5 UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
5 WASHINGTON D.C. 20460
OFFICE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
June 24, 2009
EPA-CASAC-09-011
The Honorable Lisa P. Jackson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, NW
Washington, D.C. 20460
Subject: Review of EP A's Integrated Science Assessment for Carbon Monoxide (First
External Review Draft)
Dear Administrator Jackson:
The Clean Air Scientific Advisory Committee (CASAC) Carbon Monoxide Review
Panel met on May 12-13, 2009, to review the EPA's Integrated Science Assessment for Carbon
Monoxide (First External Review Draft, March 2009). This letter has been reviewed and
approved by the chartered CASAC at a public conference call on June 17, 2009. This letter
provides CASAC's overall comments and evaluation. We highlight the most important issues
which need to be addressed as the first draft Integrated Science Assessment (ISA) is revised.
The CASAC and Panel membership is 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.
CASAC commends the EPA staff for the development of a comprehensive, readable, and
good quality first draft of the Integrated Science Assessment for Carbon Monoxide. The
document pulls together critical evidence from the past decades while emphasizing new evidence
and associated insights. The extensive literature is thoughtfully summarized. The document
makes effective use of tables and appendices. We applaud the process used by the EPA to
produce this document. The EPA has implemented a process that is consistent with current
approaches to evidence review and synthesis. It has progressively refined this process in recent
NAAQS reviews. Our major comments follow:
• A key issue is susceptible populations who might have a greater response to the inhalation of
Carbon Monoxide (CO) as it combines with hemoglobin in the blood, thereby reducing
oxygen delivery. These susceptible populations drive the CO standard. The ISA should
focus more on individuals with pre-existing cardiopulmonary disease and also address how
the consequences of CO exposure may be modified by exercise, altitude, low hematocrit, as
-------
well as by exposures to active and passive tobacco smoke. We encourage clarity and
quantification of the magnitude of susceptible and vulnerable populations.
• The report needs to give greater attention to the heterogeneity of CO concentrations within
urban areas and to the available literature on exposure modeling. Relying only on EPA's
fixed monitoring network CO measurements may underestimate CO exposures for specific
vulnerable populations such as individuals residing near heavily trafficked roads and who
commute to work on a daily basis. The degree to which the available monitoring capabilities
can reflect the temporal and spatial patterns of CO concentrations need to be characterized.
Exposure assessments should be evaluated more critically in the revised ISA. Understanding
the extent of exposure measurement error is critical for evaluating epidemiological evidence
and for using exposure assessments.
• An essential aspect of evaluation of the evidence on CO - in part because levels are declining
- is the issue of co-pollutants. In urban air, CO is always present in a mixture with other
pollutants. Distinguishing the effects of CO per se from the consequences of CO as a marker
of pollution or vehicular traffic is a challenge, which this report needs to confront as
thoroughly as possible.
• The role of CO as a participant in global atmospheric chemistry requires greater explication.
The ISA should expand the discussion regarding the indirect role of CO on climate change as
mediated by atmospheric conversion of greenhouse gases. For example, reduction of CO
emissions, in addition to potentially improving health, could mitigate greenhouse gas
concentrations. This topic could be more strongly developed in the ISA.
• We endorse the inclusion of new information on health outcomes other than those CO effects
not mediated by hypoxic mechanisms. Outcomes such as auditory system effects as well as
developmental and neonatal adverse outcomes should also be highlighted. CASAC
encourages continued tracking and integration of these active areas of research into the ISA.
With regard to the structure of the report, we found the summaries of Chapters 1-4
helpful. We would like the next ISA draft to include a summary of Chapter 5. This inclusion is
particularly important since this is the final chapter. The emphasis should be on the most
important and recent scientific evidence and conclusions.
CASAC also notes that the ISA documents a substantial decline in CO levels in urban
areas over the past two decades. This decline is noteworthy and undoubtedly benefited public
health.
-------
CASAC reiterates its expectation that the revised ISA will be accompanied by a
delineation of key changes from the first draft. This will enhance the efficiency and targeting of
subsequent CASAC reviews, and will provide a transparent record of the basis for these changes.
The CASAC looks forward to reviewing the next draft of the ISA.
Sincerely,
/Signed/ /Signed/
Dr. Joseph D. Brain, Chair Dr. Jonathan M. Samet, Chair
Clean Air Scientific Advisory Committee Clean Air Scientific Advisory Committee
Carbon Monoxide Review Panel
Enclosures
-------
Enclosure A
ROSTER
U.S. Environmental Agency
Clean Air Scientific Advisory Committee
Carbon Monoxide Review Panel
CASAC MEMBERS
Dr. Joseph D. Brain, (Chair) 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. Armistead (Ted) Russell, Professor, Department of Civil and Environmental Engineering,
Georgia Institute of Technology, Atlanta, GA
CO PANEL MEMBERS
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, The University of
Maryland, College Park, MD
Dr. Laurence Fechter, Senior Career Research Scientist, Department of Veterans Affairs ,
Research Service (151), Loma Linda VA Medical Center, Loma Linda , CA
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. Michael T. Kleinman, Professor, Department of Medicine, Division of Occupational and
Environmental Medicine, University of California, Irvine, Irvine, CA
Dr. Arthur Penn, Professor LSU School of Veterinary Medicine, Department of Comparative
Biomedical Sciences, LSU SVM - Room 2425, Louisiana State University, Baton Rouge, LA
Dr. Beate Ritz, Associate 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. Stephen R. Thorn, Professor, Institute for Environmental Medicine, 1 John Morgan
Building, University of Pennsylvania, Philadelphia, PA
SCIENCE ADVISORY BOARD STAFF
Dr. Ellen Rubin, Designated Federal Officer, 1200 Pennsylvania Avenue, NW, Washington,
DC, Phone: 202-343-9975, Fax: 202-233-0643, (rubin.ellen@epa.gov)
-------
ROSTER
U.S. Environmental Agency
Clean Air Scientific Advisory Committee
CHAIR
Dr. Jonathan M. Samet, Professor and Chair, Department of Preventive Medicine, University
of Southern California, Los Angeles, CA
CASAC MEMBERS
Dr. Joseph Brain, Philip Drinker Professor of Environmental Physiology, Department of
Environmental Health, Harvard School of Public Health, Harvard University, Boston, MA
Dr. Ellis B. Cowling, University Distinguished Professor At-Large Emeritus, Colleges of
Natural Resources and Agriculture and Life Sciences, North Carolina State University, Raleigh,
NC
Dr. James Crapo, Professor of Medicine, Department of Medicine, National Jewish Medical
and Research Center, Denver, CO
Dr. 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, IL
Dr. Armistead (Ted) Russell, Professor, Department of Civil and Environmental Engineering,
Georgia Institute of Technology, Atlanta, GA
SCIENCE ADVISORY BOARD STAFF
Dr. Holly Stallworth, Designated Federal Officer, EPA Science Advisory Board Staff Office,
Washington, DC
-------
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 B
Responses to Agency Charge Questions
1. The framework for causal determination presented in Chapter 1 was developed and
refined in other ISAs (e.g., the PM ISA). During previous reviews, CASAC generally
endorsed this framework in judging the overall weight of the evidence for health effects.
Please comment on the extent to which Chapter 1 provides necessary and sufficient
background information for review of the subsequent chapters of the CO ISA.
Chapter 1 is generally well-written, well-organized, and useful in content. The summary is
helpful.
Section 1.6, EPA Framework for Causal Determination, is appropriately very similar, or in
places identical, to Section 1.6 of the Integrated Science Assessment for Paniculate Matter (First
External Review Draft, December 2008). As EPA receives comments on this framework when
reviewed by various panels of CASAC, EPA should strive for consistency across documents.
The Particulate Matter (PM) Review Panel offered several comments. For example, "the
categorization reflects the strength of evidence and not the potential magnitude of public health
benefits." This implies that there is a distinction between weight of evidence and the potential
sensitivity or magnitude of the outcome. This distinction should be appropriately conveyed by
discussing both the weight of evidence and the magnitude or sensitivity of each health effect
endpoint. A second point is that additional clarification regarding the terms "susceptible" and
"vulnerable" would be useful - the PM Review Panel provided detailed comments along these
lines. For consistency these comments should be addressed across ISAs as well as in the Scope
and Methods Plan for Health Risk and Exposure Assessment (REA). A third suggestion is to
consider the role that publication bias might have as it relates to making weight of evidence
determinations.
The methodological framework provided in Section 1.6 is very similar to that used by the
Institute of Medicine (IOM) and the International Agencies for Research on Cancer (IARC).
However, a key difference is that that these agencies convene expert committees to review the
literature in depth and to apply criteria in order to arrive at conclusions about causality. The
process and criteria used by EPA staff to make judgments regarding weight of evidence must be
made clear and transparent. For example, there was insufficient clarity about the relative
emphasis that was given to of clinical exposure studies versus epidemiological studies.
-------
The material in Section 1.6 is methodological and thus generic. This section should be tailored
to address implementation of the methodological framework with respect to CO. For example,
Section 1.6 should introduce issues that provide a foundation for later chapters such as the role of
controlled studies, epidemiology, toxicology and other information sources. Even though
laboratory experiments have the advantage of being free of confounding or modifying effects
from co-pollutants, epidemiological studies have the advantage of addressing susceptible
subpopulations and long-term health effects that cannot be assessed via controlled clinical
exposure studies.
2. Chapter 2 presents the integrative summary and conclusions from the health effects
evidence, with the evidence characterized in detail in subsequent chapters. What are the
views of the Panel on the effectiveness of the integration of atmospheric science,
exposure assessment, dosimetry, pharmacokinetics, and health effects evidence in the CO
ISA?
Chapter 2 is a key chapter of the ISA and should remain "up front" in the ISA to inform and
assist the reader to better understand the following chapters and the most important findings and
points in those chapters. Brief recommendations are made to strengthen this chapter.
The summary of the 1st four topics in the charge question above (atmospheric science, exposure
assessment, dosimetry, pharmacokinetics) consist of only three and a half pages, while the last
topic, health effects evidence, is summarized in 11 pages. We recommend expanding the
material on the first four topics. A strength of the section on the health effects evidence, absent
from discussion of the other four topics, is the summary sentence—according to the EPA's 5-
level hierarchy-at the end of each major health effect. The Chapter 2 summary presents the
strong positive association between CO exposures in clinical settings and a) angina in human
volunteers and b) a variety of cardiovascular-related toxicology outcomes. Epidemiologic data
support associations between ambient CO levels and adverse cardiovascular, central nervous
system and birth outcomes, but the criteria for interpreting these study results in terms of
causality need to be described clearly.
A more direct examination of multi-pollutant exposures is recommended since in "real-life,"
CO-exposures are associated with exposure to numerous other traffic- and non-traffic-related
factors. Also recommended is a "take-home" statement summarizing the strength of evidence
discussing whether or not there are adverse health effects at or near current ambient levels.
Section 2.3.3, Birth Outcomes and Developmental Effects exemplifies the need to carefully
distinguish between weight of evidence and the strength of the association. The sections on
hospital visits and admissions for cardiovascular issues are two other examples.
The identification of vulnerable subpopulations is important. This should motivate consideration
of areas of focus for exposure assessment in the REA. Table 1 could be expanded to include a
summary showing whether data are based on experiments (human/animal) or epidemiology. The
numbers of subjects studied could also be listed.
-------
3. To what extent are the atmospheric science and air quality analyses presented in Chapter
3 clearly conveyed and appropriately characterized? Is the information provided
regarding CO source characteristics, CO chemistry, policy-relevant background CO, and
spatial and temporal patterns of CO concentrations accurate and relevant to the review
of the CO NAAQS?
The chapter reviews the state of the science and is accurate and up to date, but incomplete. Our
core comments are:
• The ISA needs to present a review of the literature sufficient to address the question: Is
there a compelling need to protect welfare from adverse effects on climate through
changes in atmospheric composition, oxidizing capacity, and radiative forcing?
• The panel believes that the current monitoring network is adequate to demonstrate
compliance with the NAAQS, but substantial improvement could be achieved in
coverage and detection limits to better quantify ambient CO concentrations, sources, and
exposure.
• Emissions models have been reported to disagree by a factor of two with field
measurements. This adds substantially to the uncertainty in numerical models of CO and
air quality in general.
CO plays a major role in global atmospheric chemistry and has an indirect radiative forcing of
about 25% of that of CO2 (IPCC FAR 2007). Moreover, the evidence that CO has a substantive
indirect impact on climate is growing stronger. The ISA acknowledges this in general, but needs
to summarize the policy-relevant scientific literature. How does the state of the science inform
our desire to protect welfare from adverse effects of large-scale changes in atmospheric
chemistry and climate?
The background level of CO has decreased throughout the 1990's but has since stabilized,
presumably due to increased emissions in the developing world. Reductions in emissions of CO
can have substantive beneficial effects on the radiative forcing that leads to global climate
change. The ISA needs to address this issue, and review the state of the science for both local
and global CO concentrations.
This chapter and others point out shortcomings in the monitoring network, but do not adequately
review the state of the science on available CO detectors, the actual uncertainty associated with
current measurements, or the spatial distribution and detection limits necessary to provide
sufficient information to evaluate models of human exposure, urban and mesoscale air quality, as
well as large scale effects. This relates to Question 4: The ISA concludes in section 3.7 that
central-site monitor concentration is generally a good indicator for the ambient component of
10
-------
personal CO exposure. What are the views of the Panel on this conclusion and its supporting
evidence?
The current ambient monitoring network is not well designed to characterize spatial and
temporal variability in ambient concentrations. Thus it does not adequately support detailed
assessments of human exposure or air quality modeling such as for photochemical oxidants.
Relevant microenvironments that are influenced by local factors, such as in-vehicles and in high
proximity to roadways, are not well represented. Although this point is acknowledged in various
places in the ISA, it does not seem to be consistently conveyed throughout the document. The
impact of the new NCore network should be reviewed both for the number of monitors and their
detection limits.
The paragraph of Section 3.2 on emissions models ends flatly with "EPA MOBILE6 vehicle
emissions model (http://www.epa.gov/otaq/m6.htm) now overestimates vehicle CO emissions by
a factor of ~2." This warrants deeper discussion - for example anthropogenic CO emission
sources, such as 4-stroke and 2-stroke spark ignited internal combustion engines, reflect differing
chemistry of CO formation and engine-out emissions. The role of catalytic converters needs to
be introduced, particularly as it pertains to "cold start" and "fuel enrichment" episodes of high
tailpipe CO emissions during a vehicle duty cycle. Factors that lead to on-road locations of high
CO emissions should be introduced. Whether these factors are adequately taken into account in
the comparison of emissions inventory and ambient ratios of CO to NOX should be discussed.
Moreover, EPA staff should give consideration to the role of the newly available Draft MOVES
2009 (Motor Vehicle Emission Simulator 2009, US Environmental Protection Agency, EPA-
420-B-09-008, http://www.epa.gov/oms/models/moves/420b09008.pdf) and its expected formal
successor (currently scheduled for release at the end of 2009) in improving characterization of
onroad emissions and, therefore, in better characterizing near-roadway air quality.
The inter-monitor variability described in Figure 3-28 should be presented more clearly and
more appropriately interpreted in terms of variability versus uncertainty. There is a need for
more quantitative information regarding the CO concentration gradient near roadways and a
comparison of near-roadway to area-wide monitoring data. The ISA could acknowledge cross-
media and co-pollutant consequences of oxygenated fuels. For example, use of ethanol as an
oxygenate (e.g., E5 or E10) or as an alternative fuel (e.g., E85) may lead to higher emissions of
some hydrocarbons and of nitrogen oxides. Some hybrid vehicles have many engine shutdowns
and starts during driving; whether this could have a "cold start" effect is not well known. It may
vary depending on the vehicle make and model, and duty cycle.
11
-------
4. How well do the choice and emphasis of exposure topics presented in Chapter 3 provide
useful context for the evaluation of human health effects in the ISA ? Is the discussion and
evaluation of evidence regarding human exposure to ambient CO and sources of
variability and err or in CO exposure assessment presented clearly, succinctly, and
accurately? The ISA concludes in section 3.7 that central-site monitor concentration is
generally a good indicator for the ambient component of personal CO exposure. What
are the views of the Panel on this conclusion and its supporting evidence?
In general, the discussions in Chapter 3 on CO source characteristics, CO chemistry, and policy-
relevant background CO are accurate and relevant to the CO NAAQS. As emissions from the
American vehicle fleet decrease and the number of violations of the NAAQS approach zero, it is time
to both congratulate the EPA and the State agencies for their success and to reassess our approach to
monitoring emissions and ambient concentrations of CO as well as personal exposure.
The detection limit and precision of the data from the currently deployed network of monitors is not
adequately reported in this document and is likely to add uncertainties for exposure assessment at CO
concentrations below the current NAAQS. The inclusion of measurements below the detection limit
and more precise measurements of low CO concentrations would allow us to better estimate total CO
exposure.
Total personal exposure to CO is the time weighted sum of exposure to CO in all
microenvironments including multiple outdoor environments (not just multiple indoor
environments). Increasingly, we have found that other microenvironments, such as near-road or
other hot-spot concentrations, significantly contribute to personal exposures, and we have data to
represent that exposure. Therefore the central-site monitor concentration is viewed as the best
available, albeit a limited indicator for the ambient component of personal CO exposure.
Ambient CO concentrations have been demonstrated to be heterogeneous, but this heterogeneity
is generally not reflected by central-site monitors. Therefore central-site monitor information is
limited in capturing all outdoor micro-environments that could have influenced exposure
assessments in epidemiological studies. Equation 3.4 should be reformulated to include at least
the in-vehicle and near roadway exposures (ref section 3.5.1.3 and Figure 3-34). This will also
require that the following sections (and any others) be modified to reflect that complex exposure:
1. Lines 30-31, page 3-57; lines 7-10, page 3-65 and page 3-74 lines 10-11. An analysis should
be conducted using the available monitoring and micro-environmental data to assess the likely
distribution of CO concentrations and those should be related to resulting changes in COHb,
particularly at the upper tail of the distribution. Limitations of this analysis, and likely biases,
should be identified.
5. The dosimetry and pharmacokinetics of CO are discussed in Chapter 4. Please comment
on the presentation in the ISA of the current state of knowledge on the Coburn-Forster-
Kane (CFK) model and model enhancements. Has the expected contribution of different
exposure durations (1-24 h) to COHb levels been clearly and accurately conveyed?
12
-------
Chapter 4 presents in sufficient detail various forms of the CFK model, as well as their
enhancements and limitations. It also discusses older empirical and recent multi-compartment
models. With so many different models it is, however, unclear which of the models would best
estimate venous COHb under the dynamic CO exposure conditions, e.g., an 8-h CO
concentration profile with several CO spikes. Several human exposure studies have reported
venous COHb levels during dynamic CO exposure profiles. It would be helpful to select the best
empirical, CFKE and multi-compartment model, apply them to such a profile and present the
results in a graphical form. The suggested models are Neto et al, 2008, Smith et al, 1994 and
Bruce and Bruce, 2006, respectively. The COHb estimates should provide information about
which model most closely predicts measured venous COHb and could potentially be considered
most suitable for dose estimation.
The question of the effects of different exposure durations on COHb formation was evaluated by
a mathematical model with integrated nonlinear CFK as enhanced by Smith et al., 1994. The
approach and the parameters selected, however, were not described in a sufficient detail nor were
the limitations discussed. This may lead to incorrect estimation of COHb particularly over
longer time periods (8h-24h) e.g., if the endogenous CO (COHb) value exceeds certain limits.
In addition to COHb modeling, the chapter also discusses overall pharmacokinetics of CO
transport, endogenous CO production, exogenous CO uptake and elimination. The factors and
conditions that may influence CO kinetics were discussed satisfactorily. It would be helpful to
give a range of endogenous CO values for a population at-risk, such as asthmatics or people with
metabolic syndrome, to have a better characterization of the potential increment from ambient
CO. Further consideration might also be given to adding a discussion contrasting cell signaling
and cellular biology of CO in general as derived from endogenous vs. exogenous sources.
6. The mode of action section in Chapter 5 presents information on both hypoxic andnon-
hypoxic mechanisms for CO health effects, with particular emphasis on recent studies
evaluating the non-hypoxic effects at low to moderate CO levels. Please comment on the
appropriateness of the focus, structure and level of detail in this discussion. For
example, is the evidence relating to the interaction between inhaled CO and endogenous
CO properly characterized?
The mode of action/mechanisms section of Chapter 5 provides a very important compilation of
highly diverse mechanistic studies on the interaction of CO with various biological systems. The
discussion of non-hypoxic mechanisms provides very interesting insights into the potential
pathophysiological pathways which may relate to specific outcomes such as angina, stroke and
inflammatory events, but the linkage of these mechanisms to biological responses and
morbidity/mortality is not clearly addressed. It would be appropriate to include an appraisal of
what information would be needed before these non-hypoxic mechanism outcomes would be
useful in setting the NAAQS. If non-hypoxic effects of CO are observed at environmentally-
relevant concentrations that pertained to myocardial ischemia and the resultant disturbances in
both membrane potentials and membrane permeability, the effects observed in patients with
cardiovascular disease and provide additional insights to the medical community. Thus, a better
13
-------
understanding of these phenomena could enhance the biological plausibility of the health effects
attributed to CO. We recommend that this area might be appropriately added to the future
research agenda.
Chapter 5 discusses the interaction between exogenous and endogenous CO, but more discussion
is needed in regard to how these interactions might produce health effects either via COHb as
well as direct biologic activities (non hypoxic mechanism). There might be a stronger focus on
whether these mechanisms might be a more important issue in susceptible populations. For
example, do these mechanisms play a stronger role in people with anemia? Would a better
understanding of molecular kinetics be helpful in improving the ability of the CFK equation to
predict COHb levels in exposed populations (the population demographics for susceptible
populations will be included in the risk assessment models)? It might also be useful to mention
that we still have a poor understanding of local, intracellular CO concentrations. There may be
high levels of endogenously produced CO in close proximity to heme oxygenase activity and the
added burden of exogenous CO could raise the available CO concentration to levels that could
induce inflammation or other non-hypoxic biological responses. A better understanding of the
local effects of endogenous CO production and how this will interact with intracellular CO in the
event of exposure to elevated ambient CO concentrations is needed. The impact of endogenous
CO production on COHb levels (hypoxic effect) is incorporated in the CFK equation 4.1.
Perhaps a statement could be made regarding factors other than hemolytic states that might lead
to significant levels of endogenous CO production.
7. Chapter 5 presents information on cardiovascular, central nervous system,
developmental, respiratory, and mortality outcomes following exposure to CO. To what
extent are the discussion and integration oftoxicological, clinical, and epidemiologic
evidence for these health effects scientifically sound, appropriately balanced, and clearly
communicated? Are the tables and figures presented in Chapter 5 appropriate,
adequate, and effective in advancing the interpretation of these health studies?
a. For cardiovascular outcomes, controlled human exposure studies discussed in
Chapter 5 and in previous assessments have identified cardiovascular effects in
diseased individuals following exposures near the level of the current standards,
while new epidemiologic studies provide evidence of cardiovascular effects at
ambient concentrations. What are the opinions of the Panel on the treatment of
factors influencing the interpretation of this evidence, such as the plausibility of
cardiovascular effects occurring at ambient levels, the additive effect of ambient
CO to baseline COHb resulting from endogenous and non-ambient CO, and the
challenge of distinguishing effects of CO within a multipollutant mixture (e.g.,
motor vehicle emissions) in interpreting epidemiologic study results?
14
-------
Chapter 5 lists pathologic and toxicologic evidence appropriately, but adequate integration of
these experimental study results with the epidemiologic evidence is still lacking. Priorities in
choosing the most important studies to discuss and criteria for how the evidence in these
different areas of research were assessed need to be made clearer for the reader. A more
comprehensive and transparent strategy should be used to assess the relative importance of
epidemiologic studies for making determinations in the ISA. Epidemiologic studies received
very different levels of attention and review. The brief mention of a large number of studies does
not address or present adequately the validity and relative weight of each study listed. Criteria
for identifying seminal studies should be made explicit. Important studies need to be described
in more detail. We also ask the authors to improve the organization of the material on the
pathologic/clinical and toxicological evidence and make it easier for the reader to grasp the
content of these studies; e.g., while a wide range of outcomes and different experimental
exposure protocols have been reported under 'developmental effects' (pages 5-83 on), it is hard
to understand this chapter without a summary table that facilitates reviewing the results in their
totality.
We also recommend adding an introduction at the start of each section that clarifies why it is
important to consider these health outcomes in this report. The tables and figures are generally
helpful, but they should provide more detail (e.g. sample size, study design, main biases,
exposure assessment methods, etc.). Furthermore, the information in figures and tables might
need to be re-organized; e.g. for Table 5.13 we recommend reorganizing either by trimester or
mean CO levels or outcome considered rather than country (some of these re-organized tables
could also appear in the appendix).
Finally, multipollutant-related issues need to be acknowledged upfront - they are currently
sprinkled throughout the chapter - and their implications for epidemiologic studies clearly stated
and discussed. This should include acknowledging the possibility that certain epidemiologic
studies may not be able to resolve the issue, especially when multiple pollutants are highly
correlated with CO because of common sources, i.e. multipollutant models may not provide
adequate answers.
b. Please comment on the implementation, in Chapter 5, of the causal framework
presented in Chapter 1. Does the integration of health evidence focus on the most
policy-relevant studies and health findings?
Generally, the health outcomes discussed in Chapter 5 are well chosen as the policy relevant and
important health outcomes to be considered in this report. We especially applaud that this report
addresses many innovative studies such as those in the area of fetal development and premature
birth. We expect that once Chapter 5 has been revised, the most 'policy-relevant studies' and the
'most important health findings' from epidemiologic studies can be better identified. We also
think that a re-organization might help clarify for the reader which health outcomes are
considered most policy relevant based on toxicological and clinical or epidemiologic study
results. The chapter should delineate which ones may affect the most people, affect people for
the greatest duration (such as a lifetime), or represent the most serious health events (such as
deaths or lifelong disability); and which ones are intriguing and likely quite policy relevant but
15
-------
need further study. In order to compare the epidemiologic study results among each other, it is
important that appropriate scaling factors for estimated effect sizes be used and that these
methods are presented with the appropriate clarity.
8. What are the views of the Panel on the discussion of factors affecting susceptibility and
vulnerability in Section 5.7?
Section 5.7 should be re-written. There is a need to place issues of vulnerability and
susceptibility relevant to CO in Section 5.7. These topics are included in many areas in Chapter
5, but not in Section 5.7. For example, the actual risks associated with CO for newborns/infants
were not stated. Generally there is a need for more explicit definitions of susceptibility and
vulnerability that are compatible with those used in other EPA documents. There is a clear
overlap in meaning of these terms.
Tables 5-18 and 19 require more detail with inclusion of 'at risk' groups. These tables currently
are 'shopping lists' of issues that might be relevant to CO. Some are key to understanding
vulnerability whereas others may have little or no relevance. Hence, a more careful listing of
factors is required.
Subheadings in Section 5.7 require further refinement or possible elimination. For example,
age/gender may not be the best categories as these variables may change CO uptake/elimination
rates and thus steady state COFtb.
16
-------
Enclosure C
Compilation of Individual Panel Member Comments
CAS AC Carbon Monoxide Review Panel on
CO Integrated Science Assessment (First External Review Draft, March 2009)
Dr. Milan Hazucha 18
Dr. Michael Kleinman 22
Dr. H. Christopher Frey 26
Dr. Russell R. Dickerson 35
Dr. Stephen R. Thorn 40
Dr. TomDahms 42
Dr. Paul T. Roberts 48
Dr. BeateRitz 55
Dr. Arthur Penn 61
Dr. Armistead (Ted) Russell 64
Dr. Laurence Fechter 69
17
-------
Dr. Milan Hazucha
(May 14, 2009).
This chapter is essentially an updated version of chapter 5 of 2000 AQCD with slightly
reorganized chapter headings and subheadings. I actually like this approach since it allows easy
back referencing of the material if one is interested in a more detailed presentation of the earlier
studies. The essential information from the 2000 document has been incorporated in the current
draft and merged well with the new findings.
Particularly section 4.3 has been expanded since cellular and molecular mechanism of CO has
been studied more extensively over the last decade. These studies have raised a number of
questions about potential interaction of biological effects due to these mechanisms and the
effects induced by exogenous sources of CO (addition, potentiation, etc.?) that may elicit or
enhance adverse health effects.
Charge Question 5: The dosimetry and pharmacokinetics of CO are discussed in Chapter 4.
Please comment on the presentation in the ISA of the current state of knowledge on the Coburn-
Foster-Kane (CFK) model and model enhancements. Has the expected contribution of different
exposure durations (1-24 h) to COHb levels been clearly and accurately conveyed?
The draft presents and discusses in a sufficient detail various forms of CFKE and their
limitations (4.2.1).
The Multicompartment Model section (4.2.2) covers all published models except for the most
recent one by Neto et al., J Braz Soc Mech Sci Eng 30/3:253-260, 2008. The multicompartment
models are more complex than CFKE but it is unclear how much more accurate they are
predicting venous COHb. While most of the input physiologic parameters for CFK model can be
relatively easy measured directly or estimated from a large data base, many of the parameters for
the multicompartment models must be estimated from a limited data base, which may lead to
wider predictive errors.
What I am missing is a brief discussion of the older mathematical models (Singh et al, 1991;
Sharan et al. 1990; Selvakumar et al, 1992). How does the predictive accuracy of these models
compare to CFKE and multicompartment models? Which one is the best over-all model if there
is such?
Since some models under predict while others over predict venous COHb it would be very
helpful as well as illustrative to develop a table/ graph comparing measured venous COHb values
obtained under, e.g., several typical dynamic ambient CO concentrations profiles over 12 hour
period (some older human studies provide such data) vs. predicted COHb under the same profile
employing "the most accurate" mathematical, CFK, and multicompartment model (Neto et al.,
2008, Smith et al, 1994, and Bruce and Bruce, 2006 as a suggestion).
18
-------
Section (4.2.3) discusses CFKE application under varying CO concentration and exposure
duration for a "healthy human at rest" (more detailed characterization is required). The
interaction is illustrated in Figure 4-3 (note that at 24 hrs COHb will reach equilibrium at any
CO concentration). Although accurately conveyed, there seem to be limits to the accuracy of this
model. The cited endogenous productions of 0.39% COFIb by QCP model (4.2.3) was measured
under basal conditions and is an underestimate of CO production at baseline conditions at rest,
the values reported in many studies. As discussed in section 4.5 endogenous CO production goes
up during oxidative stress, inflammation, pregnancy, in people with metabolic syndrome and
various diseases, the conditions when taken together will affect a majority of population. Under
these conditions the baseline COFIb value (endogenous production and possibly exogenous
sources) estimates are in the 1-2% range (Piantadosi, 2002). Smaller cohort studies also report
>1% COFIb level in healthy individuals. Hart et al, 2006 reports baseline mean % COHb value
for never smokers as 1.77 (n=547) for men and 1.53 (n=1901) for women. Thus, under increased
endogenous production of CO the model will proportionally overestimate venous COHB as the
time period and endogenous CO concentration increase.
The discussed models are designed to estimate venous COHb. However, the critical physiologic
endpoint is arterial COHb. Several human and animals studies have shown that breathing high
concentration of CO for a very short period of time will transiently increase arterial COHb to
levels well above the venous COHb. Among the first organs to see higher COHb is the heart and
the most active part of brain. Such, though brief exposures, may trigger pathologic response in
affected organs in at-risk individuals. Therefore, it is important to explore the capability of
COHb predictive models to predict accurately arterial COHb under transient exposure(s) to high
CO. Underground bus stations, heavy traffic in urban street canyons, and intersections, etc., may
create local environments when individuals will be transiently exposed to high CO. The issue of
peak CO concentrations, resulting transiently higher arterial COHb level and arterial-venous
COHb differences should be addressed in section 4.3.2.2.
Mass transfer of CO subsection (4.3.1.1) includes table 4-la (human) and 4-lb (mice) showing
CO cone, in different tissues, but for a brief sentence the relevance of these observations is not
discussed. Are these differences important? Are there important differences in distribution of CO
between human and mice tissues? Between tissues of other animal species? Any importance of
these differences for data extrapolation, etc.? Without addressing these questions what is the
point of having figure 4-lb? The same comments apply to page 4-13, lines 22-24. Moreover, the
statement on line 23-24 is incorrect since according to tables 4-la and 4-lb the distribution
among organs does not quite follow the same pattern and the relative concentration of CO
between tissues changes with increased ambient CO as well.
Figure 4-4: The source should be US EPA 2000 AQCD.
Subsection 4.3.1.2 Lung Diffusion of Carbon Monoxide should be expanded to include a
paragraph on changes in DLCO in disease though some of it is discussed in subsection 4.4.4. At
the end of paragraph the reader should be directed to the last paragraph on p.4-17 and section
4.4.4 for additional discussion on DLCO.
19
-------
Subsection 4.3.2.2 Blood. Throughout this Chapter and in other chapters as well it is assumed
that arterial and venous COHb are at equilibrium. However, what are the health consequences
when they are not at equilibrium, particularly during a rapid CO uptake? .What are the factors
affecting equilibrium? Moreover, a brief discussion of methodology of measuring COHb would
be helpful (CO-oximeter, Gas chromatography and others).
Subsection 4.3.2.4 Other Tissues, the statement on line 22-24 needs to be revised since the
distribution between organs changes with changing COHb level. Why we have a table 4-lb?
Explain.
At the end of the subsection 4.4.4 Health Status the reader should be directed to section 5.2
discussing cardiovascular effects.
The Endogenous CO production and Metabolism (4.5) has been substantially expanded as
compared to previous AQCD 2000. It is a nice comprehensive review. I suggest including in the
top paragraph on page 4-20 other diseases that increase endogenous production of CO like liver
disease, pulmonary hypertension, metabolic syndrome, and inflammatory diseases in general.
Cite the studies and provide measured endogenous CO, e.g., for asthma, allergies, drug-induced
increase in CO (e.g., Zocor reduces cholesterol), and others. If these various health conditions
are combined more than one half of US population will have elevated endogenous CO.
Page 4-19, line 25-26. True we do not know precisely what is the range of endogenous COHb
level (important parameter in COHb modeling) in the general population. However, numerous
studies suggest the baseline level range is 1-2% COHb. In disease population in can be higher.
This section should also include a discussion on differences and commonalities between the
effects due to endogenous CO and exogenous CO on cell metabolism, etc. The molecule is the
same but the effects may not be because there are other substances released during endogenous
production that may influence metabolic pathways.
On line 32 after Manno's reference insert a reference by Bos et al, 2006. The study provides
more updated findings on dihalothanes.
The Summary and Conclusions (4.6) should include a statement about which model is more
accurate or suitable and under what conditions (uptake, elimination) for COHb estimation. It
should also include a statement about increased production of endogenous CO in inflammatory
and other diseases.
Section 5.2 Cardiovascular effects.
This section presents numerous tables of epidemiologic studies for various CV outcomes
including long-term averages for CO. Most of the reported CO levels are at the range of
endogenous CO under basal conditions and almost all at the range of baseline COHb. If these
CO concentrations are taken at their values it is highly unlikely that they will induce any health
effects even in at-risk population. With peak CO values, which are physiologically the most
important, averaged over time how is one suppose to assess clinical significance of the findings?
The given mean CO values for these studies seem to be meaningless. An introductory paragraph
discussing the caveats in interpretation of these epi studies would be very helpful.
20
-------
Interpretation of multipollutant studies is similarly difficult without providing effect estimates
for all pollutants in the mix, e.g., CO, PMio and CO+PMi0. From my reading of CO ISA and PM
ISA the same studies are interpreted differently in each document. We cannot have it both ways
and the differences in interpretation need to be reconciled not only for CO and PM but for other
co-pollutants and their respective ISA as well.
21
-------
Dr. Michael Kleinman
9. The framework for causal determination presented in Chapter 1 was developed and
refined in other ISAs (e.g., the PM ISA). During previous reviews, CASAC generally
endorsed this framework in judging the overall weight of the evidence for health effects.
Please comment on the extent to which Chapter 1 provides necessary and sufficient
background information for review of the subsequent chapters of the CO ISA.
Chapter 1 clearly sets out the questions to be addressed in the NAAQS review (1-1). The
literature review was extensive and covered areas of epidemiology, toxicology and clinical
studies with an appropriate emphasis on elucidating the importance of exposure-response
relationships and modes of action. The chapter is very general in its approach and might
have been more CO-directed.
The EPA Framework for Causal Determination is clearly described in general terms (1-8)
however some expansion of the discussion to include specific reference to CO would be helpful.
For example the statement (pl-9, L16-17)" Data will not be available for all aspects of an
assessment and those data that are available may be of questionable or unknown quality" could
be amplified with which specific type of CO data might fall into this category.
The discussion of potential confounders could mention CO-specific confounders such as
environmental tobacco smoke and discussions of limitations of interpreting animal study data
could mention any relevant species-related differences that will be addressed in the later
chapters.
10. Chapter 2 presents the integrative summary and conclusions from the health effects
evidence, with the evidence characterized in detail in subsequent chapters. What are the
views of the Panel on the effectiveness of the integration of atmospheric science,
exposure assessment, dosimetry, pharmacokinetics, and health effects evidence in the CO
ISA?
Chapter 2 summarizes the conclusions drawn from the subsequent chapters. As such it
provides a roadmap of the critical junctures in the literature surveyed that influence the
causal determinations and the assessment of the strength of exposure-response relationships.
With regard to exposure there is some discussion of in-vehicle to roadside comparisons. It
would be helpful to mention the differences between the micro-scale (2-10m from raod) vs.
more distant (>70m from road). It would also be useful to mentions what % of the
population might be considered vulnerable because of near road or on-road exposures.
22
-------
It might be important to mention in the discussion of compensatory mechanisms (2-4, L6)
that individuals with cardiac or coronary artery disease might be unable or less able to
compensate. The point is made on p2-5 that they might not be able to endure compensatory
changes (which are defensive) but they also be unable to mount a defensive compensation
because of medication use or tissue damage.
It might be useful to discuss the Framework structure. The framework is hierarchical. If the
data are "inadequate" than one can not judge whether or not there is or is not a causal
relationship. Perhaps #4 should be suggestive of NO causal relationship and #5 should be
data are inadequate.
With regard to cardiac morbidity (2-7) a more explicit discussion is needed of the
uncertainties that lead to a designation of "likely to be causal" rather than "causal."
11. To what extent are the atmospheric science and air quality analyses presented in Chapter
3 clearly conveyed and appropriately characterized? Is the information provided
regarding CO source characteristics, CO chemistry, policy-relevant background CO, and
spatial and temporal patterns of CO concentrations accurate and relevant to the review of
the CO NAAQS?
The Chapter provides a comprehensive overview of the topics above. There are a few areas
that need to be more completely explained. For example, Fig 3-2 identifies on-road and non-
road engines as the major (-70%) of the CO emissions. However Figure 3-4 seems to
suggest that Region 1 emissions are ~2x those for Region 9 which includes S. California
where there are more cars than people (or so it seems).
12. How well do the choice and emphasis of exposure topics presented in Chapter 3 provide
useful context for the evaluation of human health effects in the ISA? Is the discussion
and evaluation of evidence regarding human exposure to ambient CO and sources of
variability and error in CO exposure assessment presented clearly, succinctly, and
accurately? The ISA concludes in section 3.7 that central-site monitor concentration is
generally a good indicator for the ambient component of personal CO exposure. What
are the views of the Panel on this conclusion and its supporting evidence?
The car/taxi data in Table 3-9 (5.7 ppm) should be contrasted with the in vehicle data Fig 3-
32 which shows that the in vehicle exposure is between 18 and 40 ppm. Is that a significant
consideration? The statement that measurement at a hot spot would "skew" community
exposure estimates upward is true but it begs the question of what part of the community is
being ignored. Perhaps it is worth discussing whether a population weighted average
exposure would be a more accurate parameter for use in late exposure-response estimations.
13. The dosimetry and pharmacokinetics of CO are discussed in Chapter 4. Please comment
on the presentation in the ISA of the current state of knowledge on the Coburn-Foster-
23
-------
Kane (CFK) model and model enhancements. Has the expected contribution of different
exposure durations (1-24 h) to COHb levels been clearly and accurately conveyed?
The CFK model is well described however the discussion of the sensitivity of the model to
uncertainties in the model parameters as a function of time (Fig 4-1) could be presented more
clearly. A concrete example(s) would be very helpful. If we pick an arbitrary fractional
sensitivity (i.e. FS = -0.5) and the parameter Vb, would it be correct to state that a +5% error
in the value of Vb used in the computation would result in a 10% underestimate of COHb at
10 min and 2 hr after exposure and a -5% error in Vb would result in a 10% overestimate of
COHb? It would also be very useful to include a table of the parameters and the range of
parameter values and uncertainties that would be used for specific estimates (as a function of
gender, age, body mass, etc.?)
14. The mode of action section in Chapter 5 presents information on both hypoxic and non-
hypoxic mechanisms for CO health effects, with particular emphasis on recent studies
evaluating the non-hypoxic effects at low to moderate CO levels. Please comment on the
appropriateness of the focus, structure and level of detail in this discussion. For example,
is the evidence relating to the interaction between inhaled CO and endogenous CO
properly characterized?
The discussion of non-hypoxic mechanisms provides some very interesting insights but the
linkage of these mechanisms to biological responses and morbidity/mortality is left as an
open question. It would be appropriate to include an appraisal of what information would be
needed before these non-hypoxic mechanism outcomes would be useful in setting the
NAAQS. This could possibly lead to some recommendations for future research. Similarly
the interaction between exogenous and endogenous CO is discussed but the way in which
these interactions can be incorporated into the definition of a NAAQS is not made clear.
There might be some focus on whether these mechanisms might be a more important issue in
susceptible populations. For example, do these mechanisms play a stronger role in people
with anemia? Another issue that might be addressed is in the area of toxicokinetic modeling.
Would the molecular kinetics be helpful in improving the ability of the CFK to predict COHb
levels in exposed populations (assuming the population demographics for susceptible pops
are including in the RA models).
15. Chapter 5 presents information on cardiovascular, central nervous system,
developmental, respiratory, and mortality outcomes following exposure to CO. To what
extent are the discussion and integration of lexicological, clinical, and epidemiologic
evidence for these health effects scientifically sound, appropriately balanced, and clearly
communicated? Are the tables and figures presented in Chapter 5 appropriate, adequate,
and effective in advancing the interpretation of these health studies?
Tables 5-4 through 5-9 would be more useful if they included a direction of change for the
endpoints and a level of significance. Because the section on health effects is long and very
24
-------
detailed it would be useful to have a table of key endpoints and whether or not there appears
to be a significant effect of CO.
a. For cardiovascular outcomes, controlled human exposure studies discussed in
Chapter 5 and in previous assessments have identified cardiovascular effects in
diseased individuals following exposures near the level of the current standards,
while new epidemiologic studies provide evidence of cardiovascular effects at
ambient concentrations. What are the opinions of the Panel on the treatment of
factors influencing the interpretation of this evidence, such as the plausibility of
cardiovascular effects occurring at ambient levels, the additive effect of ambient
CO to baseline COHb resulting from endogenous and non-ambient CO, and the
challenge of distinguishing effects of CO within a multipollutant mixture (e.g.,
motor vehicle emissions) in interpreting epidemiologic study results?
b. Please comment on the implementation, in Chapter 5, of the causal framework
presented in Chapter 1. Does the integration of health evidence focus on the most
policy-relevant studies and health findings?
It is not clear after the review of the epidemiologic, clinical and toxicological data why a
causal relationship for cardiovascular morbidity is "likely" rather than definite. Some
evaluation of what is still lacking to make that determination is needed. If the implication is
that there is never enough certainty to state that there is a causal relationship than perhaps the
framework should be restated.
Because the section on birth and developmental is long and very detailed it would be useful
to have a table of key outcomes and whether or not there appears to be a significant effect of
CO.
16. What are the views of the Panel on the discussion of factors affecting susceptibility and
vulnerability in Section 5.7?
This section would be strengthened if some demographic statistics were added to Tables 5-18
and 5-19. These are data that will factor into the risk analysis and this would be an
appropriate place to summarize them.
25
-------
Dr. H. Christopher Frey
I have prepared responses to charge questions 1, 2, 3, 4, and 8.
1. The framework for causal determination presented in Chapter 1 was developed and
refined in other ISAs (e.g., the PM ISA). During previous reviews, CASAC
generally endorsed this framework in judging the overall weight of the evidence for
health effects. Please comment on the extent to which Chapter 1 provides necessary
and sufficient background information for review of the subsequent chapters of the
CO ISA.
Chapter 1 is generally well-written, well-organized, and useful in content.
Section 1.6, EPA Framework for Causal Determination, is appropriately very similar, or in
places identical, the similar section in Section 1.6 of the Integrated Science Assessment for
Particulate Matter (First External Review Draft, December 2008). As EPA receives comments
on this material when reviewed by various Panels of CASAC, EPA should strive for consistency
across documents. The PM Review Panel offered several comments. Appropriately, "the
categorization reflects the strength of evidence and not the potential magnitude of public health
benefits." This implies that there is a distinction between weight of evidence and the potential
sensitivity or magnitude of the outcome . This distinction should be appropriately conveyed by
discussing both weight of evidence and the magnitude or sensitivity of each health effect
endpoint. A second point is that additional clarification regarding the terms "susceptible" and
"vulnerable" would be useful - the PM Review Panel provided detailed comments along these
lines, and for consistency these comments should be addressed across ISAs and REAs. A third
point is to consider the role that publication bias might have as it relates to making weight of
evidence determinations.
2. Chapter 2 presents the integrative summary and conclusions from the health effects
evidence, with the evidence characterized in detail in subsequent chapters. What
are the views of the Panel on the effectiveness of the integration of atmospheric
science, exposure assessment, dosimetry, pharmacokinetics, and health effects
evidence in the CO ISA?
The integrative summary and conclusions from the health effect evidence, presented in
condensed form, is extremely useful to the reader. In general, Chapter 2 is very useful, and
should be retained. It is very helpful to the reader to have this kind of "roadmap" as to the
bottom line policy-relevant state-of-the-science.
Table 2-1 could be modified to provide additional information regarding the weight of evidence
for each identified health effects endpoint, such as whether the finding is based on controlled
experiments, epidemiology, toxicology, or other, and a brief justification for the finding.
26
-------
On p. 2-2, line 5, it is stated that the 2002 National Emission Inventory (NET) is the most recent
data available. Perhaps that might have been true at the time that this material was drafted.
However, it would be appropriate to update to the 2005 NEI
(http://www.epa.gov/ttn/chief/net/2005inventorv.htmltfinventorydata) which is now available.
Acronyms should be spelled out the first time that they are used - e.g., ETS on p. 2-3, line 10;
CAD, on p 206, line 3.
Section 2.3.3, Birth Outcomes and Developmental Effects. This section is an example of the
need to carefully distinguish between weight of evidence and the magnitude or sensitivity of the
association. On p. 2-10, a statement is made that there is "weak evidence" of various adverse
effects. Presumably, this is a statement specific to weight of evidence. But is it also the case that
the magnitude of the effect is small? That is, is the intended mean that there is evidence of a
weak or small decrease? Are there cases in which a weak weight of evidence is also associated
with a small magnitude of effects? Section 1.6 might elucidate these kinds of situations and
offer clarification on the distinction between weak weight of evidence and small magnitude of
effects.
After reading Section 2.3.3 and 2.3.4, both of which have weight of evidence findings that are
"suggestive of a causal relationship," one might consider whether there is consistency in these
findings. Given that there are only 5 categories for weight of evidence, it is likely that there are
gradations within each category. Here, it appears that there may be a stronger case for birth
outcomes and developmental effects than for respiratory morbidity. Some comparative
assessment of the weight of evidence findings, and the strength of the associations, could be
useful.
It would help to have a "bottom line" summary of the overall assessment of the adverse effects
of CO at levels comparable to current air quality and to the current standard. It seems to be the
case that the document implies that the subsequent REA would focus on quantifying responses
based on controlled experiments, and that the epidemiological evidence tends to be weak,
associated with small effects, or confounded by co-pollutants. The chapter could offer a
synthesis and summary. For example, the current Section 2.4.1 seems to focus only on clinical
and epidemiological evidence with regard to the issue of concentration-response relationships. A
clearer summary could be offered regarding EPA staffs view of the way forward.
The identification of vulnerable subpopulations is of significant importance because it should
motivate areas of focus for exposure assessment in the REA. In particular, the relatively high
exposures associated with persons who spend time in or near traffic (roadways) and those who
exercise are of note.
3. To what extent are the atmospheric science and air quality analyses presented in
Chapter 3 clearly conveyed and appropriately characterized? Is the information
provided regarding CO source characteristics, CO chemistry, policy-relevant
background CO, and spatial and temporal patterns of CO concentrations accurate
and relevant to the review of the CO NAAQS?
27
-------
The information about Sources and Emissions of CO appears to omit some key information that
provides insight regarding conditions under which gasoline vehicles emit CO at high rates.
The statement that "Internal combustion engines used in mobile sources, by contrast, have
widely varying operating conditions and, thus, inherently higher and varying CO formation" (p.
3-1, lines 16-17) is not accurate. Internal combustion engines, also referred to as spark-ignited
(SI) engines, tend to have inherently high engine-out CO emission for reasons described below.
The reason that gasoline engines have higher uncontrolled emission rates than any other
combustion based source are because they typically operate close to the stoichiometric air-to-fuel
ratio, have relatively short residence times at peak combustion temperatures, and have rapid
cooling of the cylinder exhaust gases. The lack of excess oxygen means and the short
combustion residence time mean that carbon in the fuel is not fully oxidized to CO2, and CO
concentrations are approximately at equilibrium during the power stroke. The rapid cooling of
the exhaust gas means that the concentrations of free radicals, including the hydroxyl radical,
rapidly decline. As a result of this, it is not possible for CO to oxidize to CO2 fast enough during
cooling of the exhaust, leading CO levels to be "frozen" well above equilibrium values at any
given gas temperature in the exhaust. The very high "engine-out" CO concentrations motivate
the need for post-combustion, or end-of-pipe control, using an oxidation catalyst to promote
burnout of CO. A catalytic converter, or 3-way catalyst, serves this function, while also
oxidizing hydrocarbons and reducing nitrogen oxides.
Diesel engines have much lower engine-out CO emissions than gasoline engines because they
typically operate at very high air-to-fuel ratios. The presence of excess oxygen promotes mixing
between oxygen and the fuel, leading to improved burnout of carbon during the power stroke.
Furnaces, such as those in power plants, have much slower rates of flue gas cooling compared to
the rate of exhaust gas cooling in an internal combustion engine. Therefore, there is more time
for most of the post-flame CO to oxidize to CO2 by reaction with hydroxyl radicals, before the
concentration of the latter drops as temperature decreases.
An excellent reference that provides a scientific perspective on these issues is the textbook by
Flagan and Seinfeld on Fundamentals of Air Pollution Engineering, Prentice Hall, 1988.
Although this book is now out of print, it is far more rigorous and detailed than many more
recent texts.
There are two other key factors pertaining to CO emissions from gasoline-fueled vehicles that
should be mentioned: (a) cold start; and (b) fuel enrichment.
28
-------
A "cold start" refers to the time period after an engine start until which the catalytic converter
reaches its "light off temperature. The latter is the temperature at which the oxidation reaction
for CO becomes effective. Depending on the ambient temperature, the "soak" time (the time
since the most recent engine shutdown), and the design of the engine and exhaust system, the
duration of a cold start may be approximately one to three minutes. During a cold start, the
tailpipe emissions are as high as the engine-out emissions for CO. Cold starts are somewhat
more severe in cold weather than in warm weather, but can occur at any ambient temperature,
since the light-off temperature of the catalytic converter is substantially higher than ambient
temperatures.
Fuel enrichment refers to episodic situations during on-road operation in which there is high
power demand from the engine. Because the oxidation of CO to CO2 in the catalytic converter is
exothermic, and because high engine power demand is usually associated with high rates of
exhaust flow, the catalytic converter could overheat and become damaged. To prevent this, the
fuel-to-air ratio is increased, which leads to enhanced incomplete combustion and very low
levels of oxygen in the exhaust. Under these conditions, there is very little oxidation of CO to
CO2 by the catalytic converter, which prevents the catalyst from overheating, but leads to high
tailpipe emissions. Fuel enrichment episodes can occur for just a few seconds associated with
high accelerations, high speeds, high road grades, or combinations of these, combined with use
of accessories, or other sources of load such as having many passengers or cargo in the vehicle,
or towing a trailer. Although enrichment events occur on a vehicle-specific basis, it is possible
to have locations on a roadway network that are conducive to producing enrichment events for
many vehicles that pass through. An example would be freeway on-ramps, merges after tolls, or
accelerations that take place after a red light at a signalized intersection.
Because some of the key microenvironments of concern are near-roadway or in-vehicle, the ISA
should more fully and carefully explain the key factors that lead to episodes of high CO
emissions, particularly from gasoline-fueled highway vehicles.
Looking ahead to new vehicle technologies, a potential concern with hybrid electric vehicles
(HEVs) or Plug-in HEVs (PHEVs) is that, depending on their design, they can have many engine
starts and shutdowns during onroad driving. There is the potential that engine restarts could be
associated with a "cold start" effect if the soak time since the prior engine shutdown is long
enough. There are not yet good data on whether this effect is significant, and generally the
expectation is that HEVs and PHEVs will have lower average emission rates than comparably
conventional gasoline vehicles because they typically have significantly smaller engines.
29
-------
In addition to the role of the catalytic converter, the ISA should at least briefly summarize the
role of oxygenated fuels as a strategy for reducing engine-out, and tail-pipe, CO emissions from
vehicles. An example of an oxygenated fuel is ethanol. Although on average ethanol leads to
reductions in tailpipe CO emissions, it appears to lead to increased non-methane organic gas
(NMOG) emissions and slight increases in NOX emissions. The ISA might note that, historically,
here have been unintended consequences of the development and use of oxygenates for fuels;
notably, MTBE. MTBE has been found to be a persistent environmental pollutant, even leading
to problems associated with groundwater. Although the statutory mandate that underlies the
NAAQS does not enable EPA to take these cross-media and unintended consequences into
account, the lessons learned from such experiences can at least be summarized in the ISA.
The ISA should also give some attention to emerging trends, such as the potential for increased
use of biofuels. It is expected that biofuels, such as ethanol or biodiesel, would lead to reduced
tailpipe CO emissions since they are oxygenated. However, the reductions in total fuel life cycle
emissions, including fuel production and vehicle emissions, may be less than the reductions for
the tailpipe alone. Furthermore, there may be some geographic shifts in the location of CO
emissions, with some increases occurring in rural areas where biofuel production activities may
increase.
Regarding the discussion on the bottom of page 3-3, especially lines 15-19, the text should also
mention the finding, reported in NARSTO (2005), Chapter 7, page 200, that the MOBILE6
model correctly predicts the relative change in emission rates with respect to time (see the 2nd
column, top of column). Secondly, it may be too strong to infer that CO emissions are
overestimated by a factor of 2. The more correct inference is that the ratio of CO to NOX is
larger for the emissions inventory than for observed ambient concentrations, which could imply
that CO emissions are overestimated, NOX emissions are underestimated, or some combination of
both (See NARSTO, 2005, page 203). In particular, it is not clear that cold start emissions were
appropriately accounted for in the comparisons that conclude that the CO emissions are
overestimated by as much as a factor of two. For example, a tunnel study cannot provide insight
on this issue, since the location of the tunnel is typically sufficiently far away from the initiation
point of a trip that the vehicle would be in hot stabilized operating mode in the tunnel. NARSTO
(2005) also notes that some of the findings of previous studies were contradictory, citing in
particular a CRC (2004) tunnel study (this is probably the same as the Pollack et al. study cited
by EPA - both are reports by ENVIRON). Hence, the information contained in this paragraph
should be much more carefully interpreted. Although Parrish (2006) appears to reconcile the
contradictions in the previous study, there seems to be inadequate attention to the issue of cold
start, nor is there a plausible basis given as to why the CO emission inventory might be
overestimated.
EPA has recently released Draft MOVES 2009
(http://www.epa.gov/otaq/models/moves/index.htm). A "final" version of MOVES is currently
expected later this year, that would replace Mobile6. Draft MOVES 2009 is capable of
estimating highway vehicle CO emission rates taking into account a wide variety of driving
cycles, operating conditions, vehicle characteristics, and so on. The use of MOVES as a basis
30
-------
for estimating CO emission rates for highway vehicles, if such rates are needed to support
exposure modeling, should be considered.
p. 3-54, line 8, delete "does" and change "oxidize" to "oxidizes"
Section 3.3, chemical mechanism on top of page 3-4. This is very helpful. However, references
should be cited for the information provided on this page. As a matter of notation, HCHO is
perhaps a more common way to write the molecular formula for formaldehyde than CH2O. Free
radicals should have a "dot" - e.g., HC>2*.
Page 3-9, line 7, please use "such as" rather than "like."
Page 3-21, Figure 3-11, and similar figures. The highlighted counties (especially in yellow) that
are small in geographic area are very difficult to see on these maps. It may be necessary to add
pointers to such counties or to include a table listing all such counties in an appendix, just to
make sure that the information is conveyed completely.
Page 3-35 and related material. The analysis of the location and data for monitors in Pittsburgh
is interesting. Having lived in Pittsburgh for a number of years, I notice that monitor "A" is
located very close to the Ohio River, Monitor "B" seems to be in an urban canyon setting within
Pittsburgh's "Golden Triangle," and Monitor "C" seems to be close to roads and ramps that
represent major points of egress or ingress for the downtown area. Depending on wind direction
and time of day, Monitor "A" could be influenced by heavy traffic on the Fort Pitt bridge, and
perhaps by emissions exiting the bore of the Fort Pitt tunnel. However, the text attributes
variability among these three monitors to "mountainous" terrain. While there are hills on the
Northside (northern bank of the Allegheny River) and the Southside (southern bank of the
Monongehela River), the terrain in the immediate vicinity of the three monitors is not
significantly hilly. Not surprisingly, Monitor "A" is weakly correlated with Monitors "B" and
"C" (correlations of 0.43 to 0.52) probably because Monitor "A" is not in the downtown core and
the local wind conditions are likely to be highly influenced by the close proximity to the Ohio
River. Monitors "B" and "C have a correlation of 0.73, which is moderate, and is likely because
both are in the downtown core, for which there is likely to be very high correlation in traffic
conditions within the surrounding area that influences each of these two monitors. However,
given that these monitors are only 0.7 km apart, the correlation of 0.73 seems to indicate that
there local factors. One might hypothesize an urban canyon effect for Monitor "B" and perhaps
also some kind of near-roadway geometry effect for Monitor "C." Some discussion of the site-
specific nature of each monitor and the relative importance of various factors would provide
more insight into the variability between them, rather than the very brief discussion that ends
with a laundry list of factors on page 3.35 and lines 23-24.
Section 3.5.1.3 - page 3-39. This material is very important, especially given that near-roadway
and in-vehicle exposures are among the most important of the exposure microenvironments. It
would be useful to include some example graphs here to illustrate the concentration gradient as
one moves away from a roadway center or edge, with and without either a sound barrier or
vegetation, to support the material given on p 3-40, lines 5-18.
-------
Figure 3-22. This figure takes a lot of time to figure out. It would help if the figure panels were
labeled. The reader has to go back and forth between the legend and the caption to figure out
what each curve is.
Page 3-43, lines 1-3. The "laundry list" given here could be interpreted more specifically with
regard to the site being discussed. Rather than list many factors, which implies that they are all
equally important, is it possible to offer judgments as to which factors may be more important
than others?
Pages 3-48, 3-49. Please label the x-axes in each group or at least for the bottom most graphs.
General comment: while "diel" is a correct word to use, why not use "daily" instead?
Section 3.5.3 Associations with Co-Pollutants
Figure 3-28. Somewhat like Figure 3-22, these figures are not reader friendly. To avoid
confusion, these figures could be split into two separate groups of figures. The first group would
focus on correlations with other co-pollutants. The second group would focus on correlations
with different averaging times and forms of CO concentrations. Also, clarity is needed regarding
how correlations were calculated for data that seem to be of different averaging times - for
example, how does one get hourly PMi0 or PM2 5 concentrations if these are typically measured
using filter-based methods? Or are the comparisons to TEOM data? What are the sample sizes
associated with these comparisons?
The interpretation of Figure 3-28 given on Page 3-51 may not be correct. Figure 3-28 appears to
describe the inter-monitor variability in correlation coefficients (for what time period?). Not sure
what the figure caption means by "nationwide correlations" - shouldn't this be "variability in
correlations among national monitoring sites"? If the figure depicts variability in correlation,
then it is not correct to interpret the results as if they represent uncertainty. Variability refers to
real differences in values among members of a population; whereas uncertainty refers to lack of
knowledge regarding the true value of a quantity or distribution. One cannot infer whether
correlations are significantly different from zero by looking at a distribution of variability among
individual sites. The determination of the statistical significance of a correlation coefficient
depends on the magnitude of the correlation coefficient, the sample size of data upon which the
correlation coefficient was calculated, and the sampling distribution for random statistical error
in the estimate of the correlation for the individual site. Thus, while it might be true that a few of
the sites have correlations that are not significantly different from zero, the correlations of some
if not many of the sites were significantly different from zero for each and every season
considered. Hence, this entire paragraph needs to be carefully rewritten.
The ranges shown in Figure 3-28 are not confidence intervals. A confidence interval is inferred
from a sampling distribution. A sampling distribution is a frequency distribution for a statistic
based on random sampling error. The distributions shown here appear to represent variability
between monitoring sites. Hence, they represent frequency ranges.
32
-------
The paragraph of page 3-51, lines 1-21 should be divided into multiple paragraphs for clarity -
one paragraph should focus on the results and findings for CO and SO2, and then results and
findings for NOX, 63, PMio, and PM2.5 can be given in one or more additional paragraphs.
Page 3-51, line 24 - does this refer to daily CO concentrations and daily NO2 concentration?
Page 3-55, line 5: does this refer to area-wide or near-roadway CO concentrations?
Page 3-56, lines 20-29. The text here is a bit confusing because it is written as if the quantity
alpha (D) is defined in Equation (3-4). This quantity should be defined in a new equation for
clarity, and then discussed.
Page 3-63, lines 8-19. The text here appears to inaccurately describe the data reported by Abi
Esber and El-Fadel (2008). In their study, the did not measure "engine CO concentrations."
They measured the CO concentrations outside the vehicle - see Figure 1 of their paper which
provides a photograph of the "Out-vehicle air intake location." This needs to be corrected in the
text and in the caption for Figure 3-32.
Page 3-67. Lines 21-22, Regarding the statement about the possibility of community-to-
community differences in measurement errors, can a specific example be provided to support
this? i.e. is this of real concern or is there a specific reason to believe this is the case?
Page 3-67, line 32. Hydrocarbons are another co-pollutant, sometimes characterized as volitale
organic compounds, reactive organic gases, non-methane organic gases, and so on. These
include many species of compounds, including compounds identified as hazardous air pollutants
(HAPs) under the NESHAPs and as Urban Air Toxics. There are some compounds, such as
benzene, formaldehyde, 1,3-butadiene, and some others that are referred to a Mobile Source Air
Toxics (MSATs). These points should be introduced here. The co-emission of CO and HCs is
quite common, along with NOx and PM, from mobile sources. This section briefly mentioned
benzene and toluene on p. 3-68, line 28, but the co-varation in emissions and various classes and
species of HCs merits at least its own paragraph, if not a few paragraphs.
p. 3-74, line 13, there are repeated references to errors of the "Berkson type" - the first time this
is mentioned (earlier in the chapter) it should be defined and there should be citation to
reference(s).
p. 3-74, lines 6-7 versus lines 10-11 . It seems contradictory to state that there are significant
local factors leading to variability in exposures associated with proximity to roadways and then
to conclude that fixed site measurements are a good indicator of CO exposure. This apparent
contradiction should be resolved. Fixed site measurements are a poor indicator of exposure in-
vehicles or near roadways.
33
-------
4. How well do the choice and emphasis of exposure topics presented in Chapter 3
provide useful context for the evaluation of human health effects in the ISA? Is the
discussion and evaluation of evidence regarding human exposure to ambient CO
and sources of variability and error in CO exposure assessment presented clearly,
succinctly, and accurately? The ISA concludes in section 3.7 that central-site
monitor concentration is generally a good indicator for the ambient component of
personal CO exposure. What are the views of the Panel on this conclusion and its
supporting evidence?
In general, the exposure assessment material is well organized and appropriate. However,
central-site monitors are not a good indicator of CO exposure in microenviroments that are
influenced by local factors, such as in-vehicle and high proximity to roadways. Although this
point is acknowledged in various places, it does not seem to be consistently conveyed throughout
the document.
8. What are the views of the Panel on the discussion of factors affecting susceptibility
and vulnerability in Section 5.7?
Please see also the comments by CASAC and the PM Panel members on a similar section in the
1st draft of the PM ISA. Sometimes it is difficult to completely separate a susceptibility factor
from a vulnerability factor, and these situations should be acknowledged. For example, ability to
exercise, which is related to vulnerability, can be associated with nutritional status and other
factors related to susceptibility. This is not to say that the categories provided are incorrect;
merely to point out that it would be appropriate to acknowledge and characterize areas of
overlap. Another example is medication use, which is related to pre-existing disease. The tables
5-18 and 5-19 could more clearly indicate what are the key factors and what are surrogate
indicators of the factors. For example, is air conditioning really a surrogate for SES? Is
proximity to roadways a subset of geographic location?
34
-------
Dr. Russell R. Dickerson
The documents seem in general to be well researched and thorough. The Executive Summary
lacks punch, and the ISA would benefit from a list of top findings and recommendations. The
plan determining exposure of individuals for epidemiological studies looks sound, given
available observations. Most of the fundamental concepts concerning local air quality and global
atmospheric chemistry are at least covered. There are areas in which the ISA, and by inference
the Health Risk Plan, needs to evolve with the state of science. Comments on those follow.
Comments on the ISA.
As emissions from the American vehicle fleet decrease and the number of violations of the
NAAQS approach zero, it is time to both congratulate EPA and the State agencies for their
success and to reassess our approach to monitoring emissions and ambient concentrations of CO
as well as personal exposure. The existing network of CO monitors, designed to demonstrate
compliance with the current NAAQS, measure reliably, but with coarse resolution and
inadequate sensitivity most of the time. The ambient concentrations are more often than not
below the detection limit of the monitors. Section 3 of the ISA shows that there are insufficient
monitors for epidemiological studies. For example, at most sites the median concentration is
near the detection limit of the monitors used. This is recognized on page 3-25, but there is no
discussion of how to correct this problem.
CO is an important precursor to pollutant ozone and is useful tracer of vehicular emissions as
well as transport and mixing processes in the atmosphere. On the local scale, numerical
simulation of photochemical smog with models such as CMAQ can be effectively evaluated with
CO measurements. Because of the moderate lifetime (~ 1 month) and relatively simple
chemistry (loss by OH attack) CO offers a good tracer for evaluation of emissions and
meteorology in models. Boundary layer depth, for example impacts profoundly concentrations
of most pollutants, and if the models can capture the CO vertical profile then there can be more
confidence in their ability to capture mixed layer dynamics. Such studies require measurements
with greater sensitivity and resolution.
On page 3-11 the ISA states "The most sensitive trace-level versions of these instruments can
detect minimum CO concentrations of-0.04 ppm; the required lower detection limit for FRMs
in the EPA network is 1.0 ppm (40 CFR 53.20 Table B-l)." The issue of sensitivity of the
current and next generation of monitors deserves more attention in the ISA. There is mention of
NCORE, (not in the acronym list) but no details on the plans for superior monitors. Some
information is available on the EPA website:
http://www.epa.gov/ttn/amtic/files/ambient/monitorstrat/AAMS%20for%20SLTs%20%20-
%20FINAL%20Dec%202008.pdf
This is a little thin, but may still provide some guidance for planning. My understanding is that
this network will go into effect in 2011, and the ISA should discuss these plans and how they
relate to the environmental and health effects of CO.
35
-------
With increasing attention being paid to local and global climate change, a better understanding of
the global atmospheric chemistry of CO has become increasing important. This relates to the
need for a secondary standard for CO. The role of CO as an important local and global sink for
OH is mentioned in Section 3.3; the ISA should call for monitoring with sufficient sensitivity, in
other words new or modified instruments. This is no great technological challenge.
The mean global concentration of CO decreased through the 1990's but appears to have leveled
off [Duncan and Logan, 2008; Duncan et al, 2007]; see also Novell!, 2008.
http://www.esrl.noaa.gov/gmd/publications/annmeet2008/Poster Final.pdf
Because emissions from sources in the US have decreased does this imply that emissions from
the rest of the world have increased? Is that an environmental hazard for the US? The ISA
should have a section on consideration of a secondary standard for CO as promised on page 1 of
the "Plan for Health Risks", but I cannot find one.
The literature since the 2000 CD has been reviewed reasonably well, but there have been a series
of studies that support the contention that vehicular emissions have decreased considreably and
that MOBILE 5 and 6 overestimate emissions substantially. For example [Pokharel et al., 2002;
Pokharel et al., 2003] demonstrate improvements in tailpipe exhaust of CO for several American
cities. There is also evidence of improvements in the Diesel truck fleet emissions [Burgardet
al., 2006]. The observations ofParrish (2006) have been verified [Bishop and Stedman, 2008].
See also Stedman et al. (2009). These results have implications for Inspection and Maintenance
Programs as well as for numerical modeling of emissions.
On page 5-126 is stated "Because CO measurements tend to reflect more local impacts, due to
the location of monitors, than NO2 (which is a secondary pollutant and therefore more spatially
uniform) it is also possible that CO, the less precisely measured pollutant in terms of spatial
distribution, may "lose" in the multipollutant model. Thus, it may not be accurate to interpret
these results as evidence of 'confounding by NO2.'" Of these two pollutants CO is more
spatially uniform. NO2 is secondary only in the sense of it being formed from NO within the
first minutes of emission. The lifetime of NO2 is less than a day while that of CO is more than a
month.
Section 3.2 states that less CO is produced at higher burn temperatures, but thermodynamics
dictate that a fair amount of CO is formed from CO2 decomposition and that the equilibrium
favors CO and /^O2 at higher temperatures, especially in internal combustion. The remainder of
the para is good.
The color scale of Figures 3-11 and 3-12 is inappropriate - there are only two of the five colors
visible. Correlations of CO and ozone can be misleading - CO is a precursor for ozone but some
ozone is titrated out by NO that is co-emitted with CO. Both Os and SO2 tend to peak in the
middle of the day so 24-hr means of trace gas concentrations might reveal more wrt atmospheric
chemistry.
36
-------
References
Stedman et al., On-Road Motor Vehicle Emissions including Ammonia, Sulfur Dioxide and
Nitrogen Dioxide 19th Coordinating Research Council, On-Road Vehicle Emissions
Workshop, San Diego, CA. March 23-25, 2009.
Dalton, T.R., D.H. Stedman and J.D. Ray, Winter Motor-Vehicle Emissions in Yellowstone
National Park, G.A. Bishop, D.A. Burgard, Environ. Sci. Technol, 40:2505-2510, 2006.
Bishop, G. A. and D. H. Stedman (2008), A decade of on-road emissions measurements,
Environmental Science & Technology, 42, 1651-1656.
Burgard, D. A., G. A. Bishop, D. H. Stedman, V. H. Gessner, and C. Daeschlein (2006), Remote
sensing of in-use heavy-duty diesel trucks, Environmental Science & Technology, 40,
6938-6942.
Duncan, B. N. and J. A. Logan (2008), Model analysis of the factors regulating the trends and
variability of carbon monoxide between 1988 and 1997, Atmospheric Chemistry and
Physics, 8, 7389-7'403.
Duncan, B. N., J. A. Logan, I. Bey, I. A. Megretskaia, R. M. Yantosca, P. C. Novelli, N. B.
Jones, and C. P. Rinsland (2007), Global budget of CO, 1988-1997: Source estimates and
validation with a global model, Journal of Geophysical Research-Atmospheres, 112.
Pokharel, S. S., G. A. Bishop, and D. H. Stedman (2002), An on-road motor vehicle emissions
inventory for Denver: an efficient alternative to modeling, Atmospheric Environment, 36,
5177-5184.
Pokharel, S. S., G. A. Bishop, D. H. Stedman, and R. Slott (2003), Emissions reductions as a
result of automobile improvement, Environmental Science & Technology, 37, 5097-5101.
Further Comments on the Carbon Monoxide ISA and the Health Effects Plan
21 May 2009
Carbon monoxide is more than a primary pollutant. It is a major precursor to ozone and alters
the oxidizing capacity of the atmosphere. Because the atmospheric chemistry of CO is relatively
simple and well known, CO makes an excellent tracer for polluted air masses and is useful for
evaluating air quality models such as CMAQ. Some examples are given in the references at the
end. The lifetime is long relative to synoptic events, and the concentration of CO over the US is
driven primarily by boundary conditions, emissions (and in situ formation), and transport.
Agreement of observed and modeled temporal trends in surface CO concentrations indicates that
advection and mixing are appropriately simulated. Altitude profiles of CO can be a powerful
tool for determining how well vertical mixing is represented in a model, and vertical mixing is
critical to understanding the spatial and temporal variability of ozone and PM. In order to
evaluate modeled vertical profiles, modeled emissions must be correct, thus both high-resolution
measurements and reliable emissions estimates are necessary for evaluating chemical transport
models used for air quality planning.
To make a quantitative recommendation, with typical concentrations of about 200 ppb CO,
precision in the observations of 10 % or about 20 ppb is desirable for evaluation of numerical
37
-------
simulations. Uncertainty on the order of 20 ppb is achievable with modified commercial
instruments, and may be possible with careful operation of newer detectors as delivered, but this
remains to be demonstrated.
Concerning recent measurements of CO there is room for improvement, but also substantial
confusion on the accuracy and precision of the monitors in use. Newer instruments may have
superior sensitivity, but the detectors vary in age. For example, the State of Maryland operates
two instruments with a detection limit around 20 ppb, and Georgia apparently operates several
similar instruments, but such high sensitivity is not an EPA requirement. High resolution
monitoring is a State initiative and therefore subject to substantial variability across the US.
The ISA should make an attempt to compile information on precision, accuracy and uncertainty
of the measurements, especially where long term averages are driven by high values. Some new
instruments are improved substantially. (For the record the gas-filter correlation detectors of
which I am aware are NDIR). High sensitivity NDIR analyzers are available from Thermo
Scientific (Model 48CHL), Teledyne Instruments (Model 300E), and Environment S.A (Model
CO12M), and all report a lower detectable limit of about 0.040 ppm. All manufacturers also
reported precision or zero drift of about 0.10 ppm. It may be worth while for EPA to evaluate
these instruments, but given the reported specifications, these instruments off the shelf (as
purchased) will not provide adequate sensitivity when the typical concentration is 200 ppb (0.2
ppm). It has been possible in the past to use frequent checks of the zero point to improve the
accuracy of commercial CO detectors. New instruments may be amenable to this procedure.
How can the actual dose of CO be determined? As reviewed in the ISA, street canyon level
models exist. They must however be evaluated with high precision observations. This will
require multiple measurement points for the course of a few days because the diurnal (diel)
patterns are important.
As reported in the ISA, MOBILE6 appears to overestimate CO emissions. NOx emissions are
probably overestimated too, although by not as much. How will MOVES improve upon this?
There is at lease one report that MOVES calculates lower VOC's (and presumably CO), but
higher NOx emissions than MOBILE 6.
http://www.marama.org/cal endar/events/2009_02 Annual, html
For the abatement of the global-scale adverse effect of excess CO on atmospheric composition
and climate, an emissions limit rather than an ambient concentration standard is appropriate. The
ISA should consider a discussion of such a limit; what for example would be the total American
CO emissions if all on-road vehicles meet the current emissions standards? How would this
change if all the non-road vehicles and stationary internal combustion engines were regulated to
the same level? The IPCC reports estimates the radiative forcing (on the decadal time scale) due
to CO from which one can estimate the CO2-equivalent impact on climate, The ISA
could/should discuss the science behind pursuing such a goal and the appropriate credit the US
should get for greenhouse forcing avoided.
38
-------
References
A global simulation of tropospheric ozone and related tracers: Description and evaluation of
MOZART, version 2 Author(s): Horowitz LW, Walters S, Mauzerall DL, et al. Source:
JOURNAL OF GEOPHYSICAL RESEARCH-ATMOSPHERES Volume: 108 Issue:
D24 Article Number: 4784 Published: DEC 24 2003
Convective transport of biomass burning emissions over Brazil during TRACE A
Author(s): Pickering KE, Thompson AM, Wang YS, et al.
Source: JOURNAL OF GEOPHYSICAL RESEARCH-ATMOSPHERES Volume: 101 Issue:
D19 Pages: 23993-24012 Published: OCT 30 1996
Nighttime chemistry in the Houston urban plume, Luria M, Valente RJ, Bairai S, et al. Source:
ATMOSPHERIC ENVIRONMENT Volume: 42 Issue: 32 Pages: 7544-7552
Published: OCT 2008
39
-------
Dr. Stephen R. Thorn
Ch 1: Please comment on the extent to which Chapter 1 provides necessary and sufficient
background information for review of the subsequent chapters of the CO ISA.
The approach and background are well done. Introduction of the problem with absence of
alternate dose indicators (vs COHb) is important.
Ch 2: What are the views of the Panel on the effectiveness of the integration of atmospheric
science, exposure assessment, dosimetry, pharmacokinetics, and health effects evidence in
the CO ISA?
Again, well done.
Ch 3: To what extent are the atmospheric science and air quality analyses presented in
Chapter 3 clearly conveyed and appropriately characterized? Is the information provided
regarding CO source characteristics, CO chemistry, policy-relevant background CO, and
spatial and temporal patterns of CO concentrations accurate and relevant to the review of
the CO NAAQS?
They are standard facts - well done.
How well do the choice and emphasis of exposure topics presented in Chapter 3 provide
useful context for the evaluation of human health effects in the ISA? Is the discussion and
evaluation of evidence regarding human exposure to ambient CO and sources of variability
and error in CO exposure assessment presented clearly, succinctly, and accurately? The
ISA concludes in section 3.7 that central-site monitor concentration is generally a good
indicator for the ambient component of personal CO exposure. What are the views of the
Panel on this conclusion and its supporting evidence?
Issues are well presented and conclusion is valid.
Ch 4: Please comment on the presentation in the ISA of the current state of knowledge on
the Coburn-Foster-Kane (CFK) model and model enhancements. Has the expected
contribution of different exposure durations (1-24 h) to COHb levels been clearly and
accurately conveyed?
The discussion is well developed. I have only a small issue. I believe there is an error in Table 4-
la. Even in the original publication ther was confusion as to the units on the table, but I believe
the CO concentration should be in pmol/mg (NOT 100 g ww tissue).
Ch 5: Please comment on the appropriateness of the focus, structure and level of detail in
discussion on hypoxic and non-hypoxic mechanisms of CO health effects. For example, is
the evidence relating to the interaction between inhaled CO and endogenous CO properly
characterized?
40
-------
Once again, the authors did a very good job. There is an obvious concern pertaining to
compounding the effect of endogenous CO with an exogenous (inhaled) source. It might make
some sense to introduce the concept that we really still have a poor understanding of the local,
intracellular CO concentration in close vicinity to heme oxygenase activity. Therefore, the
proportionate effect of exogenous CO and how much this will alter intracellular CO
concentrations requires more study.
Chapter 5 presents information on cardiovascular, central nervous system, developmental,
respiratory, and mortality outcomes following exposure to CO. To what extent are the
discussion and integration of toxicological, clinical, and epidemiologic evidence for these
health effects scientifically sound, appropriately balanced, and clearly communicated? Are
the tables and figures presented in Chapter 5 appropriate, adequate, and effective in
advancing the interpretation of these health studies?
I believe they are - well done.
For cardiovascular outcomes, controlled human exposure studies discussed in Chapter 5
and in previous assessments have identified cardiovascular effects in diseased individuals
following exposures near the level of the current standards, while new epidemiologic
studies provide evidence of cardiovascular effects at ambient concentrations. What are the
opinions of the Panel on the treatment of factors influencing the interpretation of this
evidence, such as the plausibility of cardiovascular effects occurring at ambient levels, the
additive effect of ambient CO to baseline COHb resulting from endogenous and non-
ambient CO, and the challenge of distinguishing effects of CO within a multipollutant
mixture (e.g., motor vehicle emissions) in interpreting epidemiologic study results?
The document authors have handled discussion on factors influencing the interpretation of
cardiovascular risk in a fair and balanced manner. I think the data support caution and concern
that there is indeed a cardiovascular risk at near-ambient CO concentrations for individuals with
coronary vascular disease.
Please comment on the implementation, in Chapter 5, of the causal framework presented in
Chapter 1. Does the integration of health evidence focus on the most policy-relevant
studies and health findings?
The framework is logical and coherent.
What are the views of the Panel on the discussion of factors affecting susceptibility and
vulnerability in Section 5.7?
The authors have done an extremely good job.
41
-------
Dr. Tom Dahms
Chapter 3: Source to Exposure
General Comments
This chapter is very important to the ISA in that it provides the basis for the understanding of CO
sources, trends in CO levels from sources and exposures of both populations and individuals.
Sections 3.1 through section 3.4.1 are well constructed and provide an overview of sources and
trends in atmospheric CO over the past 2 decades. This material contains detailed information
along with sufficient interpretive information to provide the reader with consensus findings.
Given that much of the recent literature that pertains to health effects of CO is based on
epidemiological data, this chapter emphasizes the value of atmospheric monitoring data as the
best estimate of exposure to CO for the epidemiologist. . If one scans the Figures and Tables in
Chapter 5. Integrated Health Effects, data has been compiled from major cites around the world.
Health effect end-points are being assessed relative to atmospheric changes in CO collected from
urban networks of monitors. No insight is provided (in Chapter 3 or in Annex A) to help the
reader understand the validity of these international atmospheric monitoring systems. Given the
significant reduction in atmospheric levels of CO since 1980, these international studies are
important to the understanding of potential effects of CO. Therefore some means of altering the
material in this Chapter should be undertaken to aid with the improved understanding of these
international studies.
Charge question 4.
A. How well do the choice and emphasis of exposure topics presented in Chapter 3
provide useful context for the evaluation of human health effects in the ISA?
Given that much of the recent literature that pertains to health effects of CO is based on
epidemiological data, this chapter presents the case for atmospheric CO from fixed site monitors
as the best estimate of exposure to CO. If one scans the Figures and Tables in Chapter 5.
Integrated Health Effects, data presented in this ISA has been compiled from major cites around
the world. Health effect end-points have been assessed relative to atmospheric changes in CO in
those international locations. Unfortunately no insight is provided (in Chapter 3 or in Annex A)
to help the reader understand the validity of these international atmospheric monitoring systems.
Some background information on these international sites would be of value.
The emphasis of this chapter discussion is placed on establishing the validity of the use of
ambient monitoring information as the best available means for estimating exposure to CO.
Although the information is presented in some detail regarding the variability of exposure due to
individuals moving through different microenvironments, the variability due to these personal
exposures is discounted in its importance. This is a considerable deviation from the
consideration of these issues in previous AQCD for CO where considerable concern was placed
on the use of exposure models like pNEM/CO (APEX) that placed emphasis on personal
exposure. Given the information reviewed in this Chapter for the potential for individual
42
-------
variability in exposure to CO due to activity patterns, geographic or spatial locations (e.g. in
transit, proximity to roadways), it is very difficult to accept the premise that atmospheric
monitoring data provides the best means of assessing exposure. Data from fixed site monitors is
probably the best approximation of assessing exposure for epidemiological studies, however the
limitations clearly need to be clearly stated.
B. Is the discussion and evaluation of evidence regarding human exposure to ambient
CO and sources of variability and error in CO exposure assessment presented
clearly, succinctly and accurately?
Beginning with section 3.4.2 the presentation of the information changes to what often reads as a
long string of facts. This appears to be due to the attempt to mention so many of the recent
studies in this area without any concluding sentence that would justify the inclusion of the listed
material. The result is confusion regarding the intended focus of the information being presented.
For example on page 3-13 lines 4-6 indicates that data will be presented to determine if ambient
monitors adequately characterize population exposure. Information is presented but no
conclusion is drawn from the information in this paragraph. However in the remainder of the
Chapter, material is presented that suggests that fixed site monitors are good for estimating
exposure to ambient CO. It would be logical to make this statement early and then proceed to
defend the statement.
The Chapter would be much more readable if it had been more focused and carefully edited with
the intended reader in mind. For example many paragraphs do not have a topical sentence that
indicates to the reader what is to follow. The Chapter contains excessive jargon that makes it
difficult to follow for example: "In the context of determining the effects of ambient pollutants
on human health, the association between the ambient component of personal exposures and
ambient concentrations is more relevant than the association between total personal exposures
(ambient component + non-ambient component) and ambient concentrations. " The meaning of
ambient seems to be clear to the author but not always to the reader so there needs to be a brief
definition of terms—what is a non-ambient component that is presumably inhaled? I had to
assume that ambient in this context is taken to mean atmospheric levels of CO found in the
outdoor air at distances from the surface available for humans to inhale. Yet, the potential
exposures shown in Figure 3-31 on page 3-58 show that there are a wide variety of ambient
conditions that are all mixtures that are composed of some percentage of atmospheric levels of
CO. This material could benefit from editing with an eye toward making the material clear to
readers from a diverse scientific background.
Chapter 3 contains a significant amount of atmospheric monitoring data that appears to be
focused on justifying the use of atmospheric measurements of CO for purposes of assessing
exposure to CO. This reviewer is not an expert in modeling of atmospheric CO but the
information with which I am familiar seems to have been accurately described. It appears that the
author(s), in the attempt to be comprehensive, developed very little in depth information directly
focusing on health effect exposures to CO. This assessment has not been clearly nor succinctly
presented. For example the material found in pages 3-10 to 3-54 concerns the intricacies of
atmospheric monitoring and modeling. The application of the specific points of atmospheric
43
-------
monitoring need to be related to CO exposure assessment as a focus of the presentation of this
material.
Not being an expert in the area of atmospheric monitoring, I have checked some of the
references and the assertions in the text agree with the author's conclusions. Therefore I assume
that the material presented is accurate.
C..The ISA concludes in section 3.7 that central-site monitor concentration is generally
a good indicator for the ambient component of personal CO exposure. What are the
views of the Panel on this conclusion and its supporting evidence?
In the 1991 AQCD for CO on page 8-79 the following statement was made: "The authors
concluded that fixed outdoor CO monitors alone are, in general, not providing useful estimates
of CO exposure of urban residents" This statement was made on the basis of COHb
measurements from the NHANES II study where only 0.03% of the variance in COHb was due
to ambient CO data.
In the 2000 AQCD for CO the following summary statement was made: "Fixed-site monitors
often are used in urban areas to estimate the ambient concentrations to which individuals in the
surrounding areas may be exposed. These measurements tend to overestimate 8-h exposure
values for people living in areas of lower traffic and underestimate the exposure of people living
in areas of higher traffic. " This conclusion was reached based on the evidence from personal
exposure monitors and from the analysis of various micro-environments that showed levels of
CO not detected by atmospheric monitors. The evidence for this statement was not based on any
analysis of dose of CO (COHb) as was the 1991 statement.
The specific statement regarding the above question is found in the 2009 draft ISA_CO is found
in Section 3.6.5.3.page 3-65 lines 7-10 which is repeated in section 3.7.5 page 3-741ines 10,11.
"For the general U.S.population, exposure error analysis for epidemiologic studies indicates
that fixed-site measured ambient CO concentration is generally a good indicator of ambient
exposure to CO, as discussed in more detail below" The evidence that seems to have influenced
these statements can be found in Section 3.6.2, page 3-57 lines 15-31 where the study of Wilson
and Brauer (2006) is used as evidence. As noted this study was on 16 subjects who were studied
for exposure to PM. It is not clear what assumptions are involved in accepting the transference of
PM exposure to CO exposure.
Wallace and Ziegenfus (1985 in the 1991 AQCD for CO) actually tested the hypothesis that the
fixed outdoor CO monitors provide useful estimates of CO exposure. As noted above the data of
Wallace and Ziegenfus does not support this hypothesis. The current ISA Chapter 3 reviews data
that reaches the opposite conclusion but without data from persons exposed to CO to support the
hypothesis. If this assertion of the value of fixed-site monitors is to be convincing, the study of
Wallace and Ziegenfus needs to be carefully analyzed to show why it should be discarded.
44
-------
Charge Question 6.
The mode of action section in Chapter 5 presents information on both hypoxic and
non-hypoxic mechanisms for CO health effects, with particular emphasis on recent studies
evaluating the non-hypoxic effects at low to moderate CO levels. Please comment on the
appropriateness of the focus, structure and level of detail in this discussion. For example, is
the evidence relating to the interaction between inhaled CO and endogenous CO properly
characterized?
Although the roles for CO in signaling are increasing at a rapid pace, the application of this
information to understanding the adverse health effects of CO is limited. The information
presented in Chapter 5 reflects this situation and I believe is very appropriate in scope and focus.
There are many obstacles to applying the non-hypoxic effects of CO to the health effects data
base. The biggest hurdle to date is that the adverse health effects of CO observed in patients with
coronary artery disease occurred with partial pressures of CO of 0.012-0.015 torr or 15-20 ppm.
There are very few non-hypoxic effects observed with exposures in this range. One reason is that
most of the animal models or tissues studied are healthy i.e., not from animals that in any way
mimic the cells from ischemic heart tissue or other disease models. Another aspect of these
studies lie in the difficulties of attempting to reproduce endogenous release of CO from focal
distribution of heme-oxygenase with global exposures to CO. It is hoped that eventually this
field will develop approaches and methods that will lend themselves directly to addressing health
effects.
The limited similarity to endogenous production of CO and exogenous CO are well described in
section 5.1.3.3.
The impact of endogenous CO production on COHb levels (hypoxic effect) is incorporated in the
CFK equation 4.1. Perhaps some statement could be made regarding the factors, other than
hemolytic states, that might lead to significant levels of endogenous CO production.
If non-hypoxic effects of CO were observed that pertained to myocardial ischemia and the
resultant disturbances in both membrane potentials and membrane permeability, the effects
observed in patients with CAD might be less skeptically received by the medical community.
There are some encouraging studies along these lines: Thorn and Ischiropoulos showed that 10
ppm CO resulted in increased free NO along with other studies indicating that exposure to CO
results in increased concentrations of ROS.
Concern over study quality.
One of the outcomes of the discussion of the epidemiology data in the ISA regarding effects of
CO on cardiovascular end points, was a request that was made by Dr. Ritz for the authors of the
45
-------
ISA to provide information that pertained to study quality. In Chapter 1 lines 1 and 2 state that
the ISA is meant to be "a concise evaluation and synthesis of the most policy relevant science for
reviewing the NAAQS." In my view this document falls short on the evaluation aspect of the
studies reviewed. This document seems to be more of a compendium of current publications
without critical analysis of the information presented. Inclusion of this analysis would enable the
reader to determine which studies should take precedence or have the most influence in
supporting the conclusions drawn from the review of the literature. In fact this approach should
be utilized throughout the document. Without such evaluation of the material presented, the
reader can only drawn conclusions from the number of studies (presumed to be equal in quality)
showing effects vs those that do not show health effects. For example how many of the studies
have insufficient power to avoid a Type II error?
The consensus of the review panel based on the information in the ISA, the data that provides the
strongest evidence for support of the health effects of CO are the controlled human exposure
studies and the epidemiology studies. These two groups of studies clearly warrant the closest
evaluation in the ISA.
During the recent CASAC_CO Panel meeting, EPA staff raised the issue in several ways that
suggest a reluctance to base decisions on study quality. This point was emphasized by Dr. Ritz's
comments that she found it difficult to evaluate the epidemiological data because of the lack of
presentation of quality indicators for each study in the ISA document. The interpretation of the
exchange between Dr. Ritz and staff is that greater consideration of the data from studies of
higher quality should go into the evaluation of effects. There seems to be a tendency in the
document to look for multiple studies confirming the same effect and to use the median or range
of effects observed from those studies. The ideal situation is to look for multiple high quality
studies showing the same effects at the same level of exposure. Without belaboring the point, a
series of studies designed with inadequate power to show intended effects and therefore showing
no effect are just as dangerous. There needs to be more of an attempt to identify studies in all
areas of the CO database where quality indicators are identified and reliable data published. In
short there needs to be more critical evaluation of studies presented in stead of what currently
exists as a serial presentation of information with little insight into relative importance of the
studies presented.
Having been part of the multicenter Allred et. al. study using controlled exposure of high risk
subjects, it has become clear to me that such analysis is lacking not in just the epidemiological
data base but in other areas as well. I will review what went into the Allred study as an example
of what I think is important for producing a defensible scientific basis to support a standard. This
does not detract from other studies that have confirmed these findings but there is a clear
difference in the studies.
It was determined in the mid 1980s that the data produced by Aranow could not be relied upon
for reasons of scientific misconduct and that this data was a key piece of the basis for the 1979
NAAQS for CO. The proponents of loosening the NAAQS for CO had legal grounds for a
challenge. Therefore a study had to be designed and carried out to test the Aranow hypothesis in
such detail and with unquestionable quality assurance standards that the findings would clearly
46
-------
test the hypothesis to everyone's satisfaction. To accomplish this task required considerable
resources not normally made available to a group of investigators. However the NAAQS for CO
seemed to hinge on this study. (In fact after the release of these findings congressional staff
made the point of acknowledging that legislation in support of all NAAQS was altered because
of the findings of this study.) What set this study apart were the following key elements: its
multicenter nature, sound a priori statistical design, multiple dose design to provide potential
dose-response effects, well characterized subjects, tight exposure and dose controls, and audited
quality assurance. These features provided all parties with the assurance that the findings would
be defensible in any legal proceedings. There are other studies that confirm these findings but
without the Allred et al study, they would be subject to criticism because one or more of the
elements listed above were missing.
In away this study has been acknowledged by the intended levels of COHb that will be used in
the Risk Assessment.
Without going into great detail the authors of the ISA should provide the guidance requested by
Dr. Ritz and others as to standards that set studies apart from others in their findings.
47
-------
Dr. Paul T. Roberts
ISA Charge Question 3. To what extent are the atmospheric science and air quality analyses
presented in Chapter 3 clearly conveyed and appropriately characterized? Is the information
provided regarding CO source characteristics, CO chemistry, policy-relevant background CO,
and spatial and temporal patterns of CO concentrations accurate and relevant to the review of the
CO NAAQS?
In general, the discussions in ISA Chapter 3 on CO source characteristics, CO chemistry, and policy-
relevant background CO are accurate and relevant to the CO NAAQS review. Most of my comments
are on limitations/qualifications of the measurement data and the use of that data; see comments below.
Regarding Figure 3-2 (and Figure 3-4) and associated text: I realize that inventories older than 1990
are not comparable to more-recent inventories, but it is hard to properly compare emissions trends from
1990-2002 with air quality trends from 1980-2006 as is done in Chapter 3.5.2.1. I think that the 2005
NEI inventory is now available and should be used to at least partly update this comparison.
On line 13, page 3-5: it would be good to convert the 30 Tg to MT for the reader to also see the
comparison with other emissions data in this part of the ISA.
Regarding Chapter 3.4.1 Ambient Measurements: I am concerned that lower detection limits, zero
drift in monitors, and precision of the reported CO data are not being treated sufficiently to understand
the uncertainty of the data and thus properly use and qualify the data in exposure estimates and models.
1. Lower detection limits (line 6-7 of page 3-11): The 1.0 ppm listed here as required is sufficient
for determining compliance with the current NAAQS, but is no longer sufficient for typical
urban concentrations, since concentrations have decreased significantly. Even the 2000 CO
AQCD acknowledged that "At many existing (urban) monitoring sites, the mixing ratio is
frequently below the lower detectable limit specified in Table 2-1." (page 2-2). The DL in Table
2-1 of the 2000 AQCD was the same as in the current ISA (1.0 ppm). Again, from the same
page: "A CO monitor with precision of 500 ppb would be adequate to prove compliance with the
CO standard, but would not provide adequate input data for CTMs." Many of the manufacturers
quote a lower detection limit of 0.04 or 0.05 ppm, which would be sufficient in most cases if the
monitors met that spec. However, in practice this can only be met with a frequent (every hour or
so) automatic zero drift correction, since the zero drift can be 0.1 ppm per day, but only the
newer models have an auto zero-drift option. And some agencies don't select the auto-zero
option, at least they didn't in the past. Also note that agencies have a wide range of monitors in
service, including many older models with worse DL and zero-drift specs, and without auto zero-
drift correction. Due to these points, I do not agree with the statement regarding zero drift on
lines 11-12 on page 3-11. If all US monitors had the zero-drift option, I agree, but this is not the
case; many states or agencies have only recently bought their first CO monitor with an auto zero
drift option (and Georga, for example, has only recently finished testing their first monitor with
the auto-zero option)..
2. I suggest that the ISA should provide the results of calculations on in-use detection limits and
precision to demonstrate that the monitors being used for the reported data are sufficient for use
48
-------
in exposure models. Otherwise, what does the data mean for exposure? In fact, the text at lines
32-36 on page 3-35, referring to Table 3-8, recognizes the DL issue and says that these results
should be used with some caution. An example of how to provide the precision information is
the following excerpt from the 2000 CO AQCD, page 2-8, although the reported statistics were
already old at that time. Similar statistics for detection limit and precision should be calculated
on the recent ambient data that is being used in the ISA, based on the information reported to
AQS for each reporting site, and such statistics should be reported in the revised ISA. "The error
analysis is a statistical evaluation of the accuracy and precision of air quality data. Guidelines
have been published by EPA (Smith and Nelson, 1973) for calculating an overall bias and
standard deviation of errors associated with data processing, measurement of control samples,
and water vapor interference, from which the accuracy and precision of CO measurements can
be determined. Since January 1, 1983, all state and local agencies submitting data to EPA must
provide estimates of accuracy and precision of the CO measurements based on primary and
secondary calibration records (Federal Register, 1978). The precision and accuracy audit results
through 1985 indicate that the 95% national probability limits for precision are ±9%, and the
95% national probability limits for accuracy are within ±1.5% for all audit levels up to 85 ppm.
The results (accuracy) for CO exceed comparable results for other criteria pollutants with
national ambient air quality standards (Rhodes and Evans, 1987)." If appropriate data is not
easily available in AQS for the sites being used in the ISA, then EPA staff should calculate
detection limit and precision statistics for at least a few example sites and report those results in
the revised ISA.
3. In addition, it is especially important that data below detection limit is reported properly, since 8-
hr CO concentration averages, for example, might include several hours of low CO
concentrations. If the measured value is at or below the detection limit, or the EPA specified
detection limit, that data value is often just reported as that value, say 0.5 ppm, for example, or
even as 0.0; but using this data in averages can lead to biased averages (see the discussion of this
issue in the EPA Toxics Workbook, McCarthy et al., 2008). Using a value of DL/2 or a
distribution of values below the DL may be more appropriate in this application.
Chapter 3.4.2.1 Monitor Siting Requirements (page 3-12): Discussion on lines 6-9 covers microscale
sites. It seems like knowing the number of these sites being used for all of the following tables and
graphs would be important, since concentrations at a microscale or near-roadway site could be 2-10
times higher than nearby concentrations (see, for example, later near-roadway discussions, page 3-39 to
3-42). I suggest the number of microscale, middle scale, and neighborhood scale sites being used in
these analyses be mentioned here. However, I think it is also important to mention especially the
number of microscale sites when discussing the population representativeness (Figures 3-7 to 3-10 and
figures in Annex A), distributions of CO data (Tables 3-3 to 3-6), the seasonal distribution plots (Figures
3-14 ff), the inter-site statistics (Tables 3-7 ff), and the diel plots where data from multiple sites are
averaged together (Figures 3-26 and 3-27). In each of these cases, the number and location of a
microscale site could significantly bias the data in the table or figure (and thus bias the result of using
the data in exposure analyses). In fact, I suggest that a second set of many of these table and figures be
added for just the 70 microscale sites (or for those sites among these 70 which are judged to best
represent microscale sites).
49
-------
Tables 3-3, to 3-6 distribution of CO data: As mentioned earlier, I think the data in these tables need
some qualifications, either in the table/footnotes to the table or in the associated text. What is the
reported detection limit for the reported data, since so much of the data distribution are low? How many
microscale sites have their data averaged in with data from middle or neighborhood scale sites? And
most importantly, how do these issues influence the interpretation of the results? Maybe separate lines
or table should be developed for the microscale sites?
Population coverage, page 3-13 and Figure 3-7: I don't agree that the current Phoenix CO monitors
properly cover the total population; there are areas of significant population and population density to
the southeast and the northeast of the central area that are not covered. I suggest changing the words to
better reflect the representativeness of these sites.
Location of monitors, relative to roadways, lines 23-27 page 3-13 and Figures 3-13 and 3-15 (plus
figures in Annex A): I see the usefulness of these figures in general (central city versus boundary), but
I can not determine from them how close sites are to major roadways (and I think this is their major
purpose). To be useful, it seems like the figures need something about traffic density and something
about how close (in meters) the sites actually are to a major road. In addition, the text needs a
conclusion about this (something like "many sites are very near major roads" or "only a few sites are
near major roads" and "thus the results shown in Table 3-3ff and Figures 3-16ff are biased (or not) for
Phoenix, but not for Pittsburgh, etc. due to x and y".
Lines 4-5 page 3-24 Highest CO in Ogden: Since this concentration is so much higher than others,
please explain what caused it. Was this an 'exceptional event'?
Lines 32-36 page 3-35, lines 1-4 page 3-36 and Table 3-8: As mentioned earlier, the caution due to a
high monitoring detection limit is a significant limitation of this and other data displays. The words here
are good and name some specific sites where these limitations may be problems. I suggested that
statistics be added on detection limits and precision of the actual monitors used where ever data is
reported.
Chapter 3.5.1.3 Near-roadway discussion: This discussion on page 3-39 and 3-40 is good and
mentions the importance of this issue for exposure. However, comments could be added on the potential
influence of lower wind speeds on concentrations (to the discussion on lines 5-18 page 3-40) and on the
influence of varying wind directions on the CO distributions within urban canyons (lines 19-36 page 3-
40). Both conditions will reduce the gradients discussed and thus distribute the roadway CO more
spatially within the urban canyon.
Pages 3-42 and 3-43, Figure 3-23 using monitor comparisons for understanding neighborhood
variability: The statement on line 1 of page 3-42 is very important and thus information on CO
concentrations on this scale (microscale) need to be used in the subsequent CO exposure modeling, in
order to properly represent this issue. In addition, the results of Figure 3-23 could be significantly
different if some of the sites were microscale sites, so please state if they are or not and discuss the
implications of the result.
50
-------
Figures 3-26 and 3-27 and associated text on pages 3-46 to 3-47 hourly variation in CO: I think that
averaging data from multiple sites together to get average diurnal profiles is fraught with problems, and
I think the results in these figures illustrate those problems.
1. For example, averaging data from different sites with potentially different (or even slightly
different) diurnal profiles will distribute that profile over multiple hours and make it look like a
flat diurnal profile. On the other hand, if the sites have the same diurnal profile, it will reinforce
the peaks and valleys. Or if there is only one site (Seattle, for example), then the profile will
keep its shape and not be diluted by data from another site. At a minimum, these qualifications
should be discussed in the text with the conclusions (and I think they are significant limitations);
however, I suggest that it is more appropriate to re-do the text and figures (or show only diurnal
curves for only one representative site for each area).
2. In addition, I do not understand how the number of monitor days (N) is correct. For example, for
Seattle (only one site), 3 years of weekday only data (Figure 3-26) would be about 780 monitor
days; how can it be 1577? See also Figure 3-27, where weekend only data at 1 Seattle site would
be about 312 monitor days, so what does the 639 mean? On the other side, for Phoenix there are
5 sites, thus 3 years of weekday only data would be about 3900 monitor days. The value of 1021
implies only about 25% data recovery; this does not make sense. The text tells us that
Anchorage is included, but does not operate year-round; this statement implies that all the other
sites do operate year round. In addition, the text talks about using only data from site with 75%
data completeness (page 3-12 line 17). Maybe I am doing something wrong here, but please
explain.
Lines 18-21 page 3-51 Comment on non-collocated monitors: I agree with comment regarding
influence if CO monitor is not collocated with monitors for other pollutants, but the text at line 5 page 3-
50 says only collocated data was used for Figure 3-28 and similar figures in Annex A. What is the
actual case here? If only a few pairs include a non-collocated CO monitor, then why not just drop those
few cases and then the results don't have to be qualified?
There should be a conclusion or "so what" statement added at the end of Chapter 3.6.3.2, Measurement
Error in Personal Exposure Modeling.
In the Summary and Conclusions, Chapter 3.7.3 Ambient CO Measurements: Please add
information here on detection limits and precision, plus a discussion of the implications of DL and
precision on the descriptions of CO concentrations and thus exposure estimates. In addition, note that
the sentence on microscale monitors (lines 11-13) is new information not mentioned previously in the
details of Chapter 3; as mentioned earlier, the number and influence of microscale monitors should be
discussed earlier and then summarized here.
In the Summary and Conclusions, Chapter 3.7.4 Environmental CO Concentrations: For the
sentence on lines 7-10 which discusses the diel profiles, please add significant qualifications for
averaging data from multiple sites together, etc., as discussed earlier. In addition, it may be necessary to
change these conclusions if the plots are redone, based on my earlier comments.
51
-------
ISA Charge Question 4. How well do the choice and emphasis of exposure topics presented in
Chapter 3 provide useful context for the evaluation of human health effects in the ISA? Is the
discussion and evaluation of evidence regarding human exposure to ambient CO and sources of
variability and error in CO exposure assessment presented clearly, succinctly, and accurately?
The ISA concludes in Chapter 3.7 that central-site monitor concentration is generally a good
indicator for the ambient component of personal CO exposure. What are the views of the Panel
on this conclusion and its supporting evidence?
The exposure topics presented in Chapter 3 are appropriate and useful for the evaluation of human
health effects in the ISA. However, there are no significant discussions of CO measurement errors in
either this section of Chapter 3.4. Chapter 3.6.3.1 line 1 page 3-58: The "associated monitoring
errors" are NOT discussed in Chapter 3.4; see my earlier comments that this should be added to Chapter
3.4.1. In the rest of the paragraph, I am most concerned about differences in errors in CO concentration,
since some sites may have very old monitors with larger precision and larger zero drift while other sites
may have new monitors with auto zero drift corrections, etc. Please discuss the influence of this type of
error on health outcomes. The last sentence of the paragraph kind of leaves the reader hanging; what is
the implication of this issue for exposure estimates? And again in the last sentence at line 23 of page 3-
59, what is the influence of the potentially large personal measurement error on exposure estimates and
how should this be treated?
Chapter 3.6.5.3 CO Exposure Assessment Variability and Error (page 3-65): How does the
statement on lines 7-10 follow from the previous text, given the specific text on lines 3-7 page 3-63 for
an example of high in-vehicle exposures (or Figure 3-34 or other text earlier) or lines 12-13 on page 3-
61 for an example of near-roadway exposures? It seems like the comments/qualifications on lines 17-
21 may apply to the first 8 locations in Figure 3-34, but do not apply to the in-vehicle location in Figure
3-34. Thus, not including in-vehicle and near-road CO exposures could lead to significant errors in
exposure estimation and thus in health outcomes.
Chapter 3.6.7: In light of the currently-planned method for preparing the CO concentration fields for
the exposure model (as discussed at May 13 panel meeting and in slide 13 of the presentation), the
discussion and references cited on pages 3-70 and 3-71 (in Chapter 3.6.7) are not sufficient to support
the methods plan and should be significantly expanded. There is only one reference cited for
concentration surfaces, which will be a major tool in the analysis; many more are needed. A few
additional references that I can easily find are listed at the bottom of my comments. Note that many of
these references are for pollutants other than CO, since few studies are currently being done on CO;
however, the methods can be reviewed and used as guidance for similar applications for CO. In
addition, I think that the exposure modeling Chapter ( 3.6.7) should include much more specifically
about the methods that will be used to address in-vehicle and near-road exposures. A recent HEI report
is now available on the web at: http://pubs.healtheffects.org/view.php?id=306; this report has an
excellent summary of the current literature and thinking on near-roadway exposures and a good
reference list.
Regarding the Chapter 3.7 conclusion that central-site monitor concentrations is generally
a good indicator for the ambient component of personal CO exposure: Total personal
exposure to CO is the time weighted sum of exposure to all microenvironments including
52
-------
multiple outdoor environments (not just multiple indoor environments). Therefore the central-
site monitor concentration is not viewed as 'a good general indicator for the ambient component
of personal CO exposure'. Equation 3.4 should be reformulated to include multiple outdoor
microenvironments, including at least near roadway exposures (ref section 3.5.1.3 and Figure 3-
34). Equation 3.4 should also distribute the concentration term to both outdoor and indoor
microenvironments as a concentration within both the sum of the indoor components and the
sum of the outdoor components (into a new summation term) specifically as the concentration in
each microenvironment, Ci for both indoor and outdoor. This will also require that the following
sections (and any others) be modified to reflect that more-complex exposure: Lines 30-31, page
3-57; lines 7-10, page 3-65 and page 3-74 lines 10-11.
In the Summary and Conclusions, Chapter 3.7.5 Exposure Assessment...: Same comment as above
for lines 10-11 on page 3-74 of the Summary, i.e not including in-vehicle and near-road CO exposures
could lead to significant errors in exposure estimation and thus in health outcomes.
In the Summary and Conclusions, Chapter 3.7.5 Exposure Assessment...: On page 3-74, lines 2-5
there is a conclusion regarding the importance of commute time on CO exposure -1 do think this is
important, but I did not see it discussed in the earlier part of the Chapter. Please include a discussion of
this topic in the main Chapter (put in Chapter 3.3.5?). In general, this section and the exposure
modeling information in general, should be re-evaluated in light of the OAQPS presentation on May 13
and the approach they now propose to use for the REA; there may be portions of Chapter 3 that need
strengthening besides just Chapter 3.6.7 as discussed above.
Comments on the COHb versus CO concentration space: Both the discussion of COHb and
its response to CO concentrations (Chapter 4.2.3) and the discussions on CO uptake and
elimination (e.g. Chapter 4.4.1) could include additional information and data from open-air
exposures at higher CO concentrations. For example, there is published data on COHb levels in
people exposed to high concentrations of CO (up to maximum 8-hour averages of 20-40 ppm) in
an open-air setting at Lake Havasu, AZ; see the CDC MMWR:
http://www.cdc.gov/mmwr/preview/mmwrhtml/mm5315a3 .htm and the Journal of the American
Medical Association: http://jama.ama-assn.org/cgi/reprint/291/22/2692.pdf This data could be
used to expand to higher CO concentrations the discussion on the relationship of COFtb levels to
CO concentrations.
Minor edits and typos in the ISA:
- US EPA 2000 references (2) in second paragraph of Chapter 1.2 should be US EPA 1991
The bullet on line 9-10 of page 1-6 does not really properly describe Annex A. Annex A only
contains maps, tables, and charts of CO data
The yellow colored areas, especially when small, are very hard to see on Figures 3-11, 3-12 and
similar figures here and in Annex A.
Shouldn't the word maximum be added to Iine3 page 3-44, so that it would read "...an outdoor
worker's maximum exposure over the course of the day..."?
Change more to the most on line 4 page 3-46 (or add what it is more than).
53
-------
I suggest the word "only" be added in front of the 12% on line 11 of page 3-61.
Shouldn't the date in line 32 of page 3-68 be 1997, and the following review start with what has
been published since 1997, since the 2000 AQCD did not have literature on exposure modeling
past 1997?
Add the complete reference for Flachsbart, line 35 of page 3-69.
- The sentence that starts "Given reductions...." on line 11 of page 3-70 does not make sense to
me.
I suggest the word compared instead of "judged" in line 1 of page 3-73.
The text on line 15 of page 3-73 should read "...Figures 3-14 and 3-16)".
- Add " ...Policy-relevant Background (PRB)" to line 25 of page 3-73.
Selected, easy for me to find, references for spatial mapping (see above discussion for Chapter 3.6.7):
Gauderman, Avol, Lurmann, Kuenzli, Filliland, Peters, and McConnell "Childhood Asthma and
Exposure to Traffic and Nitrogen Dioxide, Epidemiology 2005; 16, 737-743.
Ross, Jerrett, Ito, Tempalski, and Thurston "A land use Regression for predicting fine particulate matter
concentrations in the New York City region", Atmospheric Environment 41 (2007) 2255-2269.
Hoek, Beelen, Hoogh, Vienneau, Gulliver, Fischer, and Briggs "A review of land-use regression models
to assess spatial variation of outdoor air pollution" Atmospheric Environment 42 (2008) 7561-7578.
Henderson, Beckerman, Jerrett, and Brauer "Application of Land Use Regression to Estimate Long-
Term Concentrations of Traffic-Related Nitrogen Oxides and Fine Particulate Matter ES&T 2007, 41,
2422-2428.
Molitor, Jerrett, Chang, Molitor, Gauderman, Berhane, McConnel, Lurmann, Wu, Winer, and Thomas
"Assessing Uncertainty in Spatial Exposure Models for Air Pollution Health Effects Assessment EHP
vol 115,no 8, August 2007.
Popawski, Gould, Setton, Allen, Su, Larson, Henderson, Brauer, Hystad, LIghtowlers, Keller, Cohen,
Silva, and Buzzelli "Intercity transferability of land use regression models for estimating ambient
concentrations of nitrogen dioxide" J Exposure Science & Environmental Epidemiology (2008), 1-11.
54
-------
Dr. Beate Ritz
1. The framework for causal determination presented in Chapter 1 was developed and
refined in other ISAs (e.g., the PM ISA). During previous reviews, CASAC generally
endorsed this framework in judging the overall weight of the evidence for health effects.
Please comment on the extent to which Chapter 1 provides necessary and sufficient
background information for review of the subsequent chapters of the CO ISA.
The wording in this chapter could be improved and is not always consistent with the latest
definitions and uses of terminology in epidemiology; for example instead of 'effect modification'
it should read 'effect measure modification', also instead of 'health effects' one might consider
using 'adverse health outcomes' or 'changes in (lung) function' etc.; 'effect' seems to imply an
etiologic factor that is not mentioned but has an effect on health. Also, the authors of this chapter
move back and forth between the concepts of confounding and effect measure modification as if
both are of concern for study validity. Yet effect measure modification is not a concern when
assessing bias and these concepts should not be mixed. The way these concepts are referred to
now in the text suggests a lack of appreciation for the differences in these two concepts; this
might be due to the fact that similar statistical methods (stratification) are used to assess these
two different concepts in data. In short, effect measure modification should not be subsumed
under or confused with bias assessment in observational studies.
55
-------
The criteria for causal determination detailed in table 1-2 are very similar to those used by
the IOM and the International Agencies for Research on Cancer, however one important
difference is that these agencies convene expert committees to review the literature in depth and
to apply these criteria in order to arrive at conclusions about causality; they do NOT ask staff to
perform this task for the agencies with external reviewers simply commenting. Thus, these
qualitative criteria are applied to the scientific literature in face-to face meetings that include
different groups of experts, all of whom have reviewed the literature in their fields in great detail
and thus are fully aware of the strengths and weaknesses of the studies included and weighted in
the qualitative review. Under these circumstances, these qualitative criteria suffice to guide an
expert based judgment including lengthy discussions of the strengths and weaknesses of the
evidence at hand. But without a standardized or quantitative review of the literature at hand,
these criteria are ambiguous if not outright subjective. When applied in a qualitative literature
review all judgment concerning the strengths and weaknesses of studies is left to the author and
thus subjective unless quantified or made very explicitly. The overall judgment whether an
observational study suffers from any substantial bias or to what degree they suffer from bias
remains qualitative and subject to the author's judgment and should be made open for challenge
by other experts who reviewed the literature according to the same criteria. While qualitative
reviews have been widely used in the past and may be appropriate when there are less than 5
studies published in a subject area they leave much room for a subjective and biased reading,
reporting, and interpretation of the literature. Since the epidemiologic literature on criteria air
pollution health effects has multiplied greatly in the past decade -as can be seen in Chapter 5 -
and in many areas there are now more than 5 studies available, it would be much more
appropriate to apply standardized and transparent rules for data abstraction and to derive
quantitative effect estimates based on meta-analytic procedures before drawing inferences about
the scientific literature. More important than even deriving a singular effect estimate is that a
systematic and quantitative procedure requires making the authors' assumptions explicit rather
than allowing authors to emphasize studies they agree or disagree with.
56
-------
Page 1-8 " The most compelling evidence of a causal relationship between pollutant and
exposure and human health effect comes from human clinical studies" - meaning experimental
chambers studies - this statement needs to be qualified since chamber or other experimental
studies in humans are impossible to conduct for the assessment of long-term exposures and
chronic health outcomes of interest since these types of experiments per se can only be applied in
a context of short term changes in air pollution and physiologic biomarkers that do not results in
continued harm to a subject, i.e. such experiments can only be set up for certain outcomes or
exposures. Hence, observational studies are imperative and likely present the only data available
for a number of health outcomes and exposure scenarios. This should be acknowledged and
seems to be neglected in this description. These clinical and chamber studies do not provide the
type of evidence that is 'most important' for human health risk assessment but rather the type of
evidence that can be obtain within the ethical constraints of human experimentation. Also on
page 1-10, not only do epidemiologic studies provide exposures in 'natural settings' but rather
they are often the only form of data available for certain outcomes and exposures, i.e. in
instances for which chambers studies are impossible to conduct (such as predicting mortality and
adverse birth outcomes). This general attitude of overvaluing short-term experimental human
studies seems to be carried through in this report and for the health risk assessment proposal that
proposes to only consider modeling based on short term changes in cardiac outcomes from
chamber studies extrapolated to cardiovascular morbidity (on page 11 of the Scope and Method
Plan for Health Risk and Exposure Assessment "Potential health benchmark values to be used in
the planned risk characterization linked to the exposure/dose analyses will be derived solely
based on the controlled human exposures literature"). At this point I am wondering why the
epidemiologic literature is reviewed at all if it has no bearing on these estimates.
2. Chapter 2 presents the integrative summary and conclusions from the health effects
evidence, with the evidence characterized in detail in subsequent chapters. What are the
views of the Panel on the effectiveness of the integration of atmospheric science,
exposure assessment, dosimetry, pharmacokinetics, and health effects evidence in the CO
ISA?
The same critique mentioned above applies to these summaries that ignore the epidemiologic
evidence in favor of human controlled exposure studies for cardiovascular morbidity. The
summaries by outcome category should be more explicit in stating what type of data the
causality determinations are based on, such as 'one chamber study plus x number of
epidemiologic study in which the following biases were or were not present etc etc..."
3. To what extent are the atmospheric science and air quality analyses presented in Chapter
3 clearly conveyed and appropriately characterized? Is the information provided
regarding CO source characteristics, CO chemistry, policy-relevant background CO, and
spatial and temporal patterns of CO concentrations accurate and relevant to the review of
the CO NAAQS?
57
-------
4. How well do the choice and emphasis of exposure topics presented in Chapter 3 provide
useful context for the evaluation of human health effects in the ISA? Is the discussion
and evaluation of evidence regarding human exposure to ambient CO and sources of
variability and error in CO exposure assessment presented clearly, succinctly, and
accurately? The ISA concludes in section 3.7 that central-site monitor concentration is
generally a good indicator for the ambient component of personal CO exposure. What
are the views of the Panel on this conclusion and its supporting evidence?
5. The dosimetry and pharmacokinetics of CO are discussed in Chapter 4. Please comment
on the presentation in the ISA of the current state of knowledge on the Coburn-Foster-
Kane (CFK) model and model enhancements. Has the expected contribution of different
exposure durations (1-24 h) to COHb levels been clearly and accurately conveyed?
6. The mode of action section in Chapter 5 presents information on both hypoxic and non-
hypoxic mechanisms for CO health effects, with particular emphasis on recent studies
evaluating the non-hypoxic effects at low to moderate CO levels. Please comment on the
appropriateness of the focus, structure and level of detail in this discussion. For example,
is the evidence relating to the interaction between inhaled CO and endogenous CO
properly characterized?
While this is an important discussion it seems irrelevant as long as the health risk assessment
does not take any of the non-hypoxic mechanisms for CO health outcomes into
consideration.
7. Chapter 5 presents information on cardiovascular, central nervous system,
developmental, respiratory, and mortality outcomes following exposure to CO. To what
extent are the discussion and integration of lexicological, clinical, and epidemiologic
evidence for these health effects scientifically sound, appropriately balanced, and clearly
communicated? Are the tables and figures presented in Chapter 5 appropriate, adequate,
and effective in advancing the interpretation of these health studies?
58
-------
Throughout Chapter 5, epidemiologic studies receive very different levels of attention and
review; the level of detail in the text seems to be depending on how many studies were published
for each outcome category, e.g. if there were 20 studies addressing a health outcome each studies
is described in a cursory manner with a sentence or two, while for a health outcome for which
only 2 studies have been published, these few studies are described and evaluated in much more
detail. The brief mention of studies leaves a lot of questions open concerning the validity and
methods used in the 20 studies i.e. for the reader it is impossible to assess from the qualitative
review text presented whether or not or to what degree these studies may be biased or the study
design may have been adequate in addressing the question at hand; i.e. the brief and almost
cursory mention of each study in the text does not allow the reader to inspect the actual data and
evaluate the results in the same manner as possible for the much better described fewer studies.
Also, since it is much more likely that 20 studies are heterogeneous with respect to their results
as well as method than the two studies, having more data available may end up being worse then
having less since there in this report there is also an emphasis on mentioning inconsistencies
such that data richer areas are receiving more scrutiny that data poorer areas when in fact the
opposite would make more sense, i.e. homogeneity of results for only 2 studies might be much
less meaningful and informative than heterogeneity across 20 reports. While I find the tables and
figures helpful and they should provide the necessary detail on all studies reviewed, they lack
some key information in each chapter, e.g. there is no mention of the type of study design
employed for studies of heart rate variability and study results are neither presented in tabulated
format or in a figure (why the only figure presented is for IHD hospitalizations is not clear). Also
it seems strange that a study with a total subject N of 6 in table 5.4 is given as much attention as
one with an N of 6784 without further qualifications, e.g. in table 5.4 studies that employed
ambient exposure assessment and those using personal exposure assessment could have been
grouped together to emphasize these important differences in exposure assessment. Furthermore,
many of the tables report mean CO levels and mention 24 hrs or 8 hrs in brackets, however this
misleading at least in those studies I know well i.e. pregnancy outcome studies in which the
averages are trimester, weekly, or monthly averages of 24 hour measurement rather than 24 hour
averages in lagged time series models (the Ritz et al. (2000) study of PTB is listed in table 5-12
as having a Mean CO of 2.7 ppm for the 6-9 am period - however this mean represents a mean
over the whole first month of pregnancy and the Wilhelm and Ritz (2005) study mentions a 1.4
ppm mean for 24 hrs but this is in fact a, first trimester mean of 24 daily measurements; the way
this data is shown now the bracketed 24 hour mention seems to imply similar averaging period
and comparability in effect estimates. Also it is surprising to see the Ritz et al 2007 study listed
in table 5-12 but no results for this study presented in figure 5-6 - possibly because this paper
only presented estimates per quartile of CO increase rather than per Ippm increases in CO;
however, reseating quartiles to a continuous estimates is a possibility that should be considered
rather than leaving results from important papers out of a figure that gives an overview over all
study results). According to the text, the estimated increase in CO presented in the figures have
been 'standardized', however, how this might have been done across so many different study
types and averages for differing exposure periods (rather than 24 hour averages as the authors of
these chapters seem to imply) has not been explained. Also, in figure 5.1 the title says that the
effect estimates have been standardized to a Ippm increase in ambient CO for 1-hr max CO
concentrations, 0.75 ppm for 8-h max CO concentrations and 0.5 ppm for 24 hrs avg CO
concentrations, but the figure does not tell us which scale has originally been used in which
59
-------
study and it might be questionable whether effect estimate sizes based on these different scales
and based on different length lag periods are comparable to each other, thus at least indicating
which study used which scale might be informative. Also, since many of the cardiovascular
studies investigated more than one outcome, it seems like the studies themselves could be
tabulated first in much more detail that includes information about exposure assessment and
biases; then in outcome specific sections it would suffice to only mention the specific results; as
done now the studies are being mentioned in each subchapter by outcome as if these were stand
alone documents and nowhere is this kind of information presented.
a. For cardiovascular outcomes, controlled human exposure studies discussed in
Chapter 5 and in previous assessments have identified cardiovascular effects in
diseased individuals following exposures near the level of the current standards,
while new epidemiologic studies provide evidence of cardiovascular effects at
ambient concentrations. What are the opinions of the Panel on the treatment of
factors influencing the interpretation of this evidence, such as the plausibility of
cardiovascular effects occurring at ambient levels, the additive effect of ambient
CO to baseline COHb resulting from endogenous and non-ambient CO, and the
challenge of distinguishing effects of CO within a multipollutant mixture (e.g.,
motor vehicle emissions) in interpreting epidemiologic study results?
All of these issues could be nicely addressed in a quantitative framework of a meta-analysis that
follows a standardized protocol, why this has not been done is unclear. Also, the plausibility of
cardiovascular effects occurring at ambient levels cannot be assessed without doing an in-depth
review and assessment of all epidemiologic studies based on a thorough reading of this literature
by experts in the field in lieu of a formal meta-analysis. Again from the present text, assessing
and judging this is not possible since information on study design, exposure assessment and
possible biases is not always presented in a enough detail and a standardized manner to allow a
reader of these summaries alone to come to any conclusion, I and others on this panel would
need to go back to all of the original literature to form an informed opinion.
b. Please comment on the implementation, in Chapter 5, of the causal framework
presented in Chapter 1. Does the integration of health evidence focus on the most
policy-relevant studies and health findings?
See my comments above
8. What are the views of the Panel on the discussion of factors affecting susceptibility and
vulnerability in Section 5.7?
The factors mentioned are adequately and discussed well; however, it is unclear how they will be
playing any role in the health risk assessment since epidemiologic results overall do not seem to
be informing much if any of the planned calculations.
60
-------
Dr. Arthur Penn
General Comments
Chapter 1 of the ISA provides a worthwhile Introduction, especially regarding the distinctions
between causation and association. The conclusions summarized in Chapter 2 ("Sufficient to
conclude...Suggestive... Inadequate...") based on studies especially those related to
cardiovascular diseases (CVD) described in Chapter 5, may need to be re-evaluated. Chapters 3
& 4 appear to be the strongest chapterseven though they each raise some questions.
Surprisingly, the data summarized in Chapter 5 of the ISA '09 CO draft do not provide strong
support for the contention that spikes in levels of ambient COresult in exacerbation of a variety
of health outcomes. This is true forcardiovascular diseases (CVD) despite the "sufficient to
conclude" label in Chapter 2, and for respiratory diseases and pre- &peri-natal outcomes, despite
the "suggestive of a causal relationship" labels in Chapter 2. Issues include statistical
significance vs. "real-life" health concerns (alluded to in Chapter 1); very limited changes in
outcome for large population groups in response to spikes in ambient CO levels; no apparent
correlation between responses to very high levels of CO in controlled studies with volunteers vs.
responses to transient changes in ambient CO levels (i.e., the assumption that we can extrapolate
from responses to very high levels of CO back to responses at ambient levels needs to be
supported);difficulties in distinguishing between CO and co-pollutant effects; insufficient
justification for proposed studies on risk characterization and population exposure/dose analysis;
and finally—an issue barely noted in the ISA—the growing evidence that CO at levels that are
orders of magnitude higherthan ambient levels may have important therapeutic value for certain
serious medical conditions.
Specific Comments (regarding Chapters 2 & 5)
1) There is an unstated (and unsupported)assumption in the ISA that every reported statistically
significant change represents a major change in (clinical, health-related) outcome. Summary data
are often presented in the ISA as percentage change or as increases in relative risk (RR) or in
odds ratio (OR), without any consideration of the actual magnitude of change in the units being
measured. When the actual numbers are calculated from the original sources, the results are often
underwhelming; e.g., is there any clinical relevanceto a (statistically significant) increase of
Iheart beat/min in response to an increase in ambient [CO]??
NB: see additional comments on PTB, LEW &IUGR below.
2)For CVD, the largest data sets available for analysis are from studies of outcomes (e.g., CHD,
MI, angina, CHF) "associated" with ambient CO levels that exceed the 1-hr or 8-hr limits by 0.5-
l.Oppm. In most cases the relationship between spikes in ambient CO and CVD outcomes can be
generously described as very weak associations. It is insufficient to conclude that a "relationship
is likely to exist". In the section on increased admissions for IHD (pp. 5-24 &-25), data from ~
55,000 patients collected over 7 years from multiple hospitals in So. Calif, reveal that for a 0.75
ppm increase in 8-hr max CO levels, there are a total of 4 extra admissions/wk (!) across the
entire So. Calif, region for people with IHD,but only if they also had a diagnosis of CHF.For IHD
61
-------
patients without CHF, there were only 2 extra admissions/wk. In Montreal, a 14% increase in
daily ED visits for IHD works out to only 3 extra visits/wk. In the Atlanta study of >4.4 million
people over 7 years, the effect of a Ippm increase in 1-hr max [CO] was a 1.6% increase in RR
over baseline #of CVD-related visits/day. This works out to 4 extra CVD-related visits/wk in the
greater Atlanta area, above the baseline of 260 CVD visits/wk. The ISA reports that this is of
"borderline significance" (statistical). It's likely to be of even less clinical significance. (If the
data on CVD-related visits/wk were reassessed, would those weeks corresponding to spikes in
ambient CO always have higher #s of visits than weeks where there were no spikes in ambient
CO?)
For Mis, the effect of elevated ambient CO was minimal or non-existent in the 3 studies
summarized.
3 & 4) Thefocus of the Health Assessment Plan on investigating decreased time to onset of
angina is not justified clearly. The only large population study reported to date, from Tehran for
a 0.5 ppm [CO] increase over 24 hr., resulted in an increased OR for admission of 1.005 (i.e., /^
of 1% increased OR).
On the other hand, controlled studies on human volunteers reveal clear effects on specific health
outcomes; however, these require volunteers to be exposed to CO levels orders of magnitude
higher than ambient CO levels.The results of Allred et al, (NEJM, 321: 1426-32,1989) on the
effects of CO exposure on men with angina who are exercising are instructive. Allred et al
demonstrated a dose-response for increasing doses of CO and a) time to onset of angina and b)
ST wave depression, The time to angina onset dropped 19 sec. from 8 min. 21 sec. (room air,
0.6% COHb) to 8 min. 2 sec. (117 ppm CO for 1 hr, 2% COHb),and then another 17 sec. to 7
min. 45 sec.,as the [CO] doubled (253 ppm CO for 1 hr, 4% COHb). The results from these
exposures to high levels of CO, relative to ambient CO levels are clear. No reasonable prediction
can be made regarding how male angina sufferers who are exercising would respond to spikes in
ambient CO levels.
Further, the ISA (p. 2-22) notes thatnationwide between 2005-07 there were <10 days on which
the max. 8-hr CO level was exceeded and only one day when the 1-hr max level was
exceeded..Even at these rare high CO levels, COHb levels will likely be « 1%. Q. What vital
new information can we expect to gain by repeating this study at 2% COHb and then adding tests
at 2.5% and 3% COHb?It is not clear how either the Allred study or the proposed study relates to
expected responses arising from spikes in ambient CO.
The Kiazevich results (2000) summarized on p. 5-47, yielded CO-related results in healthy
exercising adults, but to get these results, volunteers were exposed to 1000 or 3000 ppm (! !)CO
for 4-6 min. and then to maintenance levels of 27-100 ppm CO-^COHb levels of 5-20%. Again,
it is not apparent that there is any predictive value that these results might have for exposures to
spikes of CO above ambient levels.
The summary of CO effects on all CVD outcomes (pp. 5-37 thru 5-44) is not compelling. The
most pronounced effects in the graph on p. 43 are all from 1 study in Seoul, Korea. All the other
62
-------
studies report < 10% effect, regardless of outcome. When combined with co-pollutants, CO
effects often disappear.
NB: The correlations between elevated ambient CO levels and hospital admissions for stroke
(pp. 5-30 to 5-32) are stronger than for any CVD outcome group. Controlled elevated CO studies
of animal models for stroke or TIAsmight be more informative regarding a possible outcome
than the proposed human volunteer angina studies.
The data on PTB, LEW and IUGR (pp. 5-57 to 5-70) also emphasize statistical significance
rather than actual magnitude of change. The Australia data (p. 5-65) report a 21.7 gm drop in
body wt. for a 0.75 ppm increase in ambient CO levels for 8-hr exposures. This drop =3/4 oz. of
total birth wt. vs. that for control neonates. In Fig. 5-7, 16/19 studies showed a neonatal wt.
changeof <10 gm (up or down) for increases of 0.5 ppmin ambient CO. This is
-------
Dr. Armistead (Ted) Russell
In general, this first draft of the ISA suggests that the final ISA will provide the scientific
foundation for EPA staff and CASAC to make recommendations on possible changes to the CO
NAAQS. However, there are areas that need to be strengthened. In particular, I do not believe
that the issue of confounding has been adequately dealt with in regards to interpreting the
epidemiologic results, and I am not sure it can be at this time. Given the source of CO, it will be
found concurrently with other automotive pollutants, and the ISA needs to spend much more
effort identifying what species are in the mix of automotive pollutants, the suspected health
effects of these other compounds, and what that means in terms of identifying the impact of CO
on health. There really is little way around the presence of all of these other compounds (both
measured and unmeasured), and the typical epi study has not controlled for the mix of other
automobile-generated pollutants. Thus, strong clinical results are needed, and as pointed out in
the ISA, such studies have been lacking in recent years.
Also, I trust that a summary chapter is coming, and that each chapter will have a brief section
highlighting the most important conclusions that are relevant to assessing whether we need to
change the NAAQS, and if so, what the level, form, etc. of the new standard should be.
Chapter 2:
As noted above, Chapter 2 should deal more directly with confounding by other automobile-
generated pollutants and how that impacts the identification of CO-specific health effects at
atmospherically-relevant (i.e., US current) concentrations. Also, should smokers be identified as
a potentially susceptible/vulnerable population?
Chapter 3: Source-to-Exposure
Again, I like this framework for such a chapter. It tends to reduce the amount of unneeded
information (though I am not sure why we need to know that the C atom is covalently bonded to
the O atom and that it has a mass of 28.0101: remember the intended use of this document).
I like the CO emissions section showing the current emission sources and trends. I think that it
would also be useful to include a forecast of 2020 emissions given the current regulations. The
section on physics and chemistry is reasonable, though a bit bleak. By that I mean that we may
not know the detailed gas phase kinetics of many compounds, but we also know a good deal
about most of the more important species, and even without the details, we have a reasonable
understanding vis-a-vis how much CO is produced. I would provide increased focus on CO
production from biogenics and compare that to anthropogenic emissions. At the bottom of page
3-9, the ISA correctly identifies CO as a compound that reacts with OH. However, the reaction
produces HO2, so it is not a loss of odd hydrogen/odd oxygen/radicals, so the role in this case is
mixed. It can add to ozone formation, and thus increase OH.
The section on instrumentation should provide a better idea of what instrument capabilities are
out there, not just what is required. Is the typical network monitor really only good to 1.0 ppm?
64
-------
The section on associations with co-pollutants really, really needs to deal with associations with
other automobile-derived pollutants, including EC, OC, benzene, 1,3 butadiene, formaldehyde,
Cu, and other both exhaust and non-exhaust emissions. It is true such information is not as
abundant as for the traditional pollutants, but it is of much more relevance. You have results for
Atlanta, but you should search for more.
The section on PRB is in need of some rethinking. The statement is made that "PRB
concentrations can best be determined from the extensive and long-running network of ..." This
statement needs to be supported by some stronger reasoning. In particular, why would one use
monitored values to find the PRB for CO, but use modeling for find the PRB for ozone? I could
readily see using CMAQ to find the PRB for CO, and this would capture the CO formation from
biogenic emissions. Given the other ways one can calculate PRB concentrations, this section
needs to be very careful about what is said and to support the statements made.
Something missing from this chapter is a thorough description of APEX and results from prior
applications, particularly to CO. While there is a brief section on exposure modeling, it is not up
to fully supporting the future use of APEX in the REA. The consistent reliance on APEX for
conducting NAAQS-related exposure analysis should lead EPA to doing a more thorough
assessment of APEX across pollutants.
Minor:
3-13 1 14: Sentence beginning "As concerns..." is awkward.
Figs. 3-7,8,9: The way monitors are shown is sub-optimal.
Tables 3-3-6: Can you add Ogden?
Page 3-24: Explain why Ogden had such an incredibly high 1-hr CO.
Table 3-7,8: Please explain further.
Figure 3-16: Why is the third quarter of monitor A so high, first quarter so low? Mointor A
appears to behave very differently than the others.
Figure 3-24: "... highest DAILY 8-hour.
Page 3-52: "1 part per billion" (no s)
Page 3-63. To me, Figure 3-32 does not look logarithmic, and physically, that is not the
functional form expected.
Page 3-64, line 18... Isn't this getting in to dose?
3-64: Last paragraph. This paragraph is unclear.
At the end of the day, you are going to need to re-assure the various parties that there is a
reasonable chance that the health effects being seen are due to CO, not the other associated
pollutants. This requires a good deal more attention being paid to how well you deal with the co-
pollutant issues, including source characterization, atmospheric dynamics and concentrations
(particularly spatial and temporal associations), and epidemiologic study results where they have
65
-------
adequately considered automobile-derived pollutants. Controlling for PM2.5/10, and SO2 is
almost meaningless in this context. My read of the health chapters suggests that when
considering automobile derived pollutants (e.g., NO2 and BS/EC), the effects typically were
significantly reduced and became insignificantly different from zero in many cases. These
studies did not control for other automobile-derived pollutants that are of increasing concern
(e.g., metals, resuspended road dust). I think it would be good to have a very extensive
assessment of the issues associated with the concurrent exposure to the variety of automobile-
derived pollutants, and how such should be considered in the context of interpreting the
epidemiologic analyses.
Responses to Charge Questions:
1. The framework for causal determination presented in Chapter 1 was developed and
refined in other ISAs (e.g., the PM ISA). During previous reviews, CASAC generally
endorsed this framework in judging the overall weight of the evidence for health effects.
Please comment on the extent to which Chapter 1 provides necessary and sufficient
background information for review of the subsequent chapters of the CO ISA.
2. Chapter 2 presents the integrative summary and conclusions from the health effects
evidence, with the evidence characterized in detail in subsequent chapters. What are the
views of the Panel on the effectiveness of the integration of atmospheric science,
exposure assessment, dosimetry, pharmacokinetics, and health effects evidence in the CO
ISA?
As noted above, I do not believe that Chapter 2 (or any of the chapters) delves as deeply in to the
issue of co-pollutants as is necessary for the issue at hand. This issue needs its own section in
Chapter 2 with the take-home message very clearly spelled out and supported.
3. To what extent are the atmospheric science and air quality analyses presented in Chapter
3 clearly conveyed and appropriately characterized? Is the information provided
regarding CO source characteristics, CO chemistry, policy-relevant background CO, and
spatial and temporal patterns of CO concentrations accurate and relevant to the review of
the CO NAAQS?
As discussed above, Chapter 3 does a reasonable job, with a few shortcomings. CO formation
from isoprene could be brought out a bit more, and the PRB discussion needs to be better
supported, particularly since other ISA's come to an opposite conclusion regarding the use of
models versus observations.
66
-------
4. How well do the choice and emphasis of exposure topics presented in Chapter 3 provide
useful context for the evaluation of human health effects in the ISA? Is the discussion
and evaluation of evidence regarding human exposure to ambient CO and sources of
variability and error in CO exposure assessment presented clearly, succinctly, and
accurately? The ISA concludes in section 3.7 that central-site monitor concentration is
generally a good indicator for the ambient component of personal CO exposure. What
are the views of the Panel on this conclusion and its supporting evidence?
A shortcoming here is the rather short discussion about exposure modeling. Exposure modeling
will be a main focus of the REA, and as such, this section needs to be made to fully support that
future effort, with particular emphasis on model evaluation. Also, as mentioned above, the
discussion of co-exposure to other automobile-derived pollutants, including non-exhaust
components, needs to be fortified.
5. The dosimetry and pharmacokinetics of CO are discussed in Chapter 4. Please comment
on the presentation in the ISA of the current state of knowledge on the Coburn-Foster-
Kane (CFK) model and model enhancements. Has the expected contribution of different
exposure durations (1-24 h) to COHb levels been clearly and accurately conveyed?
6. The mode of action section in Chapter 5 presents information on both hypoxic and non-
hypoxic mechanisms for CO health effects, with particular emphasis on recent studies
evaluating the non-hypoxic effects at low to moderate CO levels. Please comment on the
appropriateness of the focus, structure and level of detail in this discussion. For example,
is the evidence relating to the interaction between inhaled CO and endogenous CO
properly characterized?
7. Chapter 5 presents information on cardiovascular, central nervous system,
developmental, respiratory, and mortality outcomes following exposure to CO. To what
extent are the discussion and integration of lexicological, clinical, and epidemiologic
evidence for these health effects scientifically sound, appropriately balanced, and clearly
communicated? Are the tables and figures presented in Chapter 5 appropriate, adequate,
and effective in advancing the interpretation of these health studies?
a. For cardiovascular outcomes, controlled human exposure studies discussed in
Chapter 5 and in previous assessments have identified cardiovascular effects in
diseased individuals following exposures near the level of the current standards,
while new epidemiologic studies provide evidence of cardiovascular effects at
ambient concentrations. What are the opinions of the Panel on the treatment of
factors influencing the interpretation of this evidence, such as the plausibility of
cardiovascular effects occurring at ambient levels, the additive effect of ambient
CO to baseline COHb resulting from endogenous and non-ambient CO, and the
challenge of distinguishing effects of CO within a multipollutant mixture (e.g.,
motor vehicle emissions) in interpreting epidemiologic study results?
67
-------
b. Please comment on the implementation, in Chapter 5, of the causal framework
presented in Chapter 1. Does the integration of health evidence focus on the most
policy-relevant studies and health findings?
What are the views of the Panel on the discussion of factors affecting susceptibility and
vulnerability in Section 5.7?
68
-------
Dr. Laurence Fechter
Charge question 7: Is the discussion in chapter 5 scientifically sound?
I have focused on the issues of CO's effects on the CNS and on the developing subject as these
are my primary areas of expertise.
Section 5.3 CNS effects
A general comment on this section is that the use of topic sentences to provide some orientation
to the reader would be welcome. Many subsections consist of descriptions of multiple studies of
CO exposures at various levels and various durations. Having a topic sentence suggesting a
range of values that yield consistent outcomes would call attention to the most relevant studies.
For example, section 5.4.2.1 would benefit from a topic sentence indicating a range of CO values
associated with decreased birth weight.
The epidemiological study results present data on relative risk and confidence intervals. Many of
the relative risk values, are quite modest. How much faith can we put in a RR of 1.02? Some
guidance is important for interpreting the data. Moreover, there seems to be some inconsistency
between the size of the OR for CO exposure's effects on birth weight vs. congenital anomalies
and the interpretation (i.e. larger OR for congenital anomalies yet a statement that there is little
evidence for increased risk).
Section 5.4.1.2. Birth weight, etc.
The data presented in this section are not especially consistent. It might be appropriate to add a
sentence or 2 identifying the far clearer effects of maternal tobacco smoking on birth weight
(even though MTS is a very complex exposure) as a relatively clear outcome and a possible
rationale for looking for a relationship between CO exposure per se and reduced birth weight,
prematurity etc.
Section 5.4.2.1
See comment above about use of topic sentence
Line 22 Fechter and Annau found a 5 % decrease in birth weight in rats. As written it suggests
that either CO levels or HbCO levels were 5%.
P 5-76 line 16-17 mistakenly states that Fechter and Annau (1977) did NOT find a significant
birth weight effect after prenatal CO. This statement is also inconsistent with statement on
previous page.
p 5-77 line 26 correct to read " given various protein diets...."
P 5-78 spelling of toxicity line 15
69
-------
5-80 Placenta section... .define high altitude; should this not state "chronic potential hypoxia
exposure"?
P 16 that same section. How relevant is the dose used to inhalation exposure studies?
Section 5.4.2.2 it appears that 75 ppm is commonly a NOAEL whereas ISOppm is a LOAEL.
Could this be stated directly or else suggested in an effort to facilitate the reader's task of
assessing this section?
page 5.860- line 14 guinea pigs are suggested as a good model for human CNS development.
This may require some added qualification as the newborn guinea pig is in many respects far
more mature than the human at birth.
p. 5-87 I'm not certain that the term "demasculination" is the most useful in understanding a shift
in DA release after amphetamine.(Is demasculination a word?)
P 5-88 a comment on the permanence v. transient effects observed under neonatal hyperthermia
effects on neurotransmitters would be helpful
p 5-88-5-91
An important sub-section entitled "The Developing Auditory System" delineates the results of a
series of reports published by researchers at UCLA in which the effects of postnatal CO
exposure are assessed in rats maintained in an artificial rearing system. These studies are
important to describe accurately because the CO levels selected for use include the lowest levels
employed in studies designed to evaluate nervous system development (12, 25, 50 and 100 ppm).
Moreover, the exposure levels selected do have some relevance to ambient CO concentrations.
Also presented in this sub-section is the result of a human study in which auditory function was
assessed in neonates who were offspring of non-smokers, and heavy, medium, and light
smokers. The conclusion of this section, in my judgment over-interprets the data as supportive of
an adverse effect of CO exposure at very low exposure levels on the developing auditory system.
Moreover, one must be somewhat circumspect about the laboratory animal data presented
because the nature of the artificial rearing system a rather invasive procedure. It is possible that
the developing auditory system is especially vulnerable to CO exposure. However, there is clear
evidence from adult rats showing that CO by itself can produce transient functional impairment
of the peripheral auditory system only when near life-threatening CO concentrations are
employed.
The laboratory studies described (Webber et al., 2003 ... .date missing from ref on p 5-88 line
26 Webber et al., 2005....date missing from ref on p. 5-90 line 26....Lopez et al., 2003, and
Stockard-Sullivan et al., 2003) consist of a set of studies in which neonatal rats are exposed to
low levels of CO while maintained in an artificial rearing system in which rat pups are fed
through a gastronomy tube and maintained effectively in floating cups placed in a water bath.
Notably, brain weight is reduced for both the artificially reared and artificially-reared carbon
monoxide exposed neonates compared to the maternally reared non-CO exposed control subject
70
-------
(Stockard-Sullivan et al., 2003). Most of the studies describe immunohistological changes or
qualitative histological observations in either the cochlea or the inferior colliculus of artificially
reared CO exposed rats. Whether the changes noted have functional consequences is uncertain.
In only one manuscript by this group (Stockard-Sullivan et al., 2003) were functional measures
taken from the auditory system and these studies are quite limited (e.g. measurement of DPOAE
generation as a measure of cochlear function performed over a very narrow range of frequencies
that is rather low compared to the normal rat audiogram).
The second full paragraph on p 5-89 describes the outcome of the Korres et al (2007) paper and
briefly describes two non-invasive measures of auditory function. Notably, the otoacoustic
emission (OAE) is described as an "echo" recorded by a microphone placed in the external ear
canal. Actually, what is measured is an active tone that is produced by the cochlea and not a
passive echo. Indeed, the distortion product otoacoustic emission is remarkable because it occurs
with totally different frequency characteristics than do the two primary tones that are delivered to
the ear. The description of the Korres paper, however, requires a bit more explanation.
Specifically, it needs to be pointed out that neonates were grouped by mother's smoking history
(none, low level, moderate, and high level). The transient ototacoustic emission was indeed
significantly lower among the offspring of smokers than non-smoking mothers only at the
highest test frequency used (4 kHz). This might be meaningful because high frequency hearing
might be predicted to be more vulnerable to hypoxia. However, there was no evidence of a
relationship between level of maternal smoking and the reduction in the otoacoustic emission
recorded. Thus, this study cannot be considered to be definitive for a link between tobacco
smoke exposure and impaired auditory system development.
Charge question 8. factors affecting susceptibility and vulnerability in section 5.7
While relevant potential susceptible populations are identified, clear conclusions (even those
stating that the current literature does not fully inform on the question of susceptibility) are not
always present. While susceptibility factors are identified, there is no overt attempt made to
address the likelihood that these factors would most likely be of differing seriousness. For
example, males and females may well differ in terms of intrinsic production of COHb, but there
might be other factors such as occupation that predispose one sex to higher CO exposure.
Moreover, there may well be other factors that predispose males to cardiovascular disease
rendering them more sensitive to CO. My point is that the gender issue is multifactorial and the
discussion presented does little to inform on the risk that being male per se plays in vulnerability.
• The discussion of cardiovascular disease as a risk factor is generally appropriate.
• Under obstructive lung disease, I am a bit troubled by the comment that smokers who
already have a high COHb level may have little reserve for further increases in COHb
resulting from ambient sources. It may be more accurate to focus on the potential for
ambient CO exposure to reduce the rate of elimination of CO resulting from smoking.
Finally, the issue of establishing permissible exposure levels for ambient CO in a
subpopulation that self-exposes to far higher CO levels needs to be recognized.
71
------- |