand smoking-associated
CO [although other sources of exposure within this domain may be important for population
subsets, e.g., occupationally-related CO exposure].
Chapter 5, in relation to CO exposure modeling leading to COHb is focused entirely on the
domain of ambient CO, although extensive modeling deals with scenarios of contributions of
ambient CO to indoor exposure environments. This can be a bit confusing because internal
combustion engine contributions to exposure in certain indoor scenarios are essentially point
sources (service station and auto repair GM 2.97 [PPMs although not labeled]). Of note, another
indoor facility group includes (Manufacturing facility) and is rather low-GM 0.089. One
assumes fork lifts or truck deliveries not considered. The salient point however is: these
scenarios exclude other likely concomitant sources of exposure.
It is not until well into Chapter 6 that modeling that includes the domain of "Internal Sources" of
COHb is introduced. The term Internal as used here is unfortunate, since it actually is intended to
mean "indoor- not from ambient sources" i.e., what I refer to as the 4th domain above. Iflhis
also includes endogenous production it is by no means clear. As shown in Tables 6-15 and 6-16,
excluding these indoor sources these estimated exposures levels are similar between a 2000
model and the current APEX model - but inclusion of this critical source of exposure drives up
exposure such that 5% of the population hits a 3% COHb level and roughly 2% a 4.0% COHb
-------
level. Remarkably there is no simulation or other estimation combining current indoor sources
with the ambient estimates.
Despite the short-shrift give the indoor exposure issue, a great deal of attention is then given to
endogenous CO production. This, of course, is not "Internal" as used above, but rather refers to
"normal" metabolic production of CO. This section is likely to lead to some confusion because,
quite appropriately, the modeled levels of COHb% (population mean 0.255%) are quite a bit
lower than population means for non-smokers. The reason of course if that observed non-
smoking population data reflect the sum of domains l,2,and 4, no simply endogenous production
(thus the baseline samples in the Allred study, etc). This is a point of blurred presentation, not
modeling. The modeling itself, however, could benefit from more transparency. The reader is
told that a certain number of studies informed the metabolic parameters as detailed in B but in
fact it is hard to tease-out which references there are the ones in question. These seem to be fairly
dated (Coburn's work from the 1960s) and this may be the best there is. Nonetheless, more
recent literature exploring moderately elevated COHb in certain conditions (for example, sickle
cell disease) suggests that the model specifications and simulations yielding a maximum
endogenous value of 1.54 is not likely to be reflective of population variability. This may be
driven by simulating 59 individuals only, albeit with 8,760 hours of modeled observation. To
capture endogenous variability, within person hour to hour activity inputs is not nearly as
important as between person variability. Moreover, it is likely that certain at risk groups may be
more likely to have ambient scenarios (and indoor scenarios) of higher exposure: eg. a low
income resident of south-central LA dwelling near a freeway with sickle cell disease (and
heating the home with a gas stove in the winter).
Integrating the comments above, there is no apparent modeling of COHb that includes variable
indoor not ambient sources (exclusive of direct cigarette smoking) + endogenous CO production
(anticipation a possible bimodal distribution, with certain disease states contributing a sub-group
of outliers) and ambient exposure.
7 In the Panel's view, to what extent does the modified assessment approach employed in
this second draft assessment provide results that meaningfully inform the EPA 's
consideration of the public health implications of the current standards
The modified assessment approach has two key components: estimated exposure and the
estimated at risk (susceptible) population. All of the comments above (charge questions 5 and 6)
relate to the exposure assessment. To the extent that the exposure issues raised above can be
addressed, this would serve to better inform the EPA's consideration of the public health
implications of the current CO standard. Even as is, this portion of the Risk and Exposure
Assessment is not fundamentally flawed.
Insofar as the at risk population estimates are concerned, the document as currently constituted is
prone to several sources of data uncertainties, all of which would tend to systematically
underestimate the public health impact of exposure. The document continues to rely on National
Health Interview Survey (NHIS) to provide a population estimate of persons at risk, narrowly
defined as self reported coronary artery disease. This was a matter of focused critique of the first
draft document. The revision now includes an estimate of "undiagnosed" disease (approximately
16
-------
40% in addition to the base population). Conceptually this is an important acknowledgment of
the one aspect of systematic underestimation, although the American Heart Association source of
the mathematical value used is far from convincing (and whatever value is used should
incorporate a female>male differential given the clear sex-based gap in diagnosis. In addition, in
terms of the restricted subset of susceptible persons based on CAD, supportive prevalence data
based on NHANES and the Behavioral Risk Survey are easily accessible and will generally
support the NHIS-based values. More fundamentally, the narrowly defined CAD prevalence
estimate, to the exclusion of all others with cardiac disease misreads and misinterprets the ISA
particularly Tables 4-9, 5-10, 5-11 and Figures 5-5 though 5-7. Revisiting the NHTS, the
prevalence rates for "all heart disease" are considerably more than for narrowly defined CAD. It
is very likely that most if not all of these are at risk as from adverse cardiac effects from ambient
CO in ways that cannot be distinguished epidemiologically from the CAD subset (this includes
in term of RR). It is certainly appropriate for the REA to present estimates, as it did, using a
narrow definition of "at risk"; to inform an EPA public health assessment, this approach alone is
unacceptable. It is ironic that so much effort is spent on multiple scenarios of exposure, while
falling short in the critical area of the key vulnerable population.
17
-------
Dr. Thomas Dahms
5. Does the Panel find the derivation and presentation of the modeling approach as a whole
(Chapters 4 and 5) to be technically sound, clearly communicated, and appropriately
characterized7
The detailed description of the apex model provides a helpful snapshot of the extensive nature of
the model. The application of the CHAD database in modeling the physiological changes of
simulated residents during their daily lives seems to be appropriately handled. This approach has
been applied and vetted for other regulated air pollutants. The presentation of the material seems
to be aimed at the exposure modeling community which makes it difficult to assess for someone
not in the modeling community.
Appendix C contains some helpful illustrations of the variability in time spent for an individual
in locations/activities during a year. It would be helpful to have a similar illustration of
estimated %COHb levels of %COHb in an individual during a day with near criteria levels of
atmospheric CO and also an illustration of the variability in peak levels of %COHb thoughout
the year. Additions of these illustrations to the text rather than in the Appendix should be
considered. It would assist the reader in understanding the variability in the modeled levels of
dose of CO.
The coupling of the non-linear CFK with the CHAD in the APEX model would be more
convincing if there were references to validation of this model in studies where measurements of
dose were made. Without documentation of such validation, the reader is expected to accept this
model based on years of improvements over other models.
Since the controlled human exposure data is a major factor in setting policy, it would be helpful
to see how well the exposure model predicts the measured CO dose in these experiments.
Although these exposures are for only 1 hour with subjects at rest, it would provide a means of
validation of the exposure/dose modeling used for the general population. It might also provide a
means of comparison of the various controlled human exposures.
6. Does the Panel find the derivation and presentation of the COHb estimates (chapters 5
and 6) to be technically sound, clearly communicated, and appropriately characterized7
The presentation of how the model arrives at estimates of %COHb in the population of Denver
and Los Angeles is clear in that it follows a natural progression in this field over the past 40
years, it is clearly communicated and the improvements in the modeling over the years is clearly
presented and is very rational. There are concerns with the output of the model that leads to a
level of uncertainty that could be somewhat reduced as described below.
The primary goal of this section would be to determine whether or not the estimated levels of
%COHb (from the model) using the current as is data are lower or higher than those estimated
levels of %COHb using the current standards. It is presumed that the exposure to the current
standards for CO results in a small but acceptable number of at risk persons for a given %COHb
18
-------
benchmark. If the current as is data results in a smaller number of persons at that benchmark,
there would be little pressure to change the standards. Therefore the issue that needs to be
discussed/defined is : what is a small but acceptable number of persons who can be exposed at
an acceptable benchmark. This discussion is absent.
It is not clear what background in modeling minutia the reader is expected to have~to clearly
understand and analyze the details of the material presented. The crux of the matter is that for
carbon monoxide there is an agreed upon dose metric, %COHb, that can be used to evaluate
exposure. This is not quite the case for the other regulated pollutants. There is no mention of any
attempt to validate the model being used under any circumstances. No matter how sophisticated
the model, without validation it is just a model with all of the attendant uncertainties.
The presentation of the information in section 6.4 on the influence of endogenous rates of CO
production on dose estimates is potentially a problem. The description of what endogenous rates
were used in the model is unclear and the information on Table B-3 on page B-20 is very poorly
labeled. The values used in this portion of the model need to be clearly documented and justified.
This becomes and issue because this modeling data is used to justify not including 1.0 %COHb
in the Policy Assessment document. The modeling of all of the parameters that impact baseline
%COHb into 'endogenous rates of CO production' as the primary determinate of baseline (no
CO exposure) %COHb. This comes to light in Table 6-17 on page 6-18 which shows the APEX
model to result in a median value of %COHB somewhere between 0.25 and 0.50% COHb and
the non-parametric distribution of values is considerable. The modelers claim that this data can
not be compared to any studies in the literature because of the time frame over which the data is
modeled. Unfortunately there will always be skepticism of any model that can not be validated
practically with actual measurements. The study by Allred et al observed 63 subjects in 3 cities
on 4 experimental days (repeated measurements on an individual occurred within 6 weeks) and
all of the subjects were observed over less than a 2 year period ( 270 measurements of baseline
levels of %COHb). The mean %COHb levels did not vary over this period of time. These values
ranged between 0.62 and 0.64 %COHb with a standard deviation of 0.16% COHb. The model
results does not fit these results on the population most at risk in this assessment.
The search for a alternative formats for setting standards maybe statistically enticing but would
present too many problems for implementation of public health measures when the standards
have been exceeded. News readers have a difficult time with the current standards provided in
PPM so I can only imagine how they would attempt to explain any of the proposed alternative
methods.
There is also no attempt made to employ the model in the studies dealing with controlled
exposures in subjects with CAD. I know that using APEX to model a 1 hour exposure to CO is
akin to killing a slug with a sledge hammer, but what other data base is as relevant to validation
of the model being so widely used?
7 In the Panel's view, lo what extent does the modified assessment approach employed in
this second draft assessment provide results that meaningfully inform EPA 's consideration of
the public health implications of the current standards?
19
-------
The additional information included in the current modification does not relieve the anxiety from
the use of so little monitoring data to provide guidance for setting standards. Perhaps it is in the
way in which the information is presented: there is no data presented or referred to that shows
that including more data adds nothing to the analysis. As a result of what has currently been
presented. 1 am still uneasy with the use of so little of the available monitoring data for this risk
assessment. Does the use of data from a few monitoring cites imply that we need fewer monitors
in our cities because we can accurately predict what exists in the entire metro area based upon a
few monitors? How is the reader to interpret the use of data from so few sites. To state the above
concern in another way, the document could be improved by showing the impact, or lack of
impact, on the dose variability by including data from more monitoring sites. This could be
previously published information and need not involve re-running of the models with data from
these additional sites.
There is also mention of the lack of data resolution of the LA monitors vs the Denver monitors
but the impact of the low resolution monitors in LA is not discussed. If it has no impact why was
the issue raised?
It is my impression that the estimates of risk due to exposure to CO are to apply to entire country
and not just to Denver and Los Angeles. It is clear that the detail presented for these two cities
can not be also presented for all of the major metropolitan areas of the country. However it is
incumbent upon the authors to address how these risk assessments for Denver and LA are to be
applied to the entire population of the United States. After all the document goes to great lengths
to describe how many people are in the at risk group in the country and then applies some metric
to determine how many of these individuals are in Denver and in LA.
20
-------
Dr. Russell Dickerson
2. Does the Panelfind the considerations of current ambient carbon monoxide monitoring
data, including specifically the data for the monitors included in this draft of the assessment,
and the discussion of the extent to which near roadway concentrations are represented to be
technically sound, clearly communicated, and appropriately characterized?
The revised draft is much improved, and generally meets expectations. EPA is faced with the
situation that ambient measurements are more accurate than emissions. This makes
measurement/model comparisons difficult. Although Chemical Transport Models (CTM's) are
not used in Chapter 3 or the REA, that chapter should point out that such tools should be used
and continued evaluations and improvements (if necessary) in emissions are needed.
The increased discussion of NCore and Analytical Sensitivity are useful and appropriately placed
near the front.
Key Observations (page 3-18 and 3-19) are appropriate, except I could not find much on the
uncertainties in emissions (see also comments on ISA) and their impact on model output.
Chapter 5 appears to reflect the state of knowledge.
Preliminary Comments on the ISA (relevant to REA)
In one last reading of the ISA, Chapter 2 does a thorough job of providing an overview of our
technical understanding of CO.
A few comments:
1. Page 2-3 might mention HCHF's are removed by OH
2. The caption to Figure 3-1 might include the word DIRECT so people don't go looking for
isoprene.
3. Section 3.2.2 looks really good, as does Figure 3-8. This makes an important point that should
appear in the PA.
4. The sections on ambient measurements, detection limits and NCore are all much improved.
5. Based on a quick read, Section 3.5 looks solid now.
In poking around the literature I have found a few more papers that evaluate the emissions
inventories of CO. Kuhns et al. (2004) and Yu et al. (2007; 2009) also found evidence of
overestimates of CO emissions in Mobile6 or CMAQ. Marmur et al. (2009) in contrast seem to
find that the CO emissions are underestimated. Zhang and Batterman (2010) find that Mobile 6
matches plume dispersion models and a roadside monitor reasonably well. We are just about to
submit a paper that evaluates Mobile6 CO emissions using about 100 altitude profiles and find
that modeled emissions are high but not in gross error. If the paper is accepted soon enough I
will send a preprint.
21
-------
Kuhns, H. D., et al. (2004), Remote sensing of PM, NO, CO and HC emission factors for on-
road gasoline and diesel engine vehicles in Las Vegas, NV Science of the Total Environment,
322, 123-137.
Marmur, A., et al. (2009), Evaluation of model simulated atmospheric constituents with
observations in the factor projected space: CMAQ simulations of SEARCH measurements,
Atmospheric Environment, 43, \ 839-1849.
Yu, B.C., et al. (2007), A detailed evaluation of the Eta-CMAQ forecast model performance for
Os, its related precursors, and meteorological parameters during the 2004 ICARTT study,
Journal of Geophysical Research, 112, D12S14.
Yu, S.C., et al. (2009), Eta-CMAQ air quality forecasts for 03 and related species using three
different photochemical mechanisms (CB4, CB05, SAPRC-99): comparisons with
measurements during the 2004 ICARTT study, Atmospheric Chemistry and Physics
Discussions, 9, 22955-22992.
Zhang, K., and S. Batterman, Near-road air pollutant concentrations of CO and PM2.5: A
comparison of MOBILE6.2/CALINE4 and generalized additive models, Atmospheric
Environment, doi: 10.1016/j.atmosenv.2010.02.008, in press 2010.
22
-------
Dr. Milan Hazucha
Background on assessing ambient CO exposure and risk (Chapter 2).
Charge Question 1: Does the Panel find the summary of CO exposure and discussion of
ambient CO sources, exposures, dose, health effects and risk characterization approach to be
technically sound, clearly communicated, and appropriately characterized?
Qualified yes in all respects.
In addition to already mentioned endogenous CO production and exogenous sources (p. 2-6,
1. 11-13), additional source is metabolic production of CO due to inhalation of, e.g.,
dihalomethanes, other substances and certain medication.
Do we really consider people using recreational drugs to be at-risk population and be
considered in risk assessment (p. 2-8,1.19)? If so we will have to consider smokers to be at-
risk population as well.
Since Allred et al. studies provide the key evidence for CO health effects assessment it would
be very helpful if the document, in addition to % changes of the critical endpoint, e.g., time
to angina also reported the actual mean and CI (confidence interval) values in respective
endpoint units (p.2-11,1.24-27 and p.2-12,1. 10-13). How clinically significant is shorter by
22 seconds time to angina out of nearly 9 minutes? Besides reduced time to angina, was the
duration and the intensity of angina affected as well? Did frequency of angina attacks
increased because of CO exposure? If these endpoints were not reported by the investigators,
it should be specifically stated so in RE A.
1 support very cautious approach to some epidemiology studies reports of the effects of CO
on respiratory system (p.2-10, 2-13, 2-18). I fully agree with EPA assessment that the
interpretation of CO-induced lung-related outcomes "is affected by uncertainties including
with regard to the biological mechanism that could explain CO-induced outcomes" (p. 2-13,
1.8-11).
As far 1% COHb benchmark suggested by the Panel, the staff correctly pointed out that "this
level overlaps with the upper part of the range of endogenous levels" and decided not to
focus on dose estimates (p.2-16,1. 26-34). I support this approach since this complies with
the EPA's task "to establish standards that are neither more nor less stringent than necessary
for these purposes", .i.e. public health.
23
-------
Air Quality Considerations (Chapter 3 and 5)
Charge Question 2: Does the panel find the considerations of current ambient carbon
monoxide monitoring data, including specifically the data for the monitors included in this draft
of the assessment, and the discussion of the extent to which near roadway concentrations are
represented to be technically sound, clearly communicated, and appropriately characterized?
Yes in all respects.
Characterization of Exposure. Dose and Potential Risk (Chapter 4-61
Charge Question 3: In recognition of CASAC comments of first draft REA, this draft REA is
expanded from the previous assessment in a number of ways (summarized in section 1.3 of the
draft document). The assessment study areas are in the Denver and Los Angeles study areas. We
are interested in eliciting the views of the Panel on the usefulness of this approach in informing
our review the NAAQS for CO. What are the Panel members' views on the following aspects in
which the assessment has been expanded from the previous draft?
The charge questions span across 3 chapters: Ch. 4 Overview of APEX modeling, Ch. 5
Application of APEX, and Ch. 6 Simulated exposure and COHb dose results.
However, it is difficult to comment on the chapters in general since the sub-questions are
rather specific.
A. An important change of this assessment from the first draft is the expansion of each of the
modeling domains to include a greater number of ambient monitors used as input to
APEX. Additionally, this draft assessment employs an algorithm that adjusts for temporal
and spatial heterogeneity in ambient concentrations across each study areas.
The bulleted list of modified/expanded sections is very helpful. Similarly, the flow
chart showing input points and flow of data has been very helpful as well.
How does the expansion of modeling domains in this REA compare to REA1? Were
the estimates about the same or different and how they were different?
Generally, larger number of monitors may improve exposure assessment. Figure 5.1
(p.5-5) shows a considerable overlay of air districts, meteorological zones and for Los
Angeles study areas as well. Although the overlapping districts, zones and areas were
adjusted for as far as exposure goes, how were they treated in terms of input data into
other modules of APEX? What approach was used to avoid duplication of input
data? For L. A., was one of the areas designated as a dominant source or each area
was considered separately in the assessment?
In generation of simulated individuals, demographic variables should include socio-
economic status and race. These variables will impact other APEX modules,
particularly the COHb one.
Chapter 6: The tables tabulating exposure values vs estimated number and percentage
of CHD persons affected at specific CO concentrations are very instructive and
24
-------
revelatory. Tables showing estimated COHb levels vs number of CHD persons and
persons/days are similarly instructive. Generally, these table show that the current
level of both standards is protective as required by the legislation, i.e., "standards that
are neither more nor less stringent than necessary."
Additional calculations and tabulation of endogenous COHb level using APEX has
been also very helpful. Table 6-18 (p.6-19) shows that the endogenous contribution to
a total COHb is "less than 0.5%, though for a limited number of hours, the
endogenous contribution could be over 1.0% COHb." The additional plots reflect
contribution of endogenous COHb to a total COHb level essentially following
physiologic laws of Haldane.
Section 6.5 Key Observations (p.6-24) summarizes the main observations presented
in this chapter. One of the important conclusions is that more than 95% of simulated
at-risk population of L.A. study areas will experience an annual daily maximum end-
of-hours COHb level below 1.5%. Moreover, when considering alternative standards
"only 0.1% of the CHD population was estimated to experience a maximum end-of
hour COHb at or above 2%. Similar values are provided for Denver.
B. The current draft assessment also include an increase in the number of
microenvironments modeled over that in the first draft (from two to eight) and improved
the representation of variability in estimated microenvironmental concentrations,
including in-vehicles.
Chapter 4: Table 4-4 (p. 5-21) and 4-5 (p. 4-29) lists 15 microenvironments used in
estimates in pNEM/CO model in 2000, and the same number and type in APEX4.3.
However, the number for this draft is reduced to 8. Did this change in any way
affected the estimates?
It would have been helpful instead of making a general statement on how better
APEX is, to actually list in 4.5 Key Observations section(p. 4-35) specific
enhancements of APEX over pNEM/CO.
The two important conclusions were that (1) the policy relevant background was
negligible, and (2) the fixed site monitoring data could be adjusted. The tables
provide sufficiently detailed data to evaluate all 5 scenarios.
In the current draft the number of microenvironments was increased to 4 indoor, 3
outdoor and 1 in-vehicle. Such expansion may improve strategies and enhance the
validity as well as credibility of the assessment. More realistic scenarios provide
stronger and more representative base for decision making.
C. This draft assessment has implemented the mass-balance model for estimating
concentrations in indoor microenvironments.
1 consider the selection of mass-balance model for indoor air appropriate.
25
-------
Charge Question 4. Does the Panel view the results of the draft exposure analyses to be
technically sound, clearly communicated, and appropriately characterized?
Qualified yes to all respects.
The various modules of APEX model are regularly upgraded to improve the simulation
process making it as realistic as possible. Yet the COHb module to estimate venous blood
COHb level, the ultimate endpoint remains the same, i.e. based on CFKE (p.4-34,1.20-31).
As already commented on this matter by several panel members including myself in the past
why is EPA so adamant exploring more recent and more sophisticated CFK equations?
Replacing original CFK with an enhanced, e.g. Bruce and Bruce module should be simple
enough. If there are no substantial differences, then no change is necessary. However, if
there are differences in COHb estimates, then we may search and evaluate the factors that
may have affected the change. Such information may potentially useful in standard setting.
Moreover, regardless of a mathematical model employed in COHb module, the COHb
estimates can be improved by tuning some of the explicit input variables such as THb and
DLco.
Appendix A: I agree with the summary of findings (p. A-5) that the current physiology file
data is obsolete and may even be incorrect for some variables. While some variables were
already updated, others such as race, SES, THb and DLco should be either added to or
replace the old data in the input module.
Charge Question 5: Does the Panel find the derivation and presentation of the modeling
approach as a whole (chapters 4 and 5) to be technically sound, clearly communicated, and
appropriately characterized?
Yes in all respects. The staff did an excellent job of presenting and discussing APEX model.
1 agree with well reasoned arguments and the conclusions.
Any concerns about the effect of missing concentration values on their distribution were
cleared by addition of descriptive statistics tables (5-7 through 5-10). The tables demonstrate
that the missing values whether estimated and corrected for or not do not influence the
distribution of hourly values either in Denver or Los Angeles. Similar approach to estimation
of missing temperature values required by APEX likely resulted, as stated in the document,
in negligible differences.
Charge Question 6: Does the Panel find the derivation and presentation of the COHb estimates
(chapters 5 and 6) to be technically sound, clearly communicated, and appropriately
characterized?
Qualified YES in all respects.
Section 6.3.1 referenced in the document is likely section 6.4.1.
26
-------
The section 6.4 on endogenous production of CO and its contribution to overall COHb in
combination with ambient levels of CO is very informative. Table 6-17 clearly shows that
even at 0 ppm CO in ambient air several hundred individuals will reach COHb level as high
as 1.8%. It would be instructive to identify groups of individuals (e.g. with anemia) who
exceeded 1% COHb level due to endogenous production.
Figure 6-2 shows, as expected, that endogenous CO will not influence COHB level if the
ambient CO concentration exceed the one produced endogenously (this needs to be stated
more clearly on p. 6-21,1. 6-8). Figure 6-4 indeed confirms the above statement.
Appendix B. COHB module
p. B-3: The P]CO , should be defined as a partial pressure.
p. B-5 In eq. B-l 1 and B-14 PC02 subscript should be correct to read not as CO2 (carbon
dioxide) but as Co2 (capillary 02).
Suggest to move the second paragraph on page B-8 as the first paragraph of the section,
otherwise without the explanation, the statement is misleading.
p. B-9- B-14. Section C4: The COHb module seems to be the weakest of the APEX modules.
Primarily, it is because we do not have sufficient data over the physiologic range for many
variables. However, though still limited some physiologic data are available for healthy and
at-risk groups and they should be integrated into data base for COHb module. From the
tables nor the text it does not look like that many critical variables such as Hb, DLco,
endogenous CO and others were, besides age and gender, adjusted for other physical
characteristics or disease conditions. For example, the amount of Hb will determine the rate
of COHb formation and is a critical variable. There are substantial differences between blood
concentration of Hb in whites and blacks.
Charge Question 7: In the Panel's view, to what extent does the modified assessment approach
employed in this second draft assessment provide results that meaningfully inform EPA's
consideration of the public health implications of the current standard?
I agree with the expanded approach and I believe that it will allow for more accurate
assessment and risk-characterization.
Characterization of Variability and Uncertainty (Chapter 7)
Charge Question 8: What are the views of the Panel regarding the adequacy of the assessment
of uncertainty and variability? To what extent have sources of uncertainty been identified and the
implications for risk characterization been characterized? To what extent has variability
adequately described and represented?
The staff adequately described uncertainty and variability. However, from table 7-1 it appears
that CHD has been considered to be the only sources of variability and no other disease
27
-------
conditions were considered in the model. Why no other relevant diseases were considered? Was
socio-economic status in any way considered in estimating uncertainty?
Does APEX model has build in any internal consistency check between factors used in the
calculations (p. 7-2)? For example, randomly selected oxygen uptake which may be high maybe
assigned to an individual with CHD who is unable to achieve such uptake level.
Activity patterns of persons 30 years ago used as APEX input are very much different for current
activity patterns (p. 7-8). Can CHAD data be limited only to more recent activity patterns??
28
-------
Dr. Francine Laden
1 Does the Panel find the summary of CO exposure and discussion of ambient CO sources,
exposures, dose, health effects and risk characterization approach to be technically sound,
clearly communicated, and appropriately characterized?
Chapter 2, background on assessing ambient CO exposure and risk, is well organized and
technically sound. EPA appropriately characterized what they did do, as well as what they did
not do. The discussion explaining the uncertainties associated with directly using studies of the
association of cardiovascular morbidity with measurements of ambient CO is important for
motivating the focus on COHb levels. It may not be immediately obvious to all readers why
ambient CO exposures are not the exposure of interest. One concern is that most of the
monitoring data and the laboratory data crucial to the assessment is quite old and could thus
effect the determination of risk. Is EPA confident that current situations can be extrapolated
appropriately from what was observed in the past? Perhaps some statement to this uncertainty
would be valuable, as well as acknowledgement that there are not any appropriate-more recent
studies available. For the most part the chapter is clearly communicated. However, the chapter
overall would benefit from some careful editing.
5 Does the Panel find the derivation and presentation of the modeling approach as a -whole
(chapters 4 and 5) to be technically sound, clearly communicated, and appropriately
characterized?
The derivation and presentation of the modeling approach as a whole is very well presented. I
had one trivial question: Could the prevalence of undiagnosed CHD be greater for women than
for men? The model assumes that the ratios of undiagnosed cases to diagnosed cases are
identical for each gender and also that this ratio has not changed since 1990. The text should at
least acknowledge that this might not be so.
6 Does the Panel find the derivation and presentation of the COHb estimates (chapters 5
and 6) to be technically sound, clearly communicated, and appropriately characterized?
The derivation and presentation of the COHb estimates are technically sound, clearly
communicated and appropriately characterized.
7 In the Panel's view, to -what extent does the modified assessment approach employed in
this second draft assessment provide results that meaningfully inform EPA 's consideration of
the public health implications of the current standards7
The draft assessment provides results that meaningfully inform EPA's consideration of the
public health implications of the current standards. Given that CO levels have decreased
significantly over the years, that levels rarely approach the standards, and that elevated levels of
COHb estimated by the risk assessment are quite low, the usefulness of the current standards
may need to be reassessed.
29
-------
Dr. Arthur Penn
1 Does the Panel find the summary of CO exposure and discussion of ambient CO sources,
exposures, dose, health effects and risk characterization approach to be technically sound,
clearly communicated, and appropriately characterized7
No major issues with this chapter. The summaries are well-done and the health effects and risk
characterization approach are presented clearly and seem technically sound.
30
-------
Dr. Beate Ritz
3 What are the Panel members' views on the following aspects in which the assessment has
been expanded from the previous draft7
A An important change of this assessment from the first draft is the expansion of each of the
modeling domains to include a greater number of ambient monitors used as input to APEX.
Additionally, this draft assessment employs an algorithm that adjusts for temporal and
spatial heterogeneity in ambient concentrations across each study area.
While the larger number of ambient air monitors may have improved exposure assessment, I am
not convinced that this in fact gives more correct estimates of exposure near roadways - mainly
in homes since only one singular distribution was used for all homes and this distribution may
not adequately reflect near roadways exposures in homes i.e. 1 am not convinced that spatial
heterogeneity driven by proximity to sources can be adequately captured by the ambient
monitoring network. In fact, the exposure both in homes and in work places closer to roadways
might be underestimated.
B. The current draft assessment also includes an increase in the number of
microenvironments modeled over that in the first draft (from two to eight) and improved the
representation of variability in estimated microenvironmental concentrations, including m-
vehicles
yes
C This draft assessment has implemented the mass-balance model for estimating
concentrations in indoor microenvironments.
yes
4 Does the Panel view the results of the draft exposure analyses to be technically
sound.clearly communicated, and appropriately characterized7
Yes, except for the assumption about having captured spatial heterogeneity in homes and work
places near busy roadways.
31
-------
Dr. Paul T. Roberts
2. Does the Panel find the considerations of current ambient carbon monoxide monitoring
data, including specifically the data for the monitors included in this draft of the
assessment, and the discussion of the extent to which near roadway concentrations are
represented to be technically sound, clearly communicated, and appropriately
characterized?
In general, the treatment of the CO monitoring data and the (admittedly poor) extent that these monitors
represent near-roadway concentrations, including the data used for this version of the assessment, are
improved over the 1st external draft REA, technically sound, clearly communicated, and appropriately
characterized. In addition, the use of data from more monitors as input to the exposure modeling is a
significant improvement.
Characterization of Exposure (Chapters 4 and 5)
The following changes from the 1st external draft REA are significant improvements and help the results
from this REA do a much better job of informing our review of the CO NAAQS.
• Expansion of the modeling domain to include more monitors in both Denver and LA.
• Adjusting for both spatial and temporal heterogeneity in ambient CO concentrations in each
study area.
• A significant increase in the modeled microenvironments.
• The use of a mass-balance model for estimating CO concentrations in indoor environments
(factors are reasonable estimates for the other microenvironments).
However, I am concerned about the use of the power of .621 in equations 4-11 and 4-22, which
reduces the CO concentration at an outdoor location, relative to the nearest central monitor. The
main justification for this is given on lines 33 to 35 of page 4-27 as a way to get the pNEM/CO
and APEX models to agree, but I do not understand the physical rationale for this. On page 5-
23, lines 8-10, it is suggested that the resulting "compression effect" is consistent with Wilson et
al (1995). However, even if we agree that this might be occurring near most residences, as in the
Wilson study, it does not occur at near-roadway or in-vehicle locations. In fact, the net result of
using this as part of the factor calculation for estimated CO concentrations is that near-roadway
and in-vehicle concentrations are a fair amount lower than was documented in the 1st REA,
section 5.4.2 (in-vehicle concentrations) and in the ISA for near-roadway. Since these two
microenvironments might be contributing a fair amount to total exposure, I think this is an
important issue to resolve. In addition, the use of this factor for near-road and in-vehicle
microenvironments has probably decreased the percent of people who experience the highest
concentration exposures, for example in Chapter 6.
Table 5-16 is a good summary of the various conditions in the 8 microenvironments, especially
with the distributions. I did noticed, however, in Appendix that a couple of locations codes are
probably mis-assigned, although these are probably small contributions: bicycle should be 5, as
shown in Table 5-16, and all the boat categories should be 8, since boats are uncontrolled for CO
32
-------
and produce significantly high CO concentrations (they should probably be much higher than 8,
but again this is probably a small contributor).
8 What are the views of the Panel regarding the adequacy of the assessment of uncertainty
and variability? To what extent have sources of uncertainty been identified and the
implications for the risk characterization been characterized7 To what extent has
variability adequately described and represented7
In general, the incorporation of more monitors in each area, more microenvironments, and
variability in various variables within APEX is an important method for addressing variability
and thus general uncertainty in exposure and dose.
Although 1 still think that the most significant uncertainties from this table could be better
quantified by using sensitivity runs of the model, I understand the time constraints on the current
NAAQS process. I believe that the significant improvements in representing near-roadway and
in-vehicle exposures has reduced the uncertainties associated with that end of the exposued
population, as represented in Table 7-2.
33
-------
Dr. Anne Sweeney
1 Does the Panel find the summary of CO exposure and discussion of ambient CO sources,
exposures, dose, health effects and risk characterization approach to be technically sound,
clearly communicated, and appropriately characterized?
Overall, Chapter 2 is a very well-written comprehensive background that describes the issues and
considerations that were confronted in the effort to assess ambient air CO exposure and human
health risks. The contributions of the various sources of both ambient and indoor CO levels were
clearly described and supported by numerous published studies. Exposure pathways and the
importance of the microenvironment were also weII-documented.
The justification for the utilization of persons with CHD as the unit of analysis in the quantitative
assessment is appropriate, given the lack of data on COHb levels in other potentially high risk
groups. However, characteristics of this simulated population that should be included in the
modeling include the population prevalence of income level (a surrogate for several important
covariates, e.g., residence near congested traffic areas) and smoking (also related to income
level).
34
-------
Dr. Stephen Thorn
1 Does the Panel find the summary of CO exposure and discussion of ambient CO sources,
exposures, dose, health effects and risk characterization approach to be technically sound,
clearly communicated, and appropriately characterized?
The summary is accurate and appropriate.
2 Does the Panel find the considerations of current ambient carbon monoxide monitoring
data, including specifically the data for the monitors included in this draft of the assessment,
and the discussion of the extent to which near roadway concentrations are represented to be
technically sound, clearly communicated, and appropriately characterized7
The discussion on current air quality monitoring is accurate and appropriate.
3 In recognition ofCASAC comments on first draft REA, this draft REA is expanded from
the previous assessment in a number of ways (summarized in section 1 3 of the draft
document). The assessment study areas are in the Denver and Los Angeles study areas.
We are interested in eliciting the views of the Panel on the usefulness of this approach in
informing our review the NAAQSfor CO. What are the Panel members' views on the
following aspects in which the assessment has been expanded from the previous draft7
A. An important change of this assessment from the first draft is the expansion of each
of the modeling domains to include a greater number of ambient monitors used as
input to APEX Additionally, this draft assessment employs an algorithm that adjusts
for temporal and spatial heterogeneity in ambient concentrations across each study
area.
B The current draft assessment also includes an increase in the number of
microenvironments modeled over that in the first draft (from two to eight) and
improved the representation of variability in estimated microenvironmental
concentrations, including m-vehicles
C This draft assessment has implemented the mass-balance model for estimating
concentrations in indoor microenvironments
I found the document to be generally well written. My one question pertains to the APEX
modeling, as raised in my review of the Policy Assessment document. The discussion in the
REA document includes information that most fixed monitors have a 1 ppm CO lower detectable
limit so the modelers added 0.5 ppm CO to all measured values to remove zeros and negative
numbers thought to be related to monitor drift It seems to me that this makes it exceedingly
difficult to accept estimates of the at-risk population and threshold COHb levels.
35
-------
4 Does the Panel view the results of the draft exposure analyses to be technically sound,
clearly communicated, and appropriately characterized?
1 am unsure if the draft exposure analysis is technically sound (see comment #3).
36
------- |